China’s July retail sales numbers, released on August 17, remained weak. Total retail sales grew just 0.6% year over year, with goods sales up only 0.5%, while catering did somewhat better at 1.4%. This continued the sluggish trend from the first half of 2026, when traditional retail sales grew only 1.3%. Online consumption, however, remained relatively resilient, with online sales of goods and services up 4.8% in the first seven months.

But China’s “total retail sales of consumer goods” is actually a narrow measure, covering mainly physical goods and catering. Spending on services such as hotel stays, flights, entertainment, or gym memberships is largely excluded. This creates some odd distinctions: dinner at a hotel counts, but the hotel room itself does not.

The Ministry of Commerce has itself emphasized that consumption is shifting from goods toward a mix of goods and services. Despite good retail sales grew only 1.3%, while service retail sales grew 5.3%. Put the two together, and the broader measure of goods and services retail sales grew 2.7%. Tourism, culture, sports and leisure services all grew at double-digit rates.

I remain very bullish on Chinese consumption. China still has one of the world’s largest consumer bases, backed by a huge population and substantial household savings. But the opportunity is increasingly less about households simply owning more stuff.

By 2024, the average Chinese household already owned 1.51 air conditioners, 1.05 refrigerators and 0.99 washing machines. Those are already high levels of penetration. For comparison, around 89% of U.S. homes have air conditioning, while only about 20% of European households do. At this point, many basic durables have limited room for volume growth beyond replacement.

The next stage of consumption will be more about improving the quality and experience of life. That can mean services such as travel, hotels, entertainment and sports, but also physical products that offer better design, comfort, technology or emotional value.

For investors focused on China, understanding this shift is one of the keys to investment success over the next decade.

China Is Pushing Insurers Toward Dividend Stocks

China’s National Financial Regulatory Administration (NFRA) introduced an important new rule for insurers last week. On August 21, the NFRA issued the Notice on Strengthening the Asset-Liability Management of Insurance Companies. Starting in 2027, life insurers will be required to maintain a net investment income coverage ratio of at least 100%. Put simply, recurring investment income, mainly bond interest, dividends and rental income, should be sufficient to cover the guaranteed cost of their insurance liabilities.

This matters because interest rates in China are already very low. Insurers can extend bond duration to lock in today’s yields for longer, but the yields available on new bonds have fallen substantially. That makes dividend-paying equities more attractive. A stock that consistently pays a 5% dividend directly generates net investment income, while a non-dividend-paying stock that might appreciate 30% does much less to help insurers meet this requirement.

The new rule therefore provides a long-term tailwind for high-dividend stocks. One broker estimates that if the share of high-dividend assets in life insurers’ equity portfolios rises by 20 percentage points, it could translate into roughly RMB 770 billion of incremental allocation. I would not take that number too literally, but the direction is clear: insurers are being allowed to own more equities while being pushed to generate more stable investment income.

NetEase’s PDD Investment: A Capital Allocation Problem

NetEase delivered a strong Q2 2026, with operating profit up 33% year over year. However, most of the growth came from margin expansion rather than the top line, as revenue increased only 7%. The OPEX ratio fell by about 2 percentage points, while gross margin expanded by almost 7 percentage points.

In fact, gross-margin expansion has been underway for several years. NetEase’s overall gross margin jumped from 54.7% in 2022 to 60.9% in 2023, and continued to improve to 64.3% in 2025. The initial improvement was largely driven by a shift in the games mix toward higher-margin self-developed titles and away from lower-margin licensed games. In H1 2026, gross margin took another major step up to nearly 70%. Management attributed this mainly to lower platform revenue-sharing costs. My guess is that this partly reflects a higher mix of PC distribution, as well as Apple’s decision in March to cut its China App Store commission from 30% to 25%.

What I found more interesting was the RMB3 billion investment loss in Q2, shown above. This brings us to what I see as the biggest problem with NetEase: capital allocation and governance. In 2025, dividends and buybacks together amounted to only around 40% of net income. That looks overly conservative for a company sitting on so much excess cash. By the end of Q2 2026, NetEase had roughly RMB168 billion of net cash, including cash, time deposits and short-term investments.

Hoarding cash is one thing. Using it to invest in public equities is another. NetEase first disclosed a position of about 3.6 million PDD ADSs in mid-2025, then increased it to 8.9 million shares by Q1 2026 and 9.6 million by Q2. The investment has not worked well so far and was a major contributor to the investment losses discussed above.

The issue is not whether PDD will ultimately prove to be a good investment. Perhaps NetEase will eventually earn a remarkable return. The point is that this money belongs to all NetEase shareholders. If management has such strong conviction in PDD, it can invest its own money. Using shareholders’ capital to make a large discretionary bet on another public company raises a serious governance question.

It is like a PM side-pocketing investors’ money to make a high-conviction bet. Even if the bet eventually works out spectacularly and returns several times the money, the PM has still violated his/her fiduciary responsibility.

CMB’s NIM Lags, for Now

China’s major banks have now reported their first-half results, and one divergence stood out to me. China Construction Bank, for instance, revenue grew 10.5% year over year, helped by an 8.5% increase in net interest income. Its NIM reached 1.37%, up 3bps from 2025 and continuing to improve sequentially. CMB, by contrast, grew revenue by only 4.8%. Its Q2 NIM was 1.82%, still down 1bp sequentially. The divergence is clear by looking at their past 6 quarter YTD NIM:

This partly reflects very different liability structures. CMB has long been known as China’s “king of retail banking,” with a large base of cheap retail and demand deposits. That is a huge advantage in normal times, but it becomes less helpful deep into an easing cycle. When deposit costs are already close to the floor, there is simply less room for them to fall further, while loan yields can keep declining. CCB started with higher funding costs, so successive deposit-rate cuts have given it much more room to reprice liabilities downward. In H1, CCB’s deposit cost fell 29bps year over year to 1.11%, helping its NIM recover even as asset yields remained under pressure.

CMB also faces more pressure on the asset side. Its strength has historically been higher-yielding retail lending, but weak household credit demand and continued deleveraging have made those assets harder to grow. Large state-owned banks such as CCB have more room to deploy credit into infrastructure, technology and other policy-supported areas.

But CMB’s sensitivity to the economy and interest-rate cycle works both ways. The same balance-sheet structure and business mix that are pressuring NIM on the way down could provide greater upside when the cycle turns.

That’s all for now. See you next time.

H World reported another solid quarter. In China, blended RevPAR reached RMB238 in Q2, up 1.1% year over year. This marked the fourth consecutive quarter of positive RevPAR growth, although the pace slowed from Q1. The increase was driven by ADR, which rose 2.6%, helped by the continued upgrading of H World’s core brands. Occupancy, meanwhile, declined year over year, as it has throughout 2026 so far.

One number worth watching more closely is same-hotel RevPAR, which fell about 3% year over year in Q2. This continues a trend that began in 2024, with same-hotel RevPAR now declining for three consecutive years.

I think there are three main reasons:

1. A high comparison base. Travel demand rebounded sharply after China reopened, and same-hotel RevPAR jumped roughly 50% in 2023.

2. Pressure from H World’s own upgrades. The rapid rollout of newer and upgraded hotel formats inevitably puts some pressure on older properties with dated rooms and layouts. But I see this largely as part of H World’s upgrade cycle. Some cannibalization of older hotels is the price of improving the quality, ADR and RevPAR of the overall network.

3. A weak macro environment. Atour provides a useful cross-check: its same-hotel RevPAR also fell about 3% year over year in Q2, suggesting that at least part of the weakness is industry-wide rather than specific to H World.

H World’s international business looked weaker at first glance. RevPAR declined 3.8% year over year in Q2, driven mainly by a 3.5 percentage-point drop in occupancy. Management attributed the weakness partly to the conflict in the Middle East and partly to H World’s expansion into Southeast Asia, where newly opened hotels are still ramping up. Importantly, RevPAR in the European business was actually up 1.1% year over year.

Stepping back, I think H World has done a decent job internationally since acquiring Deutsche Hospitality at the end of 2019. From 2020 to 2025, international RevPAR increased (recovered) by roughly 163%, while the hotel network only expanded modestly. The business is still relatively asset-heavy, with leased&owned hotels accounting for slightly more than half of the international portfolio, so I still view H World as being in a learning and testing phase overseas.

As I mentioned in my previous deep dive, I think going overseas is the right long-term decision. Growth in China will eventually mature, and management needs to start building the next growth engine before that happens. As the Chinese saying goes, if you do not think about the long term, you will eventually run into problems in the near term.

The other key KPI for a hotel chain is network size. H World added a net 324 hotels during the quarter, with 500 gross openings, bringing its global network to 13,539 hotels. The company appears slightly behind pace for its full-year target of 2,200–2,300 gross openings. That said, franchisee demand remains strong: the pipeline reached 3,054 hotels, the first time it has exceeded 3,000 in two years.

Management also raised its revenue guidance, and the market reacted well. Since H World reported earnings on Monday, the shares have risen roughly 15%:

But I do not think H World’s revenue number is particularly meaningful. As I discussed in my previous deep dive, consolidated revenue mixes two very different businesses: leased-and-owned hotels and manachised-and-franchised hotels. Their economics are completely different, so changes in mix can move reported revenue quite a lot without saying much about the underlying business.

I would rather focus on two core KPIs: hotel count and RevPAR. I expect H World to end 2026 with around 14,500 hotels, keeping it on track toward its goal of 20,000 hotels by 2030. If the RevPAR trend from the first two quarters continues, full-year RevPAR could grow around 2%. Put the two together, I estimate that hotel turnover/GMV and EBIT could both grow in the mid-teens this year. Operating expenses have also been well controlled, so I do not see much concern there.

Thanks to the four-month pullback since April, even after the post-earnings jump, the stock is still trading at around $48 per share, or roughly 13x EV/LFY EBIT. With about a 5% shareholder yield from dividends and buybacks, plus mid-teens earnings growth, I think a high-teens annual return over the next few years is quite plausible.

DiDi 2026Q2: Early Signs of Margin Improvement in Brazil?

If you haven’t followed DiDi recently, its shares have roughly halved over the past year. The main reason is its renewed push in Latin America, especially Brazil. Starting around Q2–Q3 last year, DiDi stepped up investment there, using much of the cash generated by its China business to fund the expansion. The market obviously hated it.

In my previous update on DiDi, I mentioned that I actually understood, and probably even supported, its decision to restart heavy investment in Brazil. DiDi was not entering from scratch. Through 99, it already had a strong local operation and a leading position in ride-hailing, backed by an established network of drivers, users and local infrastructure. Food delivery is a natural adjacent business, so expanding from mobility into delivery made strategic sense to me.

Since then, the situation has actually become more favorable for DiDi. China’s food-delivery war escalated sharply in Q2 2025, when JD and Alibaba began pouring billions of renminbi into subsidies to attack Meituan’s core market. The impact was immediate: Meituan’s Q2 adjusted net profit fell 89% year over year. Probably related, Keeta’s rollout in Brazil later became bumpier as well, including a postponed Rio launch and layoffs. I suspect this only strengthened DiDi’s conviction that it should press its advantage, which helps explain the aggressive spending over the past few quarters.

DiDi reported its Q2 2026 results on August 13, and the shares jumped about 10% after the release. The reason was simple: Brazil is still losing money, but the losses did not widen even as international GTV, or gross bookings, grew around 60% year over year. In other words, EBITDA margin began to improve. More importantly, management said during the callback meeting that it is targeting profitability for the international business by 2028.

The China Mobility segment was solid as usual. GTV grew 8.1% year over year, although this marked the fourth consecutive quarter of sequential growth deceleration. That is probably not surprising given how mature DiDi’s market position already is in China. More impressive was the take rate, which reached a record 24%.

The thesis in my deep dive is clearly broken. I had assumed the international business would turn profitable relatively soon, while the China segment has played out broadly as expected. Yet astute readers may notice that the narrative around DiDi has shifted meaningfully.

When I wrote the deep dive, much of the debate was still about the quality of DiDi’s China business: whether it really had a moat, whether ride-hailing was fundamentally a good business, and how much regulatory risk remained. Today, the narrative is almost the opposite. DiDi China is increasingly seen as a high-quality, well-defended cash cow. The concern now is that all that cash is being poured into the international business.

That leaves DiDi in what I think is a very interesting, and potentially quite asymmetric, position. Whether the stock gets rerated largely comes down to whether the market is willing to value the company on a sum-of-the-parts basis. If I put a 12x EBITDA multiple on the China business and value the international business at 0.5x forward GTV, both of which I think are fairly conservative, DiDi would be worth more than 2x its current share price.

However, I am not sure this will be the last cash hole for DiDi. Uber recently said it expects to commit more than $10 billion over the coming years to autonomous vehicles and robotaxi deployment. Given what I have seen from DiDi’s management over the years, I would not be surprised if, at some point, it follows Uber and starts pouring significant capital into robotaxis as well.

When it comes to DiDi, there always seems to be another money pit around the corner.

Gree’s 10% Shareholder Yield Looks Attractive, but I Like Midea Better

On August 18, Gree announced that it had repurchased 20 million shares so far under the buyback program launched at the end of April. The program calls for RMB 5–10 billion of repurchases, and judging from Gree’s previous buybacks, it will probably end up near the upper end of that range. If it eventually spends the full RMB 10 billion, the shareholder return becomes quite remarkable. The stock already offers a rare 7.5% dividend yield, and adding the buyback would push total shareholder yield ~ 10%.

Of course, there are reasons the yield is so high. The entire Chinese home-appliance sector is cheap. The sector benefited from government trade-in subsidies last year, creating a relatively high comparison base, while Chinese consumption remains weak.

Gree is especially cheap because growth is hard to find. Domestically, despite years of trying to diversify, air conditioners still dominate the business, and the company has yet to build another meaningful product category. Overseas, it remains far behind Midea and Haier.

That is also why I am not a big fan of the company. I struggle to see where the next leg of growth comes from. I also have some reservations about Dong Mingzhu’s leadership. She has undoubtedly played an enormous role in building Gree, but her management style increasingly feels out of step with the times.

To be fair, Gree has historically treated shareholders quite well. But shareholder returns are also pretty good among several of its peers in the home-appliance sector. Midea, for example, has completed RMB 10 billion of repurchases in 2025 and announced another RMB 13 billion program for 2026, of which more than RMB 6 billion has already been completed. So far this year, Midea ranks first among all A-share companies by total buyback amount.

Including both dividends and buybacks, Midea is effectively returning close to 100% of its earnings to shareholders. At around 14x earnings, that implies roughly a 7% shareholder yield. If earnings can compound at around 10% over the long run and the valuation multiple stays roughly unchanged, that gets you to something like a 17% annualized return.

That’s all for now. See you next time.

Tencent reported its Q2 results on August 12. Its core businesses are still doing very well: Value-added Services revenue grew 8% year over year, FinTech and Business Services grew 9%, and Marketing Services revenue jumped 22%.

Advertising growth came from both higher impressions and higher eCPM. Video Accounts benefited from higher ad load, while AI-powered recommendation and targeting improved ad conversion and pricing. Tencent’s ad load still appears relatively restrained, so advertising should remain one of its more important growth drivers.

Within VAS, domestic games revenue grew a very robust 17% year over year. Gaming is a hit-driven business, so relatively steady growth requires a few blockbuster titles large enough to anchor the portfolio. Tencent has several of these, including Honor of Kings, Peacekeeper Elite and Delta Force, which serves as bedrock of its domestic games business.

Inside FinTech and Business Services, Tencent Cloud revenue grew in the low-20% range, driven partly by AI demand. Speaking of AI, I have to mention WorkBuddy, which seems to be everywhere lately—I see its ads all over the place.

And the usage numbers are impressive: WorkBuddy reached 21 million PC visits in June, more than the next two combined—ByteDance’s Trae and Alibaba’s QoderWork.

This is quite a turnaround for Tencent. A year ago, it was generally seen as lagging China’s AI leaders, with Hunyuan’s performance behind Alibaba’s Qwen and ByteDance’s Doubao. But Tencent has taken a clever route: rather than trying to win purely at the model layer, it did a great job at the product layer, closer to the end users—China’s tens of millions of office white-collar workers—and may now be leapfrogging its rivals there. WorkBuddy first took off during the “raising lobsters” OpenClaw craze in March, before Tencent increasingly repositioned it as a full-fledged office agent.

Today, WorkBuddy is integrated with products such as Tencent Docs and Tencent’s enterprise knowledge tools, and can handle tasks ranging from research and data analysis to documents, spreadsheets and presentations. Personally, I do not find the product itself especially groundbreaking—it feels broadly similar to Gemini inside Google Workspace, Copilot inside Microsoft Office, or other general-purpose AI work agents. But the important context is that this is China, where most users have had little direct access to those products.

The downside of all these developments on AI front is higher costs. Sales and marketing expenses rose 26% year over year, well ahead of the 11% revenue growth, partly due to heavier promotion of games and new AI products such as WorkBuddy. G&A expenses also increased 22%, reflecting higher R&D spending on model upgrades, AI deployment across WeChat and other initiatives.

The bigger number, however, was CapEx. Tencent spent RMB52.7 billion in the quarter, up 176% year over year, pushing reported free cash flow to negative RMB13.8 billion. Excluding prepayments for compute procurement, FCF would still have been a healthy RMB37.6 billion.

The market clearly does not love this level of spending (although last year she also hated when Tencent lagged on AI investment). You can see the concern throughout the earnings call: analysts’ questions ultimately came back to AI CapEx, inference costs and whether all this investment will generate an adequate return.

No one really knows how AI will ultimately play out, what the competitive landscape will look like, how these products will be monetized, or even whether they will earn attractive returns. For now, all of that remains an open question.

Management seems to be sending a reassuring message: if AI does not work out, Tencent can simply cut back the spending, while the GPUs it has already bought can potentially be rented out, redeployed, or even sold. In other words, they seem to view the downside as relatively limited.

Some long-term Tencent investors seem to look at the valuation in a similar way. Using 2025 as a reference, Tencent’s core business should be capable of generating at least ~RMB200 billion of annual FCF before the impact of its AI investments. Striping out roughly RMB1 trillion of net cash and equity investments, the core business trades at less than 15x EV/FCF, which looks pretty cheap for a business of Tencent’s quality.

However, I am a little skeptical of this way of looking at Tencent. Even if the new AI business were succesful, most likely it will fundamentally changes Tencent’s business model. Traditional internet companies such as Tencent were genuinely asset-light. Once the platform was built, growth required very little incremental CapEx. AI is different. It requires continuous spending on GPUs, data centers and inference capacity.

What if AI turns into another prolonged competitive war, much like China’s food-delivery battle, where everyone keeps spending because no one can afford to fall behind? In that case, elevated CapEx could last for years.

I am not saying Tencent’s management is making a mistake, or that they are not excellent operators. In Tencent’s position, they have to invest heavily in AI. But as investors, our job is not to sympathize with management, nor to cling to some old impression of a business (even a great one). It is to assess the risks objectively.

Looking Beneath Moutai’s Profit Decline

On August 14, Kweichow Moutai released its first-half 2026 results. Revenue reached RMB 92.3 billion, up just 1.3% year over year, while net profit fell 1.95% to RMB 44.5 billion. This was Moutai’s first year-over-year decline in first-half profit in a decade, naturally triggering plenty of pessimism among investors.

But we get a clearer picture by breaking the numbers down by product. Revenue from Moutai liquor, the company’s core product, still grew 2.8% year over year, while series liquor revenue fell 6.0%. In other words, even in what is clearly a downcycle for the baijiu industry, Moutai’s core product has held up reasonably well relative to its peers.

Moutai continued to push its consumer-oriented channel reform. Direct-sales revenue grew 29.9% year over year, with revenue from iMoutai jumping 274.2%, while wholesale revenue fell 21.6%. By my rough estimate, moving a bottle from wholesale to direct sales increases the revenue Moutai recognizes by roughly 20% per bottle. Given the scale of the channel shift, that alone should have added roughly 6% to revenue. Yet Moutai-liquor sales grew only 2.8%.

I think part of the reason is that the average selling price within the Moutai-liquor mix has come down. To make Moutai accessible to a broader group of consumers, iMoutai sold more standard Feitian Moutai, while the company capped prices and limited supply of some higher-priced non-standard products, including premium and zodiac editions. The roughly 1.5 percentage-point decline in Moutai-liquor gross margin is, I think, further evidence of this shift, and it also contributed to the decline in net profit.

Overall, I don’t think this half-year report was as bad as the headline numbers made it look. I have seen several arguments that Moutai is still expensive at around 20x earnings, which I disagree.

First, some investors compare today with the bottom of the previous baijiu cycle. Moutai traded at close to 10x earnings in 2013, so 20x today still looks expensive. But I don’t think this is an apples-to-apples comparison. China’s 10-year government bond yield reached nearly 4.7% in 2013, versus only around 1.7% today. Once we adjust for the much lower risk-free rate, Moutai’s valuation today is actually quite close to its historical trough.

Second, one might argue that plenty of high-quality HK-listed internet and TMT companies trade below 15x earnings, so why pay 20x for Moutai? But A-shares and Hong Kong stocks have historically traded in different valuation regimes. Capital controls and the segmentation between onshore and offshore investors have long contributed to higher multiples in the A-share market, so I don’t think this is a completely fair comparison either.

Finally, some may still argue that there are much cheaper companies even within A-shares. Major home-appliance makers such as Midea and Haier trade at 10–15x PE. That is true, but the quality of the businesses is simply very different. Moutai has a gross margin of around 90%, versus less than 30% for major appliance makers.

Bohai’s First 100-Bcm Gas Field

On August 9, CNOOC (883.HK) announced that Phase I of the Bozhong 19-6 condensate gas field had reached full production, with daily output exceeding 5,200 tonnes of oil equivalent, including around 3.5 million cubic meters of natural gas. Meanwhile, Phase II is already under development and is broadly similar in scale to Phase I.

Bozhong 19-6 has particular strategic value as the largest offshore gas field in eastern China. China’s natural-gas resources have historically been concentrated in the west, while much of the country’s demand is in the east. Located close to Beijing, Tianjin and the broader Bohai Rim, Bozhong 19-6 can therefore provide a more stable local gas supply to one of China’s largest population and industrial centers.

For readers unfamiliar with China National Offshore Oil Corporation(CNOOC), it is China’s largest offshore oil and gas producer. Its position is difficult to replicate: its state-owned parent has the exclusive right to partner with foreign companies in offshore China, giving CNOOC a privileged position in developing the country’s offshore resources.

Unlike China’s other large state-owned oil companies, such as PetroChina, CNOOC still enjoys a relatively long runway for production growth. China’s offshore oil resources are estimated to be only around 23% explored/proved, while offshore natural gas is closer to 7%. The opportunity, however, requires increasingly shifting deeper rather than simply coming from untouched shallow-water fields.

That is why projects such as Bozhong 19-6 matter. The reservoirs are buried more than 5,000 meters below the seabed, with temperatures reaching as high as 204°C and unusually hard and complex formations. CNOOC says it developed several new drilling and reservoir-management techniques to make large-scale development possible, including a gas-reinjection approach that helps maintain reservoir pressure.

Essentially, CNOOC is learning how to economically develop reservoirs that are deeper, hotter, higher-pressure and geologically more difficult—potentially opening up another layer of resources that was previously much harder to exploit.

Shan Weijian’s View on China

This week I watched a podcast interview with Weijian Shan, a legendary figure in private equity who is sometimes called China’s “private equity king.” Today, he serves as executive chairman of PAG, one of Asia’s largest alternative investment firms, with more than $55 billion in AUM. He also sits on the board of Alibaba.

Shan’s own life story is just as remarkable. He grew up during the Cultural Revolution, spent six years in the Gobi Desert, then went on to study economics at UC Berkeley under Janet Yellen before eventually becoming a professor at Wharton.

The interview was partly a promotion for the Chinese edition of Shan’s third book, Money Machine: A Trailblazing American Venture in China. His first book was about his life during the Cultural Revolution; his second told the story of how he led the negotiations to acquire Korea First Bank after the Asian Financial Crisis. (I am reading this one now, and I honestly cannot remember the last time I picked up a book and found it this hard to put down.)

In Money Machine, Shan, the ultimate insider and chief architect of the deal, tells the story of how an American buyout firm pulled off what seemed almost impossible: acquiring control of a deeply troubled Chinese national bank, with a non performing loan ratio above 20%, and turning it into one of the country’s healthiest and most profitable financial institutions. The investment reportedly returned more than 10x in five years.

runs for about an hour and a half, so I will just share one point that I found particularly insightful and agree with: why Shan is not pessimistic about the Chinese economy.

1. China still has enormous room for both monetary and fiscal policy. On the monetary side, China’s reserve requirement ratio, the share of deposits banks must keep at the central bank, is still around 6.5%, compared with roughly 0% in the U.S., 0.8% in Japan and 1% in Europe. Every one percentage point cut could release roughly RMB2 trillion into the banking system. There is still plenty of room to ease.

2. The Chinese government has an unusual strong balance sheet. The state owns enormous amounts of assets, and Shan argues that once these are counted, the government is in a net asset position, something rarely seen among major economies, except perhaps Russia. That gives China considerable fiscal capacity. His conclusion is simple: if the Chinese government sets a growth target and is determined to reach it, it has the policy tools to do so.

3. Chinese households have plenty of money to spend. Household bank deposits now total roughly RMB160 trillion, more than China’s annual GDP of around RMB140 trillion. To Shan, this suggests that the purchasing power has not disappeared; people have simply chosen to save rather than spend, despite very low deposit rates. If confidence returns and even 5% of those deposits flows back into consumption, that would be equivalent to roughly 5% of GDP.

After making his argument, Shan told a little joke, which I will use to end this Weekly Letter:

A shoe company sent a salesman to an island. He came back discouraged: “There’s no market—nobody there wears shoes.” So they sent another salesman to take a second look. He came back thrilled: “The opportunity is enormous—nobody there wears shoes.”

That’s all for now. See you next time.

The diary farming sector in China has been in a downturn since around August 2021, when raw-milk prices began falling from roughly RMB 4.4/kg to close to RMB 3.1/kg. The main problem was a classic supply-demand mismatch: farms expanded production during the previous upcycle, just as consumer demand for dairy products began to weaken. Milk prices eventually fell below the production costs of many farms, forcing smaller operators to either shut down or cull low-yield cows to conserve cash.

On July 28, Youran Dairy (9858.HK) issued a positive profit alert, expecting to make a profit in H1 2026—its first profit since 2022. The turnaround comes as China’s raw-milk market is finally showing signs of recovery.

Spot milk purchase prices in several major northern producing regions have rebounded by as much as 24% from their lows. China’s dairy-cow herd declined by around 550,000 head in 2025 and is expected to shrink by another 150,000–200,000 in 2026. Meanwhile, the amount of surplus raw milk processed into milk powder by leading dairy companies has fallen from around 20,000 tonnes per day at its April 2025 peak to just over 4,000 tonnes per day in June 2026—a clear sign that oversupply is easing.

What’s more, On July 20th, Modern Dairy announced it has acquired 80% of China Shengmu, China’s largest organic dairy farming company. Industry consolidation is a typical early sign of a new capital cycle.

Interestingly, the stocks moved well before the underlying milk market recovered (something I still find a bit puzzling). I bought Youran Dairy in late 2024 and sold after the shares had risen 3–4x:

Although I won’t buy Youran or Modern today, as I think the setup then was much more attractive than it is today, I think it’s worthwhile to jot down my thinking process as it’s a common pattern in cyclical investing.

A typical commodity company suffers both volume contraction and unit price decline during a downturn: therefore earnings fall, and the valuation multiple compresses. Interestingly, dairy farming’s earning gets an extra distortion, as milk cows are recorded as biological assets and regularly revalued based partly on assumptions about future raw-milk prices. As the result, back in late 2024, the stock was trading at 2x-3x pre-downturn earnings.

I had little doubt that milk prices would eventually recover, given China’s still-low dairy consumption compared with developed markets. And when the cycle turns, the mechanism I described above will work in reverse: revenue rises, margins improve due to both operating leverage and upward revaluation of biological assets, then multiple expands. The combination can create substantial upside.

The real question was whether Youran Dairy and Modern Dairy could survive the downturn, especially given their heavy debt loads. I was relatively comfortable for three reasons:

1. Economic of scale: together, their milk output is equivalent to roughly one-sixth of China’s national production.

2. Their advanced equipment, processes and vertically integrated feed and forage operations allow them to produce higher-quality milk at lower cost—and therefore sell it at a premium.

3. Most importantly, their parent companies are China’s two largest dairy-product producers (56% combined market share): Youran is controlled by Yili, while Modern Dairy is controlled by Mengniu. Both Youran and Modern are strategically important parts of their parent companies’ vertically integrated supply chains, with long-term fixed-price contract that offer some protection from depressed spot milk prices.

Luckin Coffee continues its aggressive expansion

Luckin Coffee reported its Q2 2026 results on August 3. Revenue rose 28.5% year over year to RMB15.9 billion, as the result of the company’s rapid store expansion. It added 2,714 net new stores during the quarter, bringing the total surpassed 36,000—nearly 17% higher than it had at the end of 2025. The market appeared pleased with the results, sending the shares up 5.7% that day. And on a static basis, the stock looks extremely cheap, trading at around 6x EV/EBIT if I simply annualize Q2 operating income.

The quarter showed little operating leverage though, with GAAP operating margin edged down from 14.1% to 13.4% year over year. One reason was higher sales and marketing spending, partly to promote Luckin’s bottled drinks and expand their distribution through more retail channels. Cost of materials also increased as a share of revenue because of fluctuations in raw-material prices, while store rental and other operating costs rose slightly as the store network expanded.

These increases were partly offset by a 3.1% year-over-year decline in delivery expenses. I thought this was due to the food-delivery price war between Meituan, Alibaba and JD cooling down, but the management attributed the decline to lower delivery costs per order from “enhanced fulfillment efficiency.”

Interestingly, management was much more willing to blame the decline in same-store sales growth (SSSG) on the food-delivery war. SSSG at self-operated stores fell 5.3%, marking the third consecutive quarter of decline, which management attributed to a tough comparison against last year, when delivery-platform subsidies created an unusually high base. But how can we tell that the weakness is not also a result of Luckin’s aggressive expansion and growing store density?

Using the estimate of 230k coffee shops nationwide, Luckin’s 36k stores already represent roughly 16% of the market—about one in every six locations. It is not obvious to me that the market can become much more concentrated, especially as competition continues to intensify. Mixue is pushing further into coffee through Lucky Cup; Guming now sells coffee through many of its tea shops; KCOFFEE has already opened more than 2,600 locations; and Wallace has also entered the market.

And that is before considering Starbucks. It recently transferred control of its China retail operations to a joint venture majority-owned by Boyu Capital (a firm known for its princeling-affiliation). The new structure is meant to make the business more localized and asset-light, while supporting an expansion from roughly 8,000 stores today to as many as 20,000 over time.

This brings me back to the nature of the business: coffee retail in China (perhaps in the world) may simply not be a great business.

Yum China Now Owns Pizza Hut in China

On August 7, Yum China completed its $1.2 billion acquisition of the Pizza Hut brand in mainland China from Yum! Brands, giving it full ownership of the brand rather than simply the exclusive right to operate it. The deal was part of Yum! Brands’ broader divestiture of Pizza Hut, with the rest of the global business being sold to private-equity firm LongRange Capital.

Pizza Hut China paid about $69 million in franchise fees to Yum! Brands in 2025, so the purchase price is roughly 17x last year’s fee—a multiple in line with where Yum China itself trades. The buyout also feels like a sign that Yum China is serious about the turnaround it has already been pursuing at Pizza Hut, after years of its underperformance.

Pizza Hut opened its first store in Beijing in 1990, the same year I was born. So it is not much of an exaggeration to say that Pizza Hut and my generation grew up together in China.I still remember when I was a kid, Pizza Hut was widely seen as an upscale restaurant. For many Chinese, it was their first introduction to pizza, which felt novel and somewhat fancy at the time. Stores were often packed, sometimes with long lines outside.

Sadly, Pizza Hut gradually lost its appeal, and today the brand simply feels much less relevant. One reason was clearly slow product innovation. When Domino’s China started rolling out new flavors and products aggressively, Pizza Hut was slow to respond, and only more recently has it begun to try to catch up.

Service also became less consistent. More importantly, I think Pizza Hut has struggled with its positioning: it still wants to preserve some of its old dine-in, premium image, but at the same time it has tried buffets to compete for more price-conscious customers and delivery to mimic Domino. The result is a brand that can sometimes feel caught in between.

This is exactly the positioning issue management has been trying to fix lately. The direction is now somewhat clearer: Pizza Hut is being pushed toward a stronger value-for-money proposition, with lower-cost WOW stores helping it expand into lower-tier cities and a greater use of franchising.

The recent numbers suggest the strategy is gaining some traction: in Q1 2026, same-store transactions grew 5% year over year, marking the 13th consecutive quarter of transaction growth, while operating margin expanded year over year for the eighth consecutive quarter. Whether the turnaround will ultimately succeed is still too early to say.

That’s all for now. See you next time.

It was reported that DeepSeek has told prospective investors that it is putting its second fundraising round on hold for now. The news came just days after a four-hour recording of Liang Wenfeng speaking with investors was leaked and went viral. The transcript has since been removed from the Chinese internet.

I picked out ten points from the transcript that you might find most interesting:

1. Liang does not see DeepSeek as a company whose goal is to maximize profits. He wants it to do something useful for humanity and believes the company should operate with genuine goodwill toward the world.

2. DeepSeek is not managed through KPIs or layers of performance reviews. What holds the team together is a shared vision. Liang admires former GE CEO Jack Welch and believes that large organizations are ultimately driven by vision, not rules.

3. At DeepSeek, top-down assignments are not supposed to take up more than half of an employee’s time. The rest is deliberately left open for researchers to explore ideas and pursue projects of their own choosing. (remind me of Google)

4. Liang believes DeepSeek’s most important task is to keep its core team together. As long as the key people stay, he believes the company can keep moving forward and eventually achieve AGI.

5. From a business perspective, Liang believes open-sourcing AI can actually be beneficial. He does not see open source and commercial monetization as being in conflict.

6. DeepSeek has only one real priority: AGI. Consumer traffic, enterprise revenue, multimodal products etc., are secondary.

7. Liang believes the next major breakthrough after agents will be continual learning. After that, models will begin to improve themselves continuously, eventually leading toward embodied AI.

8. In his view, the only meaningful gap between China and the US is not talent, but resources, especially compute. DeepSeek’s current approach is simple: buy as many chips as it can at reasonable prices.

9. Domestic AI chips have a historic opportunity because AI is rapidly weakening the moat created by Nvidia’s CUDA ecosystem.

10. Liang believes Chinese AI chips are already competitive in both hardware and software ecosystems. The main constraint is production capacity, and he does not expect China to still be stuck on this problem five years from now.

DeepSeek was highly selective about who it allowed into the 1st round, focusing on investors whose long-term interests were aligned with the company’s. Only a small group of institutions were invited to participate, and the investment came with a number of restrictions.

According to reports, Liang Wenfeng was the largest investor, committing around RMB 20 billion of his own money. Tencent invested about RMB 10 billion, CATL-related entities around RMB 5 billion, while NetEase, JD.com, Monolith Capital and IDG Capital each contributed roughly RMB 3 billion. China’s national AI industry investment fund invested another RMB 980 million.

Laopu Gold’s positive profit alert turned out to be not so positive

On July 27, Laopu Gold (6181.HK) issued a positive profit alert, forecasting first-half 2026 revenue of RMB19.8–20.4 billion, up 60%–66% yoy, and adjusted net profit of RMB4.31–4.36 billion, up 83%–85%. The numbers looked spectacular at first glance. Yet the shares plunged nearly 24% the following day.

The reason becomes clearer when we compare the guidance with the preliminary Q1 figures released in March. Laopu had expected Q1 revenue of RMB16.5–17.5 billion and net profit of RMB3.6–3.8 billion. Taken together, the two announcements imply Q2 revenue of only RMB2.3–3.95 billion, down 76%–87% from the first quarter. Q2 profit is estimated at roughly RMB0.5–0.8 billion, down more than 80% QoQ and around 34%–56% YoY.

For those unfamiliar with Laopu Gold, it is a Chinese luxury jewellery brand specializing in “heritage gold”—gold pieces made using traditional Chinese craftsmanship and designs inspired by classical Chinese culture.

It emerged as one of China’s hottest homegrown luxury brands, with products ranging from several thousand to several hundred thousand yuan. Together with Pop Mart (9992.HK), Mixue (2097.HK) and Mao Geping (1318.HK), it was once called Hong Kong’s “New Consumption F4.” All four stocks have now fallen by more than half from their peaks.

I do think the sell-off of the F4 may have created some value opportunities. But Laopu Gold is probably not one of them.

The problem is quite obvious if we look at its cash flow:

Net income has surged, yet operating cash flow has become increasingly negative. A closer look shows why: nearly all the cash generated by the business is being absorbed by inventory. Part of the increase is due to much higher Gold price, so replenishing the same amount of inventory now costs more. But part of it is also a deliberate growth bet: management is expanding rapidly and building inventory ahead of the stronger sales it expects in the future.

If nearly all of Laopu’s internally generated cash is tied up in inventory, how is it funding the rest of its expansion? The answer is debt and dilution:

Many investors would point to Laopu’s brand value, luxury positioning and distinctive heritage-gold craftsmanship. But I would say it’s hard to separate Laopu’s lately extraordinary growth from the surge in gold prices and the resulting enthusiasm for gold. If the reverse of gold prices is permanent, the company could face a double hit: weaker demand leads to excess inventory, while falling gold prices could also force it to write down part of that inventory. Seen this way, Laopu Gold starts to look less like a luxury brand and more like a leveraged gold ETF.

Inside Pop Mart’s New Toronto Store

After Laopu Gold, let’s turn to another member of the “new consumer F4”: Pop Mart. The company finally opened a store in Toronto’s Eaton Centre, so I stopped by last weekend. For now, it is only a pop-up. An employee told me that it will eventually move into a much larger—and presumably better-located—space in the mall.

The traffic is pretty good and I saw several customers practicing what collectors call “box shaking”: gently shaking and weighing blind boxes to guess which figure is inside before buying.

I also spoke with two employees, both of whom were Pop Mart fans themselves—one preferred Labubu, while the other was a Hirono fan. You could tell that they genuinely loved the characters and the broad collecting culture. One of them told me that Pop Mart sells art, not toys — an idea repeatedly emphasized by the founder & CEO. Either Pop Mart trains its staff incredibly well, or designer toys really are becoming a new culture and trend—not just the fad many people think they are.

I’m probably a little too old to fully get the new culture:) But even I could see that Pop Mart’s figures were in a different league from the other character toys sold elsewhere in the same mall. The materials, detailing and overall finish were noticeably better.

The displays, layout and lighting in Pop Mart’s store is also another level. Other stores like Miniso are just retailers crowded with merchandise, while Pop Mart felt more like a small art exhibition, with each IP carefully displayed and the whole space designed around the experience.

Moutai raises prices again

On July 17, Kweichow Moutai announced its second price increase of the year for a 500ml bottle of 53-degree Flying Fairy Moutai. The retail price on the iMoutai platform rose from RMB 1,539 to RMB 1,639, while the ex-factory price—the price paid by contracted distributors—increased from RMB 1,269 to RMB 1,369. A few days later, reports emerged that Moutai’s offline self-operated stores had raised their price further to RMB 1,719.

This follows the price increase in March, when Moutai raised its self-operated retail price from RMB 1,499 to RMB 1,539, finally ending the eight-year “RMB 1,499 era.”

On the one hand, the price increases are part of a broader effort by Moutai to regain control over pricing and push distributors to earn their margins through actual service rather than scarcity alone.

On the other hand, raising prices twice during the traditionally weak season—and in the middle of a broader downturn across the baijiu industry—also shows the strength and scarcity of the brand. Bottles on iMoutai remain extremely difficult to secure even after the price increases.

The common investor view of Chinese baijiu companies is that they have excellent business models but often mediocre management and governance, partly because many of the industry leaders are state-owned. In my view, Moutai is the exception: it combines an extraordinarily rare business model with a strong corporate culture built around always trying to do the right thing.

Small Merchants Force Xufuji’s Pinduoduo Store to Shut Down

The official flagship store of Xufuji, a Chinese snack-food giant with nearly RMB 10 billion in annual revenue, recently came under a coordinated attack by small merchants on PDD. They placed a large number of orders and then applied for “refund without return,” overwhelming the store’s customer-service and delivery systems. The operator had to intercept parcels already in transit, cancel unshipped orders and eventually shut the store. On July 27, Xufuji clarified that the store was run by an authorized distributor, not by PDD itself.

The merchants’ anger has been building for years. PDD’s “refund without return” policy has caused real losses for many sellers, as some customers claim refunds on perfectly good products without sending them back. Merchants also worry that PDD may increasingly direct traffic toward its own retail operations.

A similar episode happened in 2023. Merchants first targeted PDD’s newly launched self-operated store, then moved on to stores linked to brands such as Xufuji and Nestlé. This time, they even dug up a list of Tmall stores operated by companies connected to Leqee, an e-commerce service business founded by Colin Huang, and used it to identify suspected PDD-operated stores. Many of the stores they targeted, however, were actually run by independent distributors.

“Refund without return” means buyers can receive a refund without sending the product back. In one recent case, a buyer ordered 1,100 pieces of clothing and obtained refunds on 1,000 of them, leaving the merchant with a reported loss of RMB 52,000. Sellers also complain that they have little room to appeal, turning what began as a consumer-protection policy into an easy way to game the system.

The policy has since been rolled back across the industry. In 2025, PDD, Taobao, JD.com, Douyin and Kuaishou all changed their rules so that post-delivery requests would generally be handled by merchants rather than automatically approved by the platform. A new regulation effective from February 2026 also prohibited platforms from forcing merchants to provide refunds without returns.

I believe PDD’s original intention was good: “refund without return” was meant to push merchants to take product quality more seriously, especially given the platform’s long-standing reputation for cheap but low-quality goods. However , the policy became an easy loophole for fraudulent refund claims. This may be an inherent dilemma for a platform focused on low-priced products: disputes between buyers and sellers are more likely, but given the low average order value, it often makes little economic sense for the platform to investigate and mediate every case.

That’s all for now. See you next time.

The plan is to send out a short letter on Sunday evening (ET) whenever I have not published a long write-up that week, covering some of the topics that have recently drawn attention, both within China’s investment community and in society more broadly, along with the latest developments at the companies I follow and, of course, my thoughts on them.

These letters will inevitably be biased toward my own interests and limited by the information I have access to, but I hope readers will still find them helpful in their own investing in China.

Anti-trust investigation on Trips.com

We finally received an update on the antitrust investigation into Trip.com, which was formally launched in January over the company’s suspected abuse of its dominant market position. The regulator concluded that, since 2020, Trip.com had used traffic allocation, platform rules and pricing tools to pressure hotels into exclusive arrangements and require them to offer the lowest prices on its platform, restricting their ability to work with competitors and set prices independently.

Trip.com was fined RMB 3.52 billion, or 7.5% of its 2025 domestic revenue, while another RMB 1.66 billion of illegal gains was confiscated. Together with RMB 122 million of deposits that it must return to hotel operators, the total financial impact is approximately RMB 5.30 billion. This is a fairly heavy penalty—roughly twice the amount I estimated in my previous post based on past case of Alibaba and Meituan.

Trip.com posted a PR right after the announcement, accepting the ruling and announced 19 corrective measures. Most importantly, it will end its “special” exclusive-distribution model, remove the “lowest price across the internet” requirement, discontinue the related merchant-ranking and traffic arrangements, and shut down tools that allowed the platform or its staff to adjust hotel prices without explicit approval. It also promised to introduce a more transparent commission and traffic-allocation system, stop forcing hotels to participate in promotions and strengthen its antitrust and consumer-protection controls.

As I argued in my previous post, the size of the fine is not particularly important. It is ultimately a one-time cost and should not materially affect Trip.com’s long-term value. More importantly, the worst-case outcome did not happen: the government did not force Trip.com to break up or fundamentally restructure its business in any way. That is clearly good news for shareholders.

The more difficult question is how the 19 corrective measures will affect Trip.com’s moat. Some of the practices targeted by the regulator—such as exclusive arrangements, lowest-price requirements and the use of traffic allocation to influence hotels—were unfortunately signatures of the platform’s competitive advantages. Perhaps more importantly, before assessing their impact, investors first need to understand how seriously Trip.com will implement these measures and how strictly the government will enforce them going forward.

These questions may not be easy to answer, and it could take time to get some clarity. Given this uncertainty, investors looking to gain exposure to China’s travel sector may find a simpler opportunity elsewhere, such as in the hotel chains I have written about before.

CXMT’s Blockbuster IPO

CXMT will list on Shanghai’s STAR Market on July 27. Its RMB 57.9 billion offering is the largest IPO in the board’s history.

The allotment results were announced this week: around 7.7 million winning numbers were issued, with each winner entitled to purchase 500 shares. The final online winning rate was about 0.47%, reportedly far higher than that of a typical STAR Market IPO. Combined with the relatively modest cash requirement of RMB 4,330 per winning lot, thanks for an IPO price of RMB 8.66/share, the offering has been extremely popular among retail investors. Indeed, almost every investor I know who was eligible to participate applied for the shares.

As China’s leading DRAM producer and the world’s fourth-largest supplier, CXMT is one of the country’s most strategically important semiconductor companies. Unsurprisingly, there is no lack of institutional participation in the IPO.

A total of 113 private fund managers secured allocations worth more than RMB 13 billion. Funds associated with Liang Wenfeng’s High-Flyer (founder and backer of DeepSeek) were among the largest participants, receiving around RMB 175 million in total. Another 36 insurance institutions were allocated approximately RMB 6.1 billion, or more than 10% of the offering. This includes long-term investors such as the National Social Security Fund and several state-backed funds.

Tencent fell by 6%

Tencent gave back almost all its gains from the previous three weeks, falling about 6% last week. There were few attempts to explain the sell off:

Firstly, Investors concerned about Tencent’s Q2 gaming performance. A Bernstein report cited Sensor Tower data showing that Tencent’s domestic mobile game receipts fell 2.6% year over year in the second quarter, with its top three games down 13%.

However, Sensor Tower date cover only iOS revenue in China, excluding Android, PC and oversea contributions, so their significance is limited. More importantly, quarterly fluctuations are perfectly normal in gaming and tell us little about the long-term health of the business.

Secondly, Chinese mutual fund holdings for Q2 was released last week. Tencent was the most heavily sold stock, with active equity funds reducing their holdings by roughly RMB 20 billion. Much of that capital have been redirected toward AI stocks, as mutual fund exposure to the electronics and communications sectors increased by more than 20% QoQ.

If this is really the reason behind Tencent’s sell off, Chinese markets are really a paradise for value investors. We all know the institutional imperative: facing redemption pressure and short-term performance comparisons, fund managers are structurally unable to be contrarians. It reminds me of the situation at Fundsmith.

Tencent certainly looks cheap on a static valuation basis, but I believe the key to Tencent’s investment today is to understand how AI will affect Tencent’s moat and earning power—a much more complicated topic that deserves a separate discussion.

Boss Zhipin’s World Cup ad campaign

The World Cup wrapped up last week. As readers of my deep dive may remember, BOSS Zhipin’s aggressive advertising campaign during the 2018 World Cup was a inflection point for the company. The campaign sharply increased its visibility, and user growth accelerated thereafter.

Its approach to advertising reminds me of Duan Yongping. Much of BBK’s early success was supported by expensive, high-profile campaigns, including prime advertising slots during the Spring Festival Gala and celebrity endorsements from Jackie Chan.

And both Zhao and Duan shared a similar logic: the absolute spending may look enormous, but when an advertisement reaches hundreds of millions of people through a national or global event, the cost per viewer can actually be extremely low, and consequently the ROI is very attractive.

BOSS Zhipin continued to advertise heavily during this year’s World Cup, but the message was noticeably different. Rather than repeatedly explaining what the platform does or urging people to find a job, its ads featuring Shen Teng, one of China’s most popular comedic actors, encouraged viewers to slow down, take a break and enjoy the match. I thought it was a creative way to keep the brand visible without selling too hard:

Two Chinese mathematicians won the Fields Metal

This week, Hong Wang, a professor at NYU and IHES, and Yu Deng, a professor at the University of Chicago, became two of the four winners of the 2026 Fields Medal. For those unfamiliar with the award, it is often described as the Nobel Prize of mathematics and is presented every four years to outstanding mathematicians under the age of 40.

Wang and Deng are the first Chinese nationals to receive the medal—and China got two winners at once. Wang is also the third woman to win the award in its 90-year history. Unsurprisingly, the news was a big deal for the Chinese public and quickly became a trending topic.

For anyone interested in their personal journeys and the mathematical problems they solved, I highly recommend Quanta Magazine’s profiles of Yu Deng and Hong Wang.

As a former researcher, I found their stories particularly relatable: the self-doubt, the ambition to solve important problems, and the pressure to publish in top-tier journals. Despite their talent, both followed uncertain and indirect paths.

Yu Deng, for example, failed to qualify for a national Go tournament in middle school, which led him to turn his attention to mathematics. After completing his postdoc at NYU, he received no tenure-track offers from his preferred institutions and came close to taking a finance job in China.

Hong Wang did not secure the single place available in Peking University’s mathematics program, so she initially enrolled in earth sciences before later switching to mathematics. While studying in France, she doubted whether she was good enough to pursue mathematical research. At one point, she switched to architecture and spent a semester interning at a firm in Paris.

What kept them going through these setbacks? The first was a pure love of mathematics. Both had enjoyed solving mathematical problems from a young age, and whenever they faced uncertainty or disappointment, they could return to the work itself and focus on the next problem.

The second was the support they received throughout their lives—from parents who encouraged their interests without placing too much pressure on them, to mentors and collaborators who gave them the patience, confidence and support to continue through difficult periods.

Their journeys show that talent matters, but so does an environment that gives people the freedom and support to pursue their passion and allows it to flourish.

Oh, and before I wrap it up: the other winner Jacob Tsimerman, a professor at UofT, announced he is joining OpenAI:)

I was in Engelberg last week for Rob’s meeting. I really enjoyed my trip to Switzerland—the snow, the mountains, and the chance to meet investors from around the world. It’s amazing how a shared passion for investing can bring together people across professions, cultures, and life paths.

As my struggle to find time for deep dives continues, I thought it’d be nice to write a shorter piece on the antitrust investigation into Trip.com Group. I’ll provide some background and lay out three possible outcome scenarios.

Some investors also worried that PDD could face a similar investigation, partly as retaliation for the fistfight last December. While I think an antitrust probe into PDD is unlikely, I found the mystery around the fistfight intriguing enough to dig deeper into what’s been circulating online. Needless to say, much of this is unverified—so take it with a grain of salt.

One quick note up front: you won’t find any investment recommendation in this post. I don’t own shares of Trip.com or PDD, and I don’t claim to be an expert on either business.

So what’s the takeaway then, beyond the fascinating stories themselves? I hope you walk away with a better sense of the complex waters that both Chinese companies and the government are navigating—let alone investors.

Antitrust investigation into Trip.com

On January 14, 2026, the Chinese government officially launched an antitrust investigation into Trip.com. It’s reported that the State Administration for Market Regulation (SAMR) have already stationed personnel at the company’s headquarter in Shanghai. Staff have been asked to surrender their computers and mobile phones and cooperate with the investigation.

The news quickly became a trending topic across the country, and Trip.com’s ADRs nosedived 17% the same day.

Background

Trip.com is China’s largest online travel agency (OTA), with roughly 56% market share measured by GMV. It operates two major apps with somewhat different customer profiles: Ctrip, Qunar. Including its ~20% stake in Tongcheng, Trip.com effectively controls close to 70% of China’s OTA market. The next largest players are Meituan and Fliggy, with market shares of about 13% and 8%, respectively.

Trip.com makes most of its money from commissions—roughly half from hotel bookings and half from transportation tickets. Yet this investigation targets only the hotel side. That’s because China’s hotel supply is still fragmented (chain penetration is just above 40%), leaving hotel owners with limited bargaining power. Ticket suppliers, on the other hand, such as airlines and railroads, are concentrated and often state-owned.

This weaker bargaining power shows up in the much higher take-rates on hotel bookings (about 10–15%) compared with transportation tickets (around 1%-2%). Trip.com’s contracts with hotel suppliers generally fall into three tiers:

1. Standard. 10% commission. No additional traffic support from the platform.

2. Preferred. 12% commission. Better exposure (shown earlier than comparable listings), in exchange for a guaranteed 5% lower price than on any other platform.

3. Exclusive. 15% commission. Even more exposure, in exchange for sole listing on Trip.com.

Both the exclusive requirement in Tier 3 and the price-parity clause in Tier 2 are obvious antitrust red flags. Trip.com enforces price parity through an algorithm called “Pricing Assistant”, which constantly scans rival platforms for the same room. Once it finds a lower price, Trip.com’s backend automatically cuts the price on its own listing—without the hotel’s consent.

The warning signs started flashing last year. In August 2025, the Guizhou Provincial Administration for Market Regulation called Trip.com in for a regulatory meeting over alleged exclusive dealing and the use of technical measures to interfere with pricing. A few months later, in November 2025, the Yunnan Provincial Tourism Homestay Industry Association issued a rights-protection statement, making similar accusations.

Hotel owners blame Trip.com for their financial struggles. Meanwhile, many users complain about algorithmic price discrimination. Basically, the more time you spend in the app, the more expensive the tickets become. No wonder everyone seems to be cheering—except Trip.com’s investors.

Some misleading media coverage didn’t help either. Trip.com was frequently described as the most profitable internet company, based on its Q3 2025 net income of RMB 19.9B—even though RMB 14B of that came from the sale of its India business.

Impact of the investigation

Some investors worried this is 2021 all over again, which I think is far-stretched. In fact, Booking.com has been facing antitrust investigation across Europe over similar issues: Price Parity Clauses, Preferential Partner Programs, etc. Even the ‘pricing assistant’ is not something invented by Trip.com - If Booking detects better prices elsewhere, it unilaterally applies a discount, called Booking Sponsored Benefit, to align prices on its platform.

In my view, the Chinese government is simply doing its job, much like regulators in Europe.

How about the impact on Trip.com’s business? Will the investigation end up weaken Trip.com’s market position and profitability? The hones answer here is too soon to know. Let’s consider three scenarios:

Bull case
Trip.com only has to write a big check. Based on past cases involving Alibaba and Meituan, antitrust fines are typically around 3–4% of the prior year’s revenue. That would imply roughly RMB 1.8B, or about 10% of Trip.com’s 2024 net income. The exact number doesn’t matter much—it’s a one-time expense.

Under this scenario, the ~20% share price drop could be a bargain opportunity. That said, I think this outcome is unlikely. What’s the point of a fine if it doesn’t come with corrective action?

Base case
Trip.com ends up paying a fine and is forced to drop both its exclusive arrangements and price-parity clauses. That would certainly affect both revenue and margins. Without these terms, Trip.com has to match the lower commissions offered by competitors, and some price-sensitive customers could shift to cheaper alternatives.

On the other hand, I’m sure Trip.com’s moat is more than just bullying hotel owners. After all, these practices are common across all OTAs. Trip.com has a sticky, high-income user base that values service quality over price. And its transportation-ticketing business should see minimal impact. Finally, we shouldn’t ignore its rapidly growing overseas business.

That said, if there’s one thing we’ve learned from investing in China over the past few years, it’s that the moat around platform businesses is often weaker than we assume—and competition is far more intense than expected. It’s not just Meituan, Fliggy, and Douyin eyeing Trip.com’s market share; even JD is moving into the space.

Bear case
The worst-case scenario is that Trip.com is forced to divest Qunar, and its stake in Tongcheng. If that happens, Trip.com could be left with only around 20% market share. And if these platforms end up competing head-to-head, the industry could turn ugly very quickly. This would be unprecedented—but nothing is impossible.

My point is simple: the prudent move, in my view, is to wait for the government’s official penalty decision. Based on past cases, that could mean a 6–9 month window (about 4 months for Alibaba, and 18 months for Meituan).

And even if the final decision ends up aligning with our base-case scenario, I would still not lean on theories to make my investment decision. I would closely monitor a set of KPIs to judge the impact on Trip.com’s business.

The fog around PDD

The Trip.com investigation has also made some investors nervous that PDD could be next—there are whispers that regulators showed up again at PDD’s headquarters earlier this week. I think that’s unlikely for one simple reason: unlike Alibaba, Meituan, and Trip.com, PDD doesn’t control anything close to 50% market share. In other words, it doesn’t even meet the basic threshold to be treated as a dominant player under the antitrust law.

The investigation is likely related to issues such as counterfeit products, food safety, fake shipments and fake orders. That could lead to a fine, but I don’t see it having a material impact on PDD’s business.

Now, why do I say “again”? In case you missed it, back in early December 2025, it was widely reported that PDD executives and officials from the State Administration for Market Regulation (SAMR) got into a fistfight on Dec 3rd. Almost immediately, reports and social media posts mentioning the incident were scrubbed from China’s internet.

Some investors worried that PDD might face retaliation from the government. But there has to be more going on beneath the surface. If PDD had truly crossed the government in a serious way, why wasn’t it publicly condemned—instead of being quietly shielded as the story disappeared? And it’s hard to believe senior management would be reckless enough to let things escalate into a physical confrontation with government officials. Too many pieces don’t add up.

In mid-December, a petition letter from a former PDD employee, one of 30 some who were reportedly fired, was leaked online. Combined with several posts from alleged insiders, it gave me a somewhat clearer picture of what may have happened, even though we’ll probably never know the full story. If you’re interested, read on—but take it with a grain of salt.

Why was the story scrubbed so quickly

Yes—there’s something called government relations. In practice, it functions a lot like lobbying in the Western world: almost every major internet company in China hires former government officials to help navigate regulation, manage communication, and reduce political risk. Of course, companies prefer it to remain in the shadow.

In PDD’s case, someone circulated a list online claiming to identify a dozen C-level executives with government backgrounds. The roles cited include a former division chief at the Public Opinion Bureau of the Central Propaganda Department, a former director of the Political Department at the Shanghai Intermediate People’s Court, a former deputy division chief at the Shanghai Administration for Market Regulation, among others.

With that context, the censorship around this incident feels less baffling. It’s not as simple as “PDD vs. the government.” The Chinese state isn’t a single monolith—there are different agencies, interests, and internal dynamics at play. PDD likely leveraged its government connections to keep the story under wraps.

Why the fistfight?

This also suggests that on Dec 3rd, the two sides involved likely knew each other well—perhaps they had even worked together before, back when they were both in government.

There is more background story: One of PDD’s GR staff, WX, previously served as a deputy inspector in the Advertising Department of the State Administration for Market Regulation. According to post from insiders, WX left PDD around mid-2025 and quickly joined Rednote, apparently without much regard for the non-compete he signed.

PDD then launched a series of investigations into WX, including surveillance, pulling bank records, and tracking his delivery/package information. Unsurprisingly, this infuriated WX. PDD’s management believed the December 3rd investigation was retaliation from WX.

Still, even if PDD hold low regard to the regulators on site that day—given the government backgrounds of some of its senior staff—and even if management believed the December 3 visit was retaliation, it still can’t really explain why they would think escalation was a good idea. You don’t need to be an executive to know that was obstruction of official duties.

The explanation from post by insiders (with specific details including the names of people supposedly involved) is that management wanted to shut the investigation down as quickly as possible to protect PDD’s interests. So they deliberately provoked the regulators in the hope of triggering a physical conflict. Once the police got involved, the investigation would have to pause.

I don’t know. It feels like I’m reading a magical realism novel. But if this is true, I wonder what was PDD trying to hide that made them take such extreme measures.

Timeline

I’ll close with a timeline based on information gathered online:

• Dec 3rd 2025. SAMR officials arrived at PDD’s Shanghai headquarters for a 3–4 day on-site inspection. It’s not clear what the investigation was focused on, as different reports cited different issues—ranging from food safety to fake orders.

• Dec 4th. A physical conflict broke out between PDD executives and SAMR officials. The SAMR team lead fractured his thumb. Police from Shanghai’s Changning District arrived and initially suggested resolving the matter privately, because people on both sides knew each other.

  The SAMR side found the proposal unappealing, so the police issued a five-day administrative detention to the person identified as the lead participant in the fight. Higher-level regulators were still unhappy and escalated the matter to the State Council.

• Dec 5th. The Ministry of Public Security issued a directive to the Shanghai Municipal Public Security Bureau, demanding that the incident be handled seriously. The Shanghai police then assigned two patrol cars to provide 24/7 protection for the SAMR team until they completed their inspection on December 6. Meanwhile, the case was filed by the Shanghai police as obstruction of official duties.

• Dec 10th. PDD reportedly fired around 30 employees across its government relations and compliance teams, including co-founder Fan. One of the dismissed employees, Tu, who received no severance and lost his year-end bonus claimed that he was in Beijing at the time of the altercation and was made a scapegoat. His petition letter to SAMR was later leaked online.

• Dec 19th. PDD held its AGM and announced a series of senior management changes. Newly promoted co-CEO Zhao Jiazhen said the company plans to “build another PDD in the next three years.”

The author of this write-up do not owns shares in the companies mentioned and may purchase or sell shares without notice. This write-up represents only the author’s personal opinions and is not a recommendation to buy or sell a security. No information presented in the write-up is designed to be timely and accurate and should be used only for informational purposes. Readers of the write-up should perform their own due diligence before making investment decisions.

So why write another piece on H World? Two reasons.

First, looking back, that old write-up was missing a few important pieces. I didn’t explain why a strong loyalty program matters; I didn’t build a proper model or valuation; and my growth projection was too rough and not rigorous enough. A recent conversation with a new subscriber (thanks for reaching out!) reminded me of these gaps. This write-up is my attempt to fill them.

Second, hotel chain industry is one of the most defensive business out there. This feels especially relevant today, amid the ongoing instant-delivery war in China. It’s such a popular topic that everyone is trying to analyze (guess) who will win. But the truth is that the problem itself may simply be too hard—or even impossible—to solve.

In my view, rather than trying to solve an impossible problem, we should focus on the lessons it offers so we can stay vigilant when the next one comes — or avoid it altogether.

A big driver of this war is that the old Big Two are now under siege, as shown by the market share evolution in e-commerce:

I learnt two things from this figure.

1. The fragmentation of the market shows that the moats of China’s major e-commerce players are weaker than many once believed. Few would have expected today’s landscape back in 2017, when most thought Alibaba and JD would settle into a stable duopoly. (I recently learnt an anecdote that Jack Ma back then actually believed JD survived only because he allowed it to.)

2. While COVID or the CPC often get blamed for the collapse in e-commerce stock prices, the market-share trends from 2017 to 2019 were already showing early signs of trouble — a reminder of why tracking quantitative metrics matters.

A symptom of an industry with unstable market share is the constant price wars. Two characteristics, in my view, make businesses especially prone to cutthroat competition:

1. High frequency, elastic demand with low price ASP. These characteristics are breeding ground for price or discount wars, especially in geographies like China with price-sensitive consumers.

2. A pure asset-light model. Bandwidth and servers are commodities today, which lowers the barrier for any new entrant with deep pockets, strong execution, and a bit of innovation. The only asset-heavy component of e-commerce is fulfilment; but even logistics has become crowded and commoditized in China, with at least seven players operating nationwide networks.

Interestingly, JD has maintained a relatively stable market share (according to the outdated plot above). It also operates at a higher ASP and has a strong logistics network. That seems to support my theory.

If we follow Charlie’s advice and invert the problem, then businesses with the opposite characteristics should be far more defensible: 1) Low-frequency, inelastic demand with high ASP; 2) A physical radius.

And as we can see, the hotel-chain business fits both conditions. Thanks to the franchise model, a hotel chain is financially asset-light while still anchored by a large physical network.

Characteristics of the hotel chain industry

High penetration rate

In the US, hotel chains account for roughly 70% of all rooms. I don’ think this is just a US phenomenon — high chain penetration is almost inevitable, given that both travelers and hotel owners naturally prefer affiliated hotels over independent ones.

On the demand side, travelers choose hotels affiliated with large brands for their consistent service quality and predictable experience. Unless someone is on a very tight budget, few are willing to risk a bad night’s sleep, and potentially ruin the next day’s plans.

On the supply side, a strong brand is attractive for hotels owners because it’s an efficient way to achieve a high ROI. Specifically, owners enjoy

1. Higher revenue due to the RevPAR premium enjoyed by strong brands

2. Lower customer acquisition costs. Brands with strong loyalty programs generate their own direct traffic and charge CRS fees that are far lower than OTA commissions.

3. A full playbook, with support across design, construction, operations, and financial management. Without it, a new hotel owner would face a steep learning curve.

4. Lower construction and consumables costs because the chain negotiates bulk purchases across its entire network.

This value proposition to both the customers and hotel owners seems universally valid, regardless of culture and geography. 1 That’s why I think hotel chain penetration in China shall continue improve.

In fact, the value proposition of affiliating with a hotel brand is so strong that owners are willing to cover all the opening costs, bear the full downside risk, and still let the hotel chain take a cut of every dollar they earn. This is the franchise model — and it’s what makes hotel chains such powerful cash generators, and why industry leaders in this space typically were given high valuation multiples.

To close this section, let me make a comparison between hotel chains and restaurant chains. While consistent food quality and safety are real advantages for restaurant brands, it’s also at disadvantage in front of human’s nature to seek variety in what they eat, and the preference of street food or local shops over standardized chains. This challenge is even more pronounced in China, where people take great pride in a long and sophisticated food culture. The recent controversy around pre-prepared dishes (预制菜) is, in my view, closely tied to this cultural preference. This is not a problem for Western fast-food chains in China.

High concentration ratio

It’s not hard to see the network effect at work from the discussion above: a strong brand brings more loyal customers and a RevPAR premium, which in turn drives the opening of more self-operated hotels and franchisees. As the chain grows, it gains more data to refine its management system and can invest more in technology and brand upgrades. Combined with a larger footprint, this attracts even more customers — completing the flywheel.

Consequently, strong brands with high RevPAR and large networks tend to dominate the industry. Unless a top player hurts itself through worsening quality, new entrants face significant barriers. It’s no exaggeration to say that the hotel-chain business is one of the most defensive in the world.

To see the advantage of a stronger brand quantitatively, let’s compare the profitability of H World’s franchisees with those of Jinjiang and BTG, as shown in the table below:

In 2024, blended RevPAR for H World was ￥235, compared with ￥159 for Jinjiang and ￥147 for BTG — a 47% to 60% premium. Also note that the industry-wide RevPAR was only ￥118, which speaks volumes about the value of a strong brand.

Next, let’s look at traffic acquisition cost. Thanks to its strong loyalty program with 300 million members, H World sold 75% of its room nights through its own channels, with 66% coming from its CRS (the rest from B2B channels etc.). Jinjiang reports a 56.9% CRS sale in its recent prospectus, and BTG is reported to be in the 45%–50% range. 2

Assuming OTAs charge 12% of the room price (this number can be as high as 20%) and B2B channels take 9%, H World’s franchisees enjoy roughly 1%–2.5% lower traffic-acquisition costs compared with Jinjiang and BTG. And independent hotels that rely almost entirely on OTAs? They end up paying nearly twice as much (indeed OTA is also a great business).

Putting everything together, H World’s franchisees not only earn a 47%–60% RevPAR premium, but also benefit from 1%–2.5% lower traffic-acquisition costs. In total, a hotel affiliated with H World can earn roughly ￥3 million more per year than its competitors (for a 100-room hotel in a similar location). The economic advantage enjoyed by hotels affiliated with the CR3, compared with smaller or lesser-known chains, is even more pronounced.

H World’s growth potential in China

So far, we have established that hotel chain industry tends towards high penetration rate and high concentration. So what does this imply for H World?

Industrial growth

According to Frost & Sullivan, by the end of 2024, hotel-chain penetration in China had reached 42.8%, compared with 43.6% globally and 73% in the US. It has taken China two decades to get to this point. If the US is a useful reference, which I think it is, China likely has another 20 years of runway ahead.

If we assume that all future room growth in China comes from chains, then reaching a 70% penetration rate would require roughly 30M affiliated rooms (and assumes no growth in independent hotels). That implies more than 3x growth potential for the hotel-chain industry — which also serves as a lower bound for H World’s long-term upside.

Market share growth

Since 2016, H World’s room share has increased by roughly 1 percentage point to 11.5%. While that may seem modest, Jinjiang and BTG have lost about 2.5% and 6.4% over the same period.

Compared with North America, where Marriott and Hilton hold more than 25% and 20% market share respectively,3 the hotel-chain market in China is still relatively fragmented. In my view, this lower concentration reflects the fact that the industry in China is still in a growth phase. Once growth slows and the market begins to mature, leading players will start to take share from smaller ones. In theory, H World’s market share could still double from here over the long run.

Considering both industry growth and further consolidation, H World still has room to grow roughly 6× before reaching its ceiling in China.

But that’s not the full picture — this does not yet factor in RevPAR growth. The long-term inflation rate of a mature economy (around 2%) should serve as a lower bound. H World has done much better over the past decade, delivering a 5% RevPAR CAGR even with the COVID period included. For H World specifically, I think consumer expenditure per capita is a more relevant benchmark, as the company has consistently upgraded its brands to capture China’s consumption upgrade.

For a long-term investor, once a company’s runway is clear, most of the work left is tracking and verifying its competitive advantages over time. However, predicting where a company will be in 3-5 years is not an easy task. Still, we need some form of valuation to understand the price we’re paying to own the business. The remainder of this post will address that.

2025 — 2029 topline projection

From the table below, we can see that China’s hotel-chain industry experienced a rapid growth phase for about 15 years before entering a period of consolidation around 2015. The industry then enjoyed a good four-year run — H World’s stock rose nearly 10x — until COVID hit.

The pandemic was devastating for independent hotels, whose room count shrank by more than 20%. With no revenue during the lockdowns, many properties without sufficient liquidity were forced to close. By contrast, major hotel chains provided financial support to their franchisees to keep them afloat. As a result, brand-affiliated rooms have grown at a 12% CAGR since 2019, and chain penetration jumped from 26% to 40% coming out of COVID. Independent hotels did rebound after the pandemic but have largely remained flat since.

On the other hand, RevPAR trends are less encouraging. Industry RevPAR briefly surpassed its 2019 level in 2023, but has since slipped back. Meanwhile, RevPAR for major brands in the US has already reached new highs.

During this turbulent period, H World (and Atour) outperformed the industry on both metrics, which I view as clear evidence of their competitive advantages:

Recall that in my last write-up I argued that a hotel chain’s market share should be measured by its GMV — the total revenue generated from all room nights across its network. Looking only at room count is like comparing Tmall and PDD by the number of items sold. Both H World and Atour disclose this metric, but most others do not. Using RevPAR × number of rooms as a proxy,4 I estimated the following market-share changes since 2019:

Back to the topline projection: a bull case would simply extrapolate H World’s performance over the past five years — roughly 14% CAGR in room count and 5% CAGR in RevPAR. This kind of straight-line extrapolation is often the biggest source of error in investing, but under this assumption, H World would have no difficulty reaching its goal of 20,000 hotels by 2030.

What if we have entered a down cycle starting in 2025? The industry-wide decline in RevPAR this year is not encouraging. I worry that this is something that should have happened after 2019 but was delayed by the pandemic.

However, even in a down cycle, I think it’s unlikely that H World will slow its network expansion — it wouldn’t be the first time they expanded counter-cyclically. But to stay conservative, I assume a 10% CAGR in room count and a 2% CAGR in RevPAR for the bear case. Whether oversupply becomes an issue will largely depend on how quickly China exits deflation and how fast consumer confidence recovers.

In summary, under my bull (bear) case, H World would have 2.0 million (1.7 million) rooms in its network by 2029, with blended RevPAR reaching ￥331 (￥287).

Financial model and valuation

Once we estimate the total revenue generated by all hotels in H World’s network, the next step is to build a financial model to determine how much of that revenue ultimately flows to H World’s bottom line.

I find the way the income statement is presented — along with several of the metrics management highlights — not very informative, if not somewhat misleading. This is a common issue across hotel brands due to segment mix and pass-through items. Let me explain.

Mixture of 1P and 3P

The revenue of H World is a mixture of two segments: Leased & Owned (L&O) hotels and Manachise & Franchise (M&F) hotels. The former represents the full revenue generated by hotels operated directly by H World, while the latter consists of fees — 8.5% of hotel revenue — collected from franchisee.5

The problem with this mixture is that the contribution of each hotel differs dramatically between the two segments: one L&O hotel generates roughly 10× the revenue of an M&F hotel, yet its gross margin is below 30%, compared with 100% for the M&F model.6

To see why aggregate revenue can be misleading, consider a scenario where H World converts an L&O hotel into an M&F one. Total revenue would drop, but that doesn’t mean the business has worsened in any way. A similar issue arises when looking at SG&A as a percentage of total revenue — with L&O and M&F mixed together, I’m not sure how meaningful that metric really is.

Passing through items

Extracting the gross margin of each segment from the income statement is actually not possible, because the reported costs are also aggregated figures:

Sharp readers may wonder why the M&F segment even has “hotel operating costs” in the first place. The answer lies in pass-through items.

The most obvious pass-through item is the hotel manager salaries. Each manachised hotel is assigned a manager by the brand, and the manager’s salary (typically ￥20k–￥30k per month, depending on the city and brand tier) is reimbursed by the hotel. Economically, this has no impact on the group’s P&L, but accounting-wise it is recorded as both revenue and cost within the M&F segment.

You can verify for yourself that, once hotel manager salaries are included, the “gross margin” of the M&F segment drops to around 80%. However, the most recent investor presentation shows a gross margin of about 65% (One thing I appreciate about H World is how they’ve improved the their presentations in small ways over the years):

So why the discrepancy? It turns out that M&F revenue also includes the OTA commissions (except Trip.com) that H World collects on behalf of hotels, and the corresponding remittances are recorded under “Others” in Hotel operating costs:7

My calculation estimated that around half of that Others costs is OTA commissions over the past three years.

My model

If the previous two subsection felt confusing, here is how I approach it. Think of H World as an e-commerce business: L&O is the 1P model, and M&F is the 3P model. You begin by projecting GMV (as I did in the last section), then calculate revenue and gross profit for each segment, and finally project SG&A as a percentage of GMV.

Specifically, 100% of 1P GMV plus 8.5% of 3P GMV gives you total revenue. The 1P gross margin is tricky because 1) it’s volatile because of high fixed costs, and 2) H World only began reporting it separately starting in 2025 Q2. With just three quarters of data, the margin has ranged from 14% to 28%. To keep things simple, I use 20% in my model. The 3P gross margin, as discussed earlier, is 100%.

What is a reasonable level of SG&A expenses? Let’s look at SG&A as a percentage of GMV for H World, Marriott, and Hilton:

Now, you might think Marriott and Hilton spend less on OPEX and are therefore more efficient than H World. That’s actually not the case due to another pass-through item that does not exist for Chinese companies! This points to an interesting difference between the business models of Chinese and US hotel chains.

You can see this difference by noticing that Marriott and Hilton report no Sales and Marketing expenses on their income statements. It turns out all of those expenses are treated as reimbursements from their franchisees.

You can think of it this way: brands charge a franchise fee (5%–6%) and an advertising fee (around 4%). The advertising fee is contractually required to be spent entirely on sales and marketing activities for the brand, and the brand is not supposed to earn a profit on it.

Using Marriott as an example, both the advertising fees and the related expenses are recorded under Cost Reimbursement Revenue and Reimbursed Expenses. Over the long run, these two items should offset each other.

For H World and other Chinese hotel chains, the closest equivalent is the CRS usage fee, which is entirely the brand’s revenue and can be used at its discretion. At first glance, this suggests Chinese hotel chains could be more profitable than their U.S. peers. However, in my view, if H World wants to remain competitive with major international players on the global stage (and that is clearly the founder’s ambition), it will need to reinvest essentially all of its 3.5% fee into marketing.

Therefore, in my model for H World, I assume SG&A at 4.5% of GMV (3.5% + 1%), which is roughly in line with its historical average, excluding the COVID period.

Valuation

My projections and valuations are attached below:

My take is that at the current price of $47 per share, the setup leans more toward a hold than a buy. That said, H World did dip to $31 per share this July, which was a solid buying opportunity.

I have completely ignored H World’s international business in this write-up. It would be interesting to examine their progress in Europe since the 2019 acquisition in a separate post. It’s great to see they are testing waters in the ASEAN region.

The author of this write-up owns shares in the companies mentioned and may purchase or sell shares without notice. This write-up represents only the author’s personal opinions and is not a recommendation to buy or sell a security. No information presented in the write-up is designed to be timely and accurate and should be used only for informational purposes. Readers of the write-up should perform their own due diligence before making investment decisions.

A lot seems to have happened since my last post on DiDi. In July, Goldman Sachs initiated coverage on DiDi with a buy rating, after which the stock has been volatile. Last week, DiDi disclosed in their Q2 release that it had settled a shareholder class-action lawsuit for $740M.

These developments make may investors believe that a DiDi relisting is becoming increasingly likely. While I also think a relisting is plausible, it’s important to remember that this is just speculation. Ultimately, the success of this idea depends on whether DiDi can execute well and whether it is undervalued—not on where it is listed, or even if it is listed at all.

For the first half of 2025, DiDi reported ￥4.9B in adjusted EBITA (ex-SBC), up 116% yoy. Even assuming a flat second half, and with ￥62.6B in cash & equivalents, the stock is trading at 15.6x EV/2025E EBITA—this is after a 30% appreciation since my initial write-up. This outcome is the result of strong performance in China and contained losses in international operations and other initiatives. Let’s break it down.

China Mobility: Continued strong performance

For H1 2025, total transactions rose 15.2% YoY, while GTV increased 13.7% YoY, hitting the upper end of my topline growth assumption range of 3%–13%. Adjusted EBITDA margin reached a new high of 4.5% in Q2, putting it on track toward my 6% target. Overall, I think DiDi’s China segment has executed exceptionally well so far in 2025. In fact, DiDi has consistently performed well since 2023 as shown below.

It’s worth taking a quick look at what has happened since 2022. Order numbers were flat that year due to both the suspension of new user registrations and Covid lockdowns. Despite this, the margin in 2022 was relatively high at 3.5%, although it was masked by the 8 billion RMB fine paid (notably, DiDi chose not to exclude this from adjusted EBITDA).

As I discussed in my deep dive, the company’s immediate return to growth highlights its moat. Order numbers went on a streak of 10 consecutive quarters of growth. The margin decline in 2023 suggests that more subsidies were deployed to reclaim market share from competitors. By 2025, the margin of 3.5% achieved in 2022 was finally surpassed.

Speaking of margin, on August 22nd, multiple ride-sharing platforms were widely reported to have announced reductions in take rates, causing DiDi’s stock to drop 6% that day. A closer look, however, shows that the announcement was about reducing the maximum take rate to 27%. I’m not worried about this for several reasons. 1) it appears to be largely a gesture to the government. 2) recall that my calculations suggest DiDi’s average take rate is only 24%, so the reduction doesn’t have any material impact. 3) as I noted in my writeup, I expect a large portion of margin expansion driven by lower consumer subsidies and operating leverage.

Loyalty Program Upgrade

Before moving on to DiDi’s international operations, I want to highlight its recent loyalty program upgrade. Recall the key drivers of DiDi’s growth are MAU and trip frequency. In other words, DiDi needs to strengthen user stickiness and boost user engagement, and the loyalty program is a good example of how the company is doing so.

Following the upgrade, the program now features eight membership tiers. Each tier is earned by reaching a certain mileage threshold over the past three months and comes with progressively better perks. Below is an example of perks enjoyed by a v7 member (translation is mine):

A user can unlock Level 7 perks by accumulating 1,100 mileage points over the past three months. Different service has different multipliers. For instance Level 7 needs 1,100 miles from DiDi Express or 370 miles from Chauffeur. In addition to several ride-hailing–related perks, the more interesting feature lies in the Partner Perks (highlighted in the red box). Let’s go through each of them and, more importantly, assess their value:

1. H World Platinum Membership – Obtained by upgrading from Gold membership (￥219) with 40 room nights. Using ￥300 per night (H World’s Q2 ADR was ￥296), this perk is worth ￥12k.

2. Hilton Gold Membership – Requires 40 room nights. At an estimated RMB 600 per night, this perk is worth about ￥24k.

3. Atour Platinum Membership – Requires 24 room nights. With an ADR of ￥440 according to Q2 2025 report, this perk is worth roughly ￥10.6k.

4. Haidilao Golden Sea Membership – Requires ￥6k in spending to unlock.

Putting everything together, DiDi’s Level 7 membership is worth at least ￥42k—you just need to keep riding with DiDi to get it. What a very clever way to boost customer engagement and loyalty. I’ll wrap up this section with two questions for you to chew on:

1. Would these partner brands want to form similar partnerships with DiDi’s low-price competitors on Amap? Are customers chasing a ￥1–2 discount really their target audience?

2. Would it be difficult for DiDi to charge, say, ￥50 for a membership worth >￥10k? With 10 million paying users (vs. current MAU of over 100 million), that would add ￥500 million directly to the bottom line.

International Business: High growth, early days for profits

For the first six months of 2025, the number of transactions reached 6.6B, up 25% yoy (a better metric to study trends, as GTV was affected by forex fluctuations). Adj. EBITDA during this period enlarged by 9.4% to ￥-0.9B. As shown below, the EBITDA margin has been shrinking toward break-even but has plateaued since 2023.

In my deep dive, I highlighted that the biggest uncertainty for DiDi’s international business is when it can achieve profitability. Indeed, it is still early to discuss profits—especially considering DiDi has just announced the relaunch of its food delivery service in Brazil and the chaos brought by Meituan.

In April, 99Food announced its return to Brazil, outlining an expansion plan with R$1 billion in investments to reach 100 cities by mid-2026.1 s this a wise move, given that iFood already holds a 80% market share in the country? It will certainly be a challenging task, but it makes sense to me for two reasons:

1. Brazil is the largest market in Latin America, with the highest population and GDP. Today, the food delivery sector completes only 0.02 orders per person, indicating significant growth potential. As 99’s managing director in Brazil noted, “Growing fivefold in the next 3–5 years is possible.” I’d add that during this 5× growth, the competitive landscape could change dramatically—and no one should underestimate the capabilities of a Chinese team.

2. DiDi (99Taxi) already holds a 30%–40% market share in Brazil, with 700,000 registered motorbike couriers across more than 3,300 cities. It would be a missed opportunity not to leverage this network to expand into the food delivery business. The potential synergy is highly compelling.

However, things got messy when Meituan decided to enter Brazil immediately following DiDi’s announcement. The two have been battling in the Central Court of São Paulo recently .

The dispute began with Keeta (Meituan) suing 99Food for 1) bidding aggressively on the keyword ‘Keeta’ on Google to confuse its search results and 2) preventing some of its partner restaurants from signing contracts with more than two delivery platforms (similar to what iFood did to DiDi back in 2021).2

Meituan counter attacked by filing lawsuit accused Keeta of using colors, graphics and font similar to 99's platform to "catapult its commercial operations" in the Brazilian market.3 I have to say, the following pictures gave me a good laugh 🙂

Honestly, Meituan’s move puzzles me, as they have no prior presence in Brazil like DiDi. During the ongoing food delivery war in China, Meituan seemed like the victim of JD. Perhaps it’s time to echo what Confucius said 2,500 years ago:

Do not do unto others what you would not have them do unto you
己所不欲勿施于人

In summary, with the new investment in Brazil and the intrusion of Meituan, I wouldn’t be surprised to see DiDi’s international segment losses edge higher in the remainder of 2025.

Wrap up

Finally, let me point out that we saw a record-low loss for Other Initiatives in 2025Q2—let’s hope this trend is sustainable.

Together, this results in ￥4.9B in Adj. EBITA for the first six months of 2025. Share-based compensation has been trending down over the past three years, which should help close the EBITA–earning gap.

Another welcome update is that, under the new share repurchase program commenced in March, DiDi has spent $1.26B to buy back around 6% of its outstanding shares, with the majority of repurchases occurring before June.

Overall, I’m quite pleased with DiDi’s performance so far, particularly in its China segment. I don’t see much reason to adjust the assumptions in my projections from my deep dive.

The author of this write-up owns shares in the companies mentioned and may purchase or sell shares without notice. This write-up represents only the author’s personal opinions and is not a recommendation to buy or sell a security. No information presented in the write-up is designed to be timely and accurate and should be used only for informational purposes. Readers of the write-up should perform their own due diligence before making investment decisions.

Introduction

Boss Zhipin is the largest online recruiting platform in China. Together with ByteDance and PDD, they are the three Chinese companies that built their products and business models upon feed recommendations. In my view, replacing search with recommendation is the greatest innovation and contribution by the second generation of Chinese tech entrepreneurs. As a testament to the power of this disruptive model, all three companies not rose and reshaped their industries’ competitive landscapes in just around five years.

From an investor’s perspective, Bytedance is out of the picture as it stays private. Among the remaining two, I find it interesting to ask myself: which one would I choose to hold for the next 10 years? My pick is BZ, hands down—both for its competitive strength and growth potential. Even better, BZ is still relatively unknown to many investors compared with its more prominent peers. Here’s why I favor BZ.

Better competitive position

First of all, recommendation is much more needed in recruiting than e-commerce. As I will explain in the main text, recruiting and job searching is a matching problem where search barely works. This is not entirely true for e-commerce - the need of a shopping experience or high-quality products will always be there. Therefore, recommendation feed only helps PDD to establish a differentiation (via the online Costco model), but it enables BZ to truly disrupt online recruiting. This is the fundamental reason why BZ is replacing its predecessors, while PDD is co-existing with them.

Investors should look for widening moat, not just wide moat. One key factor in a business’s ability to widen its moat is how well it can withstand cutthroat competition—a particularly important question when investing in China. Unfortunately, PDD operates in a world where cash-burning subsidies, discounts, and aggressive ad campaigns are the norm. As I’ll explain later, those tactics, common in e-commerce, food delivery, and ride-sharing, barely work in online recruiting. That’s why I believe BZ’s moat is far more robust than PDD’s.

Higher ceiling

Putting these two points together, I’m confident that BZ will continue dominating and capturing more market share—still under 10% today. In contrast, PDD’s growth is limited by the overall industry growth.1 After all, how confident can anyone be that PDD will take a significant chunk of market share from BABA, JD, or Douyin?

BZ’s founder, Peng Zhao, once said he sees Recruit Holdings as the ceiling for this business. Today, Recruit’s market cap is around $90B, achieved through diversification (staffing, etc.) and overseas acquisitions. I think BZ, however, doesn’t even need that kind of expansion to reach a $100B valuation—a 10x growth from here—because China’s employee base is roughly 8 times larger than Japan’s.

Here is the math: China’s GDP is around $19T, with more than half going to employee compensation.2 If enterprises allocate just 1% of that $8T to online recruitment platforms, the industry TAM would be $80B. If BZ captures a 30% share and maintains a 50% margin, that implies $12B in profits. Applying a 10x multiple gives a $120B valuation.

A founder like no other

The real source of BZ’s widening moat lies in its culture—and in the founder who shaped. It’s no exaggeration to say that this report would not exist without my deep admiration for Peng Zhao. His story - built BZ with zero tech background - is legendary. The combination of his deep business insight, management skills, and exceptional integrity are truly rare.

I’ve observed that many exceptional managers aren’t always shareholder-friendly. And people often make exceptional effort to accommodate them. With Peng Zhao, you don’t have to worry about that. He is always open and transparent in earning calls, and he believes that shareholder return is the right thing to do. I’ve scattered his quotes throughout the report so you can get a sense of his character and brilliance. Here is the first one:

“… we have long been insist on shareholder return, which is a basic ethic for the company. Confidence from both our employee and our shareholders come from real monetary returns."

— Peng Zhao, 2024 Q2

Thesis Summary

BZ’s stock has been stuck below $25 since 2022, even as the company kept posting record revenue and profit. It’s now back at the highs again — let’s see if it can finally break out this time.

My valuation puts BZ at $35–$57 per ADS, depending on the assumptions in my bear, base, and bull cases. Even after an 80% rally since April, the stock still trades about 30% below my bear case estimate.

Key Thesis Drivers

My five-year cash flow projection is based on the continued execution of BZ’s current strategy—prioritizing user growth while remaining conservative on monetization. The key points are outlined below and discussed in detail later.

• BZ disrupted online recruiting with its MDD model (Mobile, Data, Direct Recruit), addressing the pain points of the traditional job-board approach. This new model delivers higher matching efficiency and opens access to the massive, previously underserved market of SMEs and blue-collar workers. Together, these factors enable BZ to build a scalable two-sided network with much greater growth potential than its predecessors’.

• BZ’s moat lies in the fact that its model can’t be easily replicated, as competitors operate under a very different culture. BZ is a tech-driven platform focused on serving job seekers, while 51Jobs and Zhaopin has been sales-driven businesses serving to large enterprises.

  
  BZ also enjoys the privilege of widening its moat with little resistance. Cash-burning marketing tactics is ineffective in this space, as recruiting is inelastic, low frequency behaviour. You can’t pay people to change jobs or subside companies to create hires.

• User growth will remain the main driver for BZ over the next five years. I project the number of verified enterprise users (along with job seekers) to grow at 15% annually, a conservative assumption compared with the 20%–30% growth seen over the past three years. Under this assumtpion, enterprises on BZ would represent only 42% of all employed companies in 2029.
  
  Paying ratio is closely linked to economic conditions and the supply-demand balance in the job market. I assume it will recover by 1% per year starting in 2025, 2026, and 2027 under my bear, base, and bull cases, respectively. Although ARPPU remains relatively low, management is unlikely to push it aggressively in the near term, so I assume it stays largely flat over the next five years.

• BZ enjoys strong operating leverage. Its current Sales & Marketing spend is already twice that of its competitors, yet represents the smallest percentage of revenue. I assume R&D expense and capex will decline by 10% over the next two years due to reduced investment in LLM. I’m very pleased with management’s measured and disciplined approach to AI. Under my base-case projection, EBIT margin (excluding SBC) reaches 56% by 2029.

• A 20% paying ratio and around ￥1,000 ARPPU leave plenty of room for further monetization beyond 2029, making my bear-case exit multiple highly unlikely. For comparison, companies like Recruit and especially SEEK have historically traded between 20x and 30x EV/EBIT. BZ has also started experimenting with overseas expansion, though it’s too early to draw conclusions. I assume an EV/2029 EBIT of 15x, 20x, and 25x for my bear, base, and bull cases, respectively.

• Outstanding options and RSUs are fully included in my calculation of intrinsic value per share. I also assume a 2% annual SBC in projecting owner’s earnings. No stock buybacks are factored in, making my valuation conservative given BZ’s generous buyback activity in recent years.

Catalysts

1. Share buybacks. BZ spent ￥2.67B on buybacks over the past three years, and management just announced a new $200M plan.

2. Declining SBC: High share-based compensation following the IPO and dual listing has dragged down BZ’s bottom line. SBC dropped 15% in 2025 Q1 and 24% in Q2. As SBC continues to decline, BZ’s true earning power will become clearer.

3. Recovery of China’s economy: A stronger economy would support both the paying ratio and ARPPU.

The Rising of Boss Zhipin

From day one, this industry never had to worry about its revenue model. But by the thirteenth year, around 2010, it had reached a point where making a modest profit was easy and enough to get by, yet the chance to make real money was gone. So the question became: how do you transform the industry? How do you make world-class investments, attract top talent, stay at the forefront of internet products and technology, and offer services that truly belong to this new era? Making that decision was anything but easy.

— Peng Zhao

The recruitment industry was one of the earliest to move online—well ahead of sectors like food delivery, tourism, education, and payments. Platforms such as Zhaopin and 51Job, founded as early as 1997, were pioneers in this space. The online recruiting businesses was lucrative since day one, while other internet companies were still exploring profitable models. 51Job went IPO in the US in 2004, became the best performing Chinese stock that year.

However, as Horace said, ‘many shall fall that now are in honour’. In 2021, Wechat was released — officially kicked off the generational wave of mobile over the next decade. Incumbents who active embraced the mobile trend survived and thrived. Jack Ma’s command All-in Mobile in 2013 was legendary.

Unfortunately, that’s not the case for 51Jobs and Zhipin. They got too comfortable with their lucrative business and are numb to the new, disruptive trend. Then BZ was born and released their mobile app in 2014. Same year, Liepin, who started as a headcount companies in 2006, release their mobile app.

Things changed rapidly since then, BZ became the largest online recruiting platform in 2019, measured by any metric, and is growing at an exponential rate. Zhaopin went private in 2017, followed by 51Job in 2020.

However, the issues of traditional online recruiting platforms go far beyond the need of a mobile app. The pain points run at the deep level of their business model.

Problems of the Job Board Model

I more or less understand the pain points in recruitment—I know that search doesn't really work.

— Peng Zhao

Tradition recruiting platforms run a job board model, where job seekers search for job posts, submit resumes and waiting for contact from interested employers. To some degree, the model host a long list for users on each side to search. As I will explain in detail, this model suffers from low matching efficiency, poor user experience, and large underserved market. In the language of network effect, this leads to clogged flywheel and low node number, preventing the network from reaching critical mass and exponential growth.

Uneven traffic distribution

Fundamentally, recruiting is a one-to-one matching problem, which means there are always internal competitions between job seekers and recruiters. Competitions mean negative direct negative network effect on both sides — opposite of ‘the more the merrier’. Competitions also leads to traffic concentration toward top players on both sides (large companies and strong candidates).

Taking the recruiter side as an example. It’s not difficult to understand how the following leads to traffic concentration:

1. Submitting a resume costs job seekers nothing, so most will apply to well-known companies regardless of whether they are actually competitive candidates.

2. Since job descriptions tend to be generic and lack differentiation, small companies, even if ranked highly, are often ignored in favor of large, well-known enterprises.

Interestingly, e-commerce platforms largely free from this problem. There is no competitions among buyers, as each seller can, in theory, absorb infinite demands. There are competitions among sellers, however, it’s only a problem for smaller sellers. The traffic concentration is only welcomed by top players. Actually, such phenomenon attributes greatly to PDD’s early success.

Back to recruiting, the competitions and uneven traffic distribution lead to terrible user experience—large companies complain about being flooded with unqualified resumes, smaller companies receive little to no interest despite paying for exposure, and most job seekers never hear back after applying. It's hard to think of another internet product with such a poor user experience.

Largest segment remain underserved

A resume-centric model is inherently unfriendly to both small and medium-sized enterprises (SMEs) and blue-collar job seekers. On the recruiter side, this model works best for large companies with dedicated HR teams to screen resumes. SMEs typically rely on managers to hire directly. On the job seekers side, blue-collars with low education levels have little to write about their previous experiences. Blue-collar roles are often high-turnover and easily substitutable, requiring fast communication and decision making. The traditional process of writing and screening resumes is simply too slow for such needs.

Why does this matter? 99% of China’s enterprises are SMEs and blue-collar workers accounts for around 50% of job seekers. In other words, the resume-centred model fails to serve the very groups that make up the bulk of the job market in China. Without enough number of nodes on both sides, platforms before BZ struggle to reach the critical mass needed for strong network effects.

How BZ reinvented Online Recruiting

I understand that the essence lies in the boss directly finding the right person. I don’t know how well this approach will work, and I don’t know when I’ll fully figure it out, but I’m certain that search won’t work.

— Peng Zhao

It's no exaggeration to say that Boss Zhipin was created precisely to solve the industry pain points I discussed earlier. Its answer—embodied in its business model—rests on three foundational pillars:

This was called the MDD (Mobile, Data, Direcruit) model. As we’ll see, these three building blocks working together to improve matching efficiency, reach a broader segment of the job market, and generate strong network effects.

From Search to Recommendation

As discussed earlier, traditional recruitment platforms suffer from low matching efficiency. Why? Because these platforms never truly leveraged internet technology to facilitate the matching process. Instead, they simply offered two long lists—one of job seekers, one of job postings—and left both sides to manually search through them.

BZ addresses this issue through two-way recommendations. Whether you're a job seeker or a recruiter, the AI algorithm suggests higher-quality matches based on user profiles, resumes, and interaction history.

The recommendation-based approach creates a direct, two-way connection between employers and job seekers, giving long-tail users meaningful exposure. Furthermore, the ‘scrolling addiction’ improves engage and ultimate algorithm accuracy. Together, matching efficiency is significantly improved.

Recommendation also addresses the ambiguity often present in recruitment. Unlike shoppers who typically know what they want, many job seekers—especially early-career, aren’t sure which roles suit them best. In such cases, machine-assisted matching becomes especially valuable. This is also convenient for blue-collar workers — someone looking for a job at a bubble tea shop could just as easily work in a restaurant. These kinds of flexible, substitutable roles are far better served by recommendation than by manual search.

Replacing Resumes with Direct Chat

As for recruiters initiating direct conversations—this isn’t some groundbreaking innovation; it’s just how things work in real life. Liu Bei didn’t hire a headhunter to recruit Zhuge Liang... It’s been this way since ancient times.

— Peng Zhao

BZ also pioneered a direct chat model. Instead of starting with resume submissions, the first point of contact in online recruiting is replaced by instance messaging chat box, which can be initiated by either side. Once a conversation begins, both sides can choose to exchange WeChat, resumes, or other information with mutual consent.

Why was direct chat so quickly embraced by users? Here are three reasons:

1. It brings recruiting back to its natural form. Hiring, at its core, should begin with a simple introduction, and whether it progresses further should depend on mutual interest. Direct chat captures this dynamic far more authentically than traditional platforms.

2. Resume-based models inherently place job seekers in a subordinate position—essentially "begging" for a chance. This leads to a poor user experience: applicants send out resumes without knowing whether anyone even looked. On BZ, recruiters must respond and request access to a resume only with the candidate’s permission. Job seekers feel respected on BZ — what can be more attractive than that? This also aligns with China’s broader labor trends, where companies are now competing harder to attract talent.

3. As mentioned earlier, the resume-screening model is more suited to large enterprises with dedicated HR teams. Even for them, growing demands for innovation and flatter management have led to more direct hiring by line managers. Most importantly, by deprioritizing the role played by resumes, BZ is better positioned to serve China’s vast SME sector and the blue-collar workforce—groups often overlooked by traditional platforms.

Mobile native: product belong to the new era

As I mentioned earlier, BZ emerged alongside the rise of the mobile era. But what, to borrow Peng’s words, truly is a product belong to this new era? Is it just taking a website and turning it into an app? Definitely not. A closer look at BZ’s MDD model shows that every aspect of it is designed in full alignment with the mobile infrastructure.

At the heart of BZ’s model is its algorithm, which relies on large volumes of personalized data for training. That data is generated through recommendation feeds (swipe or click) and direct chat—interactions that are only possible on portable mobile devices. The PC environment simply couldn't support that level of engagement.

Strong network effect

Putting together what we have discussed so far, this is a snapshot of BZ’s consumer model:

As discussed earlier, several pain points severely limited the strength of network effects on the job board model. Therefore, we should expect BZ’s new model, which was designed specifically to address those pain points to enjoy much stronger and more scalable network effects.

Indeed, under the MDD model, not only the negative direct network effect is reduced, the size of network is also significantly expanded. Specifically,

1. By reaching both SMEs and blue-collar job seekers, BZ dramatically expands the number of network nodes. This is achieved directly due to the large number of SMEs and blue-collar workers, and indirectly due to the short life cycles of SMEs, as well as the high job-switching frequency of blue-collar workers.

2. Negative direct network effect due to competition was mitigated, as a result of BZ’s two-way recommendation engine. It increases visibility for long-tail users—those who would otherwise be buried in traditional search-based models—thus smooth out the severe imbalance in traffic distribution.

3. BZ’s reach into SMEs and blue-collar jobs introduces a new kind of positive direct network effects, thanks to the cross-industry liquidity. Unlike specialized verticals, blue-collar jobs and service roles are interchangeable. For instance, companies in AI compete for a small pool of Masters and PhDs. In contrast, a bubble tea shop doesn’t need to only consider someone who worked for bubble tea shop before, it can hire someone from a hot pot restaurant, convenience store, or even with little work experience. This drastically reduces same-side competition among employers. The same logic applies to blue-collar job seekers: they don’t need to restrict their search to one narrow category. This cross-industry flexibility reduces competition on both sides and effectively creates positive direct network effects, where more participants make the platform better for everyone without crowding each other out.

BZ’s Moat

Evidences of moat

Readers familiar with my DiDi write-up know that I like to use ‘non-operational crises’ as a stress test for a company’s moat. Similarly to DiDi, BZ was hit with a year-long cybersecurity review, during which new user registration was suspended. Competitors attempted to seize the opportunity, ramping up marketing spend in an attempt to close the gap.

Result? BZ returns to high growth immediately after the ban lifted and the gap with its competitors is getting larger. According to the management, there were 50M failed user registration during the suspension!

Another evidence of BZ’s moat is the complaining about price from HRs. Based on my survey on social media and conversations with HRs, the common feedback is that BZ is getting more and more expensive. Since 2000, BZ has been steadily raising prices, and its annual package now ranges from ￥16,800 to ￥38,800.3 By comparison, Zhaopin and 51job charge only about one-third of that. Even Liepin, which focuses on mid-to-high-end recruitment, caps its most expensive package at just RMB 23,800 per year.

In other words, even with cheaper alternatives available, recruiters still stick with BZ. All they can do is to complain about it. Clearly there are something about BZ that are not replicable.

Why incumbents cannot replicate BZ's model

I tell my team: if your foot gets stepped on in a race, it means you’re not leading by enough — less than two strides. If you’re eight strides ahead, how could anyone step on your foot? They couldn’t even grab your clothes.

— Peng Zhao

Astute readers might ask why 51Jobs and Zhaopin cannot copy the MDD model if everyone can find it in BZ’s prospectus. Well, to some degree, they did:

The UIs look almost exactly the same and each has a button where user can go into a messaging box. But if we look closerly, 51Jobs and Zhaopin failed miserably in their copying. For instance, on BZ, the button says ‘communicate now’ while in its competitors’ app its says ‘Apply now’. Even worse, if I click the apply button in 51Jobs, the first message is “Here is my CV” and my CV is sent automatically to the recruiter! For Zhaopin, things seem better — recruiter do ask persmission for your CV, but most of these are bot auto replies. Just look in Rednote and see how many HR complains that they didn’t even choose to initiate the conversation!

So you see the problem now. BZ’s competitors copied the shell and forgot about the soul. The existing MDD model, or whatever new policy or functionality BZ will launch in the future, is a result of BZ truly aims to improve user experience, to give applicants the respect they deserve, to facilitate matching.

Let’s go deeper and ask why even today BZ’s competitors cannot offer true direct chat? Why they cannot strives for a better user experience? My answer is the culture difference. BZ’s mission is to better serve both enterprise and job seekers by leveraging technologies. Essentially it’s a research and R&D company. But its predecessors are sale company. Such difference cannot be more clear by look at their R&D spending:

Even worse, historically Zhaopin and 51Jobs primarily serve large enterprises—clients who are used to applicants begging to send in their resumes. Now imagine telling these core customers, the very foundation of their revenue, that from now on they’ll need to request permission to access resumes, and that job seekers will be placed in front of them instead. On top of that, management would need to slash the sales force and start paying top salaries to hire AI experts. The level of courage required to overturn such a deeply entrenched corporate culture and confront inevitable internal resistance is enormous. In reality, I’d say it’s almost impossible.

In summary, BZ’s model can be replicated in theory, but not in practice, by today’s major players. In that sense, BZ’s business is not a commodity. The difference in terms of user experience will only be enlarged as time goes by due to the compounding effect, and 51Jobs and Zhaopin are doomed to be eliminated eventually.

At the end of the day, our field just has to move more slowly. Because usage is infrequent, product managers and data engineers learn much more slowly, which in turn flattens the growth you can get out of each dollar invested. That’s why, slow is fast.

— Peng Zhao

Online recruiting is free from cut throat competition

Cut throat competition is a phenomenon cannot be ignored by investors investing in China. Specifically, it means burning billions in short terms to grab users from competitors by giving subsides, discounts and launching ad campaigns. If these tactics works for recruiting, maybe competitors can incentivize user to choose a slightly worse user experience. However, this turns to be not true for the following reasons:

• Recruiting and job searching are fundamentally in-elastic demand. Unlike e-commerce or ride-hailing where subsides could encouraging more orders, more rides etc, this won’t work for recruiting. People looking for jobs when they need them, not because they’re offered a coupon; and employers hire when there’s a real vacancy. The same problem goes for ad campaigns, people won’t start looking for jobs because they see a big banner in the subway. That’s why these campaigns are only active in graduate seasons.

• Recruiting and job searching are low-frequency behaviour, which also render short-term cash-burning campaign highly ineffective. Consumers might book rides or place orders multiple times per week, making it easy to influence their habit and mind share with short-term campaign. In contrast, the number of average work duration, even for 95 generation who are known for low loyalty, is 1.27 years. Paying people to sign up as job seeks can only lead to fake, inactive profiles or unqualified applicants.

• Incentive of switching is low for enterprises. The cost of paying a platform is small comparing to the salaries they will pay or the loss it incur for not able to fill a needed role. Annual plan enough for several positions for recruiters on BZ costs from ￥10k to ￥30k, small company with few demand can choose to pay much less. The average annual salary in China is 100k. Therefore, recruiters cares more about whether there are enough job seekers using the platform than price.

In summary, several properties of the recruiting behaviour renders cut throat competition tactics ineffective, adding another layer of protection around BZ’s moat.

Growth drivers

BZ’s topline is fairly straightforward to analyze—it ultimately depends on how many enterprises are willing to pay and how much they pay. Breaking down the first factor, BZ’s enterprise growth is driven by three levers: penetration rate, paying ratio, and ARRPU.

While I only use enterprise-related metrics in my projection, this by no means diminishes the importance of KPIs on the job-seeker side. In fact, a large pool of active job seekers is the very reason enterprises remain engaged and continue paying. As we will see, the number of users on both sides of the platform is closely linked, a balance that BZ actively manages through deliberate modulation of supply and demand.

Verified User Growth

The number of registered enterprises (excluding self-employed) quadrupled from 13 million in 2012 to 51 million in 20224 , and reached about 61 million by the end of 2024.5 Over the past three years, despite a weak economy, the average annual growth rate has still been around 5%. Given the already high enterprise density compared with developed countries (36 enterprises per 1,000 people), I assume this pace continues to be conservative. This implies that by 2030, China will have roughly 80 million employer businesses.

BZ’s verified enterprise count has increased fivefold since 2019. The hypergrowth phase appears to have passed, with annual growth hovering at 20–30% over the past three years—with no new user registered for half year. I’m confident in the continued expansion of BZ’s user base for two reasons: (1) its business model is particularly well suited to SMEs and blue-collar hiring, supported by strong network effects; and (2) SMEs account for 98% of all enterprises, while blue-collar roles make up 70% of total non-farm employment.

To be conservative, and align with CEO’s expectation, I assume verified enterprises grow by 15% annually. Under this assumption, verified enterprises on BZ reaching around 33M in 2030 - this is just 42% of total employed companies in China.

We have several key structural growth drivers for our revenue, which is still unchanged. The first thing is the user growth. Even though of all these uncertainties, we expect that we can still have at least 15% of overall user growth.

— Peng Zhao, 2024Q2 Earning

Since the ratio between enterprises and job seekers has been relatively stable, due to active control of supply demand balance on the platform, growth on both sides are highly correlated. Therefore, under my assumption, total verified job seekers shall be around 426M or around 75% of non-farm employment in 20306.

Paying Ratio

The upward trends of BZ’s paying ratio peaked in 2022 and has declined for two years:

To understand this we need to understand that paying ratio is directly correlated with the supply demand in the job markets - often measured by the job opening-to-application ratio. When this ratio is high, many companies compete for one applicant. Not only BZ will charge a fee for posting ‘popular positions’, companies are also more willingly to pay, hence higher paying ratio. When the ratio is low, the opposite happens, companies grasp greater bargaining powers, leading to lower paying ratio.

To confirm this theory, we need to study the correlation between job-to-application ratio and paying ratio. However, Chinese government stopped publishing the job opening-to-application ratio after 2022. To get pass this barrier, I find that the ratio of verified enterprises to job seekers on BZ closely tracks the national benchmark (left figure), indicating that BZ’s platform is a representative sample of the broader job market. Therefore, we can use this proxy instead—and it indeed shows a strong correlation with BZ’s paying ratio. (right figure)

The decline in the job opening-to-application ratio—and consequently in the paying ratio—reflects the sluggish economy following the lockdowns. Therefore, in the short term, this driver will largely depend on the pace of China’s economic recovery. However, if we zoom out and look at the past decade, the longer-term trend points to a growing labor shortage:

I believe this shortage is likely to persist over the next decade. According to the 2024 China Statistical Yearbook, there were about 180M people aged 10–19 in 2023, compared with 247.5M people aged 40–59. This implies a net annual decline of roughly 6.7M in the working-age population over the coming ten years. The government is addressing this challenge through two main avenues: boosting birth rates and accelerating urban migration. The former includes policies such as the three-child initiative and a new nationwide childcare subsidy program, while the latter involves reforms to the hukou system. Both measures, however, will take time to show meaningful impact.

Back to my projection, to factor in the uncertainty of China’s economy recovery, I assume the paying ratio for BZ start to recover at 1% annually since 2025, 2026 and 2027 for my bull, base and bear case.

Average Revenue per Paying User

The ARPPU increased by around 36% over the past 5 years, slightly above government’s 5% GDP annual growth target. The number of small-size enterprise (contribute < 5000RMB) is more stable than that of key accounts and mid-size firms, reflecting BZ’s strong position in SME.

On one hand, ARPPU of 1000 RMB is quite low — around 2% of mid level annual salary of three tier cities. This leaves ample growth potential. On the other hand, CEO says clearly and multiple times that he will not consider aggressive increase this driver in the near team.

…our future growth strategy is that we will continue to attract more and more enterprises to pay for our service for their recruitment. That's my basic growth strategy. Based on this strategy of increasing dollar market share, especially our current market share is very low, so we are actually aiming at a company that has not paid for online recruitment at all. We should be very cautious in taking the payment strategy. In other words, I will not pursue the improvement of ARPU.

— Peng Zhao 2024Q1 earning call

And the third driver is our ARPUU. As far as we have got the first and second driver, which is enough, we are not in a hurry to aggressively increase our ARPUU. So it will keep at a stable and slightly improvement trend.

— Peng Zhao 2024Q4 earning call

I think there are two reasons why the management would not like to push hard on ARPUU:

1. Users are complaining. As mentioned before, many HRs are complaining about BZ getting more and more expensive. On one hand, this means BZ’s moat is strong; on the other hand, this put some pressures on management to increase price aggressively.

2. Peng believes in slow is fast. Chinese consumers’ willingness to pay for service needs to be cultivated over a long period. The top priority for now is to continue improving user experience.

Therefore, I assume BZ’s mixed ARRPU grow at long-term GDP growth rate of 2%. This is conservative considering China’s long term goal for the next 10 years is 5%.

2025-2029 Revenue Bridge

Putting everything together, below is my base case revenue projection for next five years:

Specifically, I project BZ’s revenue in 2029 to reach ¥17B, ¥17.5B, and ¥18.2B under my bear, base, and bull case scenarios, respectively. CIC projects China’s online recruitment market at ¥220B in 2025. Even assuming no further market growth, BZ’s share would be only about 8% in 2029, indicating that my projections are far from aggressive.

Expense assumptions

Operating leverage

In terms of operating expenses, BZ allocates the majority of its revenue to sales and marketing, followed by R&D, with G&A being the smallest component. Since 2019, BZ’s revenue grown by 7x, while S&M only grow by 2x, R&D by 5x, and G&A by 7x. Stock-based compensation is not included here, but it will be factored into the valuation

The operating leverage is clearly demonstrated from the able above, and the CEO understand this perfectly.

We have very strong operating leverage, so as long as we can grow, most of revenue will turn into our profit. This is quite clear and simple.

— Peng Zhao 2024 Q2

What’s the source of operating leverage? Two things:

1. The nature of a tech-empowered asset-light platforms. Since 2020, BZ’s headcount has grown by only 68%. For example, the use of AI in security reviews has improved audit efficiency by 30%. This allows the company to handle the significantly higher workload from a rapidly expanding user base with only modest increases in staff. It’s remarkable that a workforce of just 5,000 people supports over 200 million users in the job market.

2. Economy of scale. Table below shows BZ and its competitors marketing expense and as percentage of their revenues:

   

   Competitors spent a higher proportion of their revenues but only achieve half of BZ’s marketing expense. This effect will only be more evident as BZ grows further thanks to its strong network effect.

Below are my projection of each expense terms.

Sale & Marketing

I believe it’s unlikely that BZ will significantly increase its spending in the near term for two reasons. First, there is no real need—the company’s absolute spending is already more than double that of its competitors. Second, BZ has been carefully managing the supply–demand balance on its platform. Even if it could ramp up marketing and attract more job seekers than enterprises, the result would likely be a poorer user experience and reduced willingness to pay among employers. Therefore, I assume sales & marketing expenses will remain largely flat, rising only modestly from ￥1.8B in 2024 to about ￥2B in 2029

R&D and Capex

My projection for R&D spending rests on two key points:

First, while BZ’s R&D expenses will continue to grow over the long term, they will not outpace revenue growth, thereby preserving operating leverage. BZ is fundamentally a tech company, so that investment in R&D will continue. But R&D is a long-term effort. There is no point aggressively ramping up short-term spending which would only hurt profitability.

Second, I expect R&D expenses and capex to decline in the near term as a correction to the AI-hype-driven overspending of the past two years. I am very satisfying with management’s approach to generative AI. Rather than chasing hype, they have taken a measured, disciplined stance. This is best explained by CEO himself:

Regarding the generative AI, … there are two points. First, in a scientific perspective, we are executing a taillight project, which is for our research guys, our tech guys to understand, to know what the most developed technology in this area is and what they are doing. But we are not planning to invest more resources. Actually, we're not affordable to do that, to actually do that. We just know what they are doing and kept up with the most advanced technology.

— Peng Zhao, 2024Q3 earning call

And what has BZ learnt after this short-term investment about AI?

I still think that the entire industry, not just our industry, but all industries in AI, In terms of AI application, the reality is the high prospect of the technology cannot correlate with the real actually application scenario. So there's a very loud lightning, but very small rain. So that's an industry fact.

— Peng Zhao, 2024Q4 earning call

A few words on my thinking about AI. LLMs are, and will increasingly become, a commodity. Smaller but equally powerful models will soon be widely accessible. Ultimately, even if AI applications gain widespread adoption, the real winners will be the companies that possess critical industry knowledge and proprietary data.

Back to projection. As BZ already catch on up generative AI, BZ will cutback on its investment on this front. This already shown in capex down by 10% yoy in 2024, and the R&D expense in 2025Q1 down by 9.4%. Combining my two thesis points, I project BZ’s R&D spending decreased by 10% in 2025 and 2026, and grow again at 10% starting from 2027.

Other assumptions

• G&A expense has been quite consistently stay at around 9.2% of the tot revenue. Other than this, I don’t have any extra insight, so I assume it says at such level in the future.

• Gross margin dropped 87% to as low as 83% in 2022 and 2023. This is due to the slow revenue growth (5%) in 2022, increasing security measures and heavy investment in AI. As the growth came back and BZ dial down its investment in generative AI. Gross margin has been recovering in 2024. I assume this to continue and stabilize at 85% to be conservative.

• Tax expense. PRC enterprises in general are subject to a 25% income tax. However, BZ’s effect tax rate has been below 15%, due to both a preferential tax rate of 15% for qualified High and New Technology Enterprise (HNTE) and tax reduction from certain quality R&D expense. I expect the HNTE status maintained in the foreseeable future as BZ will continue invests heavily in R&D. So I assume tax rate of 15% in my projections.

• Share-based compensation. SBC spiked to more than 15% of revenue over the past few years, due to IPO in 2020 and the dual listing in HK in 2022. As management mentioned, SBC will go down in the future. Most of such one-time SBC will be taken into account as dilution when I calculate value per share later. However, BZ will continue issue new SBC to attract talents in the future. I assume a long-term level of 2% of revenue for that — a reasonable amount for matured tech company.

2025 — 2029 FCF Bridge

Since my I have considered the impact of the capex in my expense assumptions, I use NOPAT to approximate FCF assuming D&A is roughly equal to maintain capex. Combining my revenue projection from last section, below is my base case 2025 -2029 FCF projection:

Under my projection, EBIT margin achieves 56% due to operating leverage. Around 46% of revenue goes to owner’s earning.

Valuation

Share-based compensation

I have not consider the impact of SBC so far, as they are non-cash. However, they are dilutive thus real costs to shareholders. At the end of 2024, there are 31.5M RSUs outstanding, and 44.7M share of exercisable options with a weighted average exercise price $3.11 ($6.22 for each ADS). Since today BZ is trading far above that, I count all of them ITM in my calculation of value per share.

Since I’m using 2025Q2 number for cash and shares outstanding etc, I need to estimate the latest outstanding SBC to be consistency. Very crudely, I assume the outstanding options and RSU continue decline at previous years’ rate and linear over quarters:

There are 883M ordinary shares outstanding at the end of 2025Q1, so the total dilution (excluding cash received from option exercise) is around 7.7%.

Exit multiple

The multiple the market assigns to BZ in 2029 will largely depend on its growth prospects at that time. If growth potential appears limited, a low multiple—say around 15x EV/EBIT—would be justified. However, this seems unlikely given 1) my projection of 25% profit growth in 2029, and 2) my five-year forecast assumes only a continuation of management’s current strategy.

At present, BZ focuses on expanding its user base while being conservative on monetization. Once user growth begins to plateau and stickiness strengthens further, management will almost certainly shift emphasis toward raising the paying ratio and ARPPU. With a potential 20–30% paying ratio and ￥1,000 ARPPU, the monetization runway beyond 2029 remains substantial. That will be the growth story beyond 2029.

Let’s look at comps. There are four public companies with online recruiting businesses that I’m aware of: Recruit Holdings, SEEK, Liepin, and ZipRecruiter. The last two are not doing well, with revenue declining over the past three years. Liepin, which focuses on gold-collar positions, may not be a fair comparison to BZ anyway.

Recruit and SEEK has established international footprints via acquisition. Recruit is more diversified, operating not only in recruiting but also in staffing and matching platforms across travel, food, and other verticals. SEEK may be the closest comparable, as it is entirely focused on online recruiting platforms and has a strong presence in Asia, including Hong Kong, along with a stake in Zhaopin. Both Recruit and SEEK have historically traded in the 20x–30x EV/EBIT range.

Therefore, I consider EV/2029 EBIT of 15x, 20x, 25x for my bear, base and bull case scenarios.

How about new business opportunities? Broadly, there are two paths. One is to penetrate deeper into the HR industry, which is roughly four times the size of the recruitment services market. The other is to replicate BZ’s model in other countries. Based on my understanding of Peng—who is cautious about entering unfamiliar markets and reluctant to compete on others’ turf—option one seems unlikely in the near term. In fact, BZ has already begun experimenting with overseas expansion, though little detail has been disclosed so far.

…we believe a business and revenue model that has proved in China, we hope we can deploy it in other countries/regions. And hopefully in the next 3-5 years that will be our new growth driver.

Peng Zhao, 2023Q2 earning call

DCF

Putting everything together, I calculate BZ’s intrinsic value to be $35 — $57/ADS. Below is the DCF model for the base case:

Note that I assume all cash remains in the bank, which is too conservative. From 2020 to 2024, BZ spent ￥2.67B on share buybacks, and management has just announced a new $200M buyback plan. If they continue allocating around 30% of operating cash to repurchases in the coming years, it could be highly accretive to BZ’s intrinsic value.

CEO and founder

If you’ve read this far, you already have some sense of BZ’s founder and CEO, Peng Zhao. Except for the quotes from earnings calls, the ones I shared come from a book based on an extensive interview with him:

I highly recommend this book and have read it at least twice. Each time, I ask myself how someone like him can remain largely unknown to value investors focused on China.

Peng’s background

In fact, much of his character can already be glimpsed from his usual background:

• 1994 – Graduated from Peking University Law School.

• Same year entered government work; served as Project Development Director at the China Youth Volunteers Association.

  • Helped organize the pioneering “College Students Volunteering in Western China” program.

  • By July 2020, program sent nearly 300,000 graduates to 22 provinces, with 20,000+ volunteers active annually.

• 2005 – Left government as a division director after 11 years at the Communist Youth League.

• 2005 – Joined Zhaopin as PR manager.

• 2010 – Promoted to CEO of Zhaopin.

• 2013 – Founded Boss Zhipin

What can we tell from this mini-resume?

1. He is a seasoned leader, with extensive management experience both in government and at Zhaopin.

2. He has a deep understanding of the industry and business in general. He learn old model’s pain points with first hand experience in Zhaopin.
   
   Without these two points, it’s impossible for him to build a successful tech company with zero tech background.

3. He has strong integrity. The volunteer program he started is still active and making a real impact today. In earnings calls, he often emphasizes how BZ strives to help young graduates find jobs they truly enjoy.

Finally, it’s worth noting that Peng’s government experience gives BZ an edge in navigating China’s complex policy landscape—an advantage most other tech companies simply don’t have.

The author of this write-up owns shares in the companies mentioned and may purchase or sell shares without notice. This write-up represents only the author’s personal opinions and is not a recommendation to buy or sell a security. No information presented in the write-up is designed to be timely and accurate and should be used only for informational purposes. Readers of the write-up should perform their own due diligence before making investment decisions.

However, I noticed that this incident received little attention from Western media or hasn’t been covered by Substacks that focus on the Chinese market. So I decided to translate it by myself (with the help of ChatGPT)2 and share it here. I've also included a few footnotes and figures to help clarify and contextualize the content.

I believe my translation could be valuable for many of my subscribers who invest in China but can’t read Chinese. Whether you’re interested in Alibaba, the e-commerce sector, or simply curious about the local dynamics in China over the past few years, I think you’ll find this letter insightful.

What’s my take on Alibaba’s situation after reading this letter? At first glance, many of its problems seem tied to the broader slowdown in China’s economy—so in theory, if the economy rebounds, Alibaba should too. But interestingly, in a follow-up interview, the author of the letter offered an insightful disagreement:3

I don't think they actually read my post carefully — or maybe they didn’t understand it — because we’re not even talking on the same level. They’re saying that as long as the company finds a new growth direction, the problem will be solved. But what I’m asking is: what kind of company is actually capable of seizing new growth opportunities in the first place? Alibaba has missed several major opportunities in the past — why did that happen? That’s what I’m trying to explore.

They argue that growth can cure the symptoms of a large, sluggish company. But the very problem is that this kind of organizational disease is precisely what makes it hard to find real growth. They’re focused on treating the symptoms, while I’m trying to understand and address the root cause.

In fact, I believe the dilemma Alibaba faces is a miniature of the broader challenges confronting China today. On one hand, China urgently needs a new growth model to move beyond the increasingly excessive competitions across industries and the widespread sense of involution (内卷) felt across all social classes. On the other hand, if resources and energy continue to be wasted on cutthroat, inefficient competition, it will become increasingly difficult for the country to innovate and unlock new sources of growth. It’s a vicious cycle—and breaking out of it will require effort from both sides.

As for Alibaba, the very fact that this letter was openly circulated on its internal forum4—and acknowledged by Jack Ma himself—gives me some confidence in the company’s future. Perhaps the spirit of the “Ali man” isn’t dead after all. Rather than a time teller, I believe Jack Ma is a clock builder. And if you believe Alibaba is one of those companies built to last, then its recent setbacks may well be buying opportunities.

Below is the translation — I hope you enjoy it.

Alibaba’s mission is to make it easy to do business anywhere. Its vision is to be a good company that lasts 102 years.

Back then, I was inspired by this mission to join the company, hoping to accompany it through that long journey—because Alibaba was a great company.

When I first joined, we were working in the Chuangye (Entrepreneurship) Building.5 At lunchtime, we’d go to nearby restaurants where everyone wore their employee badges. The shops gave us discounts, but more than that, we felt a sense of respect and gratitude. Back then, respect didn’t come from high salaries, but from the kindness of Alibaba employees and the impact of our products. People around us could feel that Alibaba was truly helping solve real problems. The respect we received from society came from recognition of our character and values.

There was even a popular joke at the time: Mothers-in-law in Hangzhou prefers Alibaba employees as sons-in-law—because Alibaba had already done the screening for good character. Alibaba people were trustworthy.

I can’t remember exactly when it changed, but over the years, it became harder to wear that badge with pride. A string of negative events and public criticism, combined with shifting employee mindsets, changed things drastically—both outside and within. Maybe those mothers-in-law still want Alibaba sons-in-law, but now it’s probably because they earn good money.

After fifteen years, I must say this isn’t just about how others see us— we ourselves whave changed. We’ve lost that original dream of making the world a better place. Now we talk about KPIs, salaries, stock, real estate. We treat our customers and users as traffic and data. We discuss how to operate data, how to extract value from customers. Internally, we burn energy on the ‘361’ battles.6

Watching our core values slowly erode worries me deeply. How much longer can we hold onto our mission? Can we still achieve our vision?

To be clear, my own performance has been good. I have no complains with my current team or manager. I’m resigning as part of a personal plan, not a layoff package. So this post isn’t directed at any specific team or individual. What I say applies to any part of the company where these situations resonate. This is a reflection based on my own fifteen years at Alibaba, and conversations with friends across different BUs. It’s mainly meant as feedback to the company and leadership—a few final suggestions before I go. If you feel everything is great at Alibaba right now, feel free to skip this.

I’ll skip the praise and focus on the problems. If anything I write is incorrect, I welcome corrections. Once again, I stress that the issues I raise do not apply to all departments or all individuals at a given level. These problems are widespread, though not ubiquitous. If anything I say comes across as offensive, I apologize in advance.

Once Glorious

Let’s first talk about what made Alibaba great. Everyone may have their own explanations, but I would sum it up in four key factors: the macro trends, Jack Ma’s vision, shared values, and supportive corporate system.

Tailwinds of the era created enormous opportunities

Reform and opening-up, along with globalization, ushered in a surge in productivity. In terms of USD, China’s GDP grew from $1 trillion in 1999 to $11 trillion in 2010, and to $18 trillion in 2024. This massive economic growth was the tailwind beneath everything. Internet users in China rose from 8.9 million in 1999 to nearly 1 billion by 2020. The widespread adoption of smartphones pushed the number of mobile internet users from almost zero in 2005 to 986 million by 2020. The rapid development of both PC and wireless technologies gave rise to countless application scenarios.

Jack’s vision

Jack had both the foresight and the resolve to seize these opportunities. From B2B to Taobao, Alipay, Alibaba Cloud—one strategic decision after another allowed Alibaba to stay ahead. With resource integration centered around e-commerce, the Alibaba ecosystem was built step by step. Of course, none of this would have been possible without tremendous determination, especially in the case of Alibaba Cloud. At the time, the outside world was skeptical, and internal support was weak. Results weren’t visible—but the company persisted. That required serious conviction from senior leadership.

Unique team built on shared values

With strategy and resolve in place, the next step was to form a strong, capable team to execute. This was made possible by Alibaba’s unique culture. It brought together like-minded people who united around a common mission. We called each other "classmates"—we helped each other, we learned from each other. We used nicknames7 regardless of rank, reflecting a culture of equality rather than rigid hierarchy. We wanted to make the world a better place. We wanted to solve real problems. Not everyone who joined was exactly the same, of course—but before 2015, there was a common spirit among Alibaba people.

Well-designed corporate system empowering the team

Our core value —Customers first, employees second, shareholders third—was embedded in our system design, which empowered employees to fight for both the company and themselves. A free and open culture allowed everyone to speak their minds. The stock option system gave employees a sense of ownership and shared interests. The HRG system8 ensured employees had support when they ran into problems and channels for voicing concerns when they felt treated unfairly.

Signs of Weariness

When did we start to feel like we were weakening? There’s no clear boundary or defining event—some might point to the Ant Group incident—but my personal feeling is that it began around 2017. Qin Yan,9 however, posted warnings even earlier than that.

Growth momentum cooling down

After more than a decade of rapid growth, starting in 2017, the growth rate of internet and mobile users dropped to single digits and began to decline. By 2024, the number of users surpassed 1.1 billion, marking the near-complete penetration of the addressable domestic population—growth has essentially come to a halt. Most mainstream internet products had already established their foothold before 2017.

Most acquisitions ended in failure

As various business units within the group reached their peak, Alibaba began making strategic investments. Unfortunately, most of these acquisitions ended in failure. Here are a few examples:

• Local Services: In 2006, Alibaba acquired Koubei.com (口碑网), back when O2O was still in its infancy. In 2010, Dianping was still a small player and Meituan had just been founded. After a series of moves, Koubei was shut down in 2011. It was relaunched in 2015 with traffic support from Alipay but remained lukewarm and was eventually merged into the Local Services division.

• Food Delivery: In 2018, Alibaba acquired Ele.me. At the time, Ele.me and Baidu Waimai held over 50% of the food delivery market—solidly in the lead. Yet within two years, Meituan overtook them.

• Bike Sharing: In 2017, Alibaba invested in ofo. After ofo collapsed in 2018, it invested in Hello Bike, which now barely holds its ground alongside Didi’s Qingju and Meituan.

• Payments: Starting in 2014, Alipay’s offline payment business was continuously eroded by WeChat Pay. By 2017, it had lost direction. Even now, it’s painful to look back.

• Music: In 2013, Alibaba acquired Xiami Music and Tiantian Dongting, which together held a 21.9% market share—the top position at the time. Tiantian Dongting shut down in 2016, and Xiami followed in 2021.

• Video Streaming: In 2016, Alibaba acquired Youku Tudou with a 35% market share—ranked No. 1. Today, it has fallen to third place, with only about half the traffic of Tencent Video and iQIYI.

• E-commerce (International): In 2016, Alibaba acquired Lazada, then the leading e-commerce platform in Southeast Asia. It has since been overtaken by Shopee, and TikTok has taken a significant share of the market as well.

• Additionally, there were acquisitions of department store chains like Intime (Yintai) and Sun Art (RT-Mart) in 2017, costing tens of billions of RMB. By 2024, these assets began to be divested at a loss. We never truly transformed these traditional retail sectors—they remained what they were.

Over the years, the only acquisitions that could be considered successful were AutoNavi (Gaode) and possibly UC Browser.

I personally experienced both phases of Koubei. To be honest, based on what I saw at the time, it was clear it wouldn't succeed. Our approach was extremely simplistic and crude: throw money at operations to hit data metrics. As for the real fundamentals—recruiting merchants, building services, developing the product—those were seen as dirty work, exhausting, slow to deliver results, and had little short-term impact on KPIs. So nobody wanted to do them, or those who did weren’t rewarded with good performance reviews.

Coincidentally, in 2016 I was transferred to Alipay’s offline payments team to support the field effort. Within six months, I realized it was hopeless. The problem wasn’t the fierce competition or that WeChat was too strong—it was our performance evaluation system and company culture. We had lost the patience and resolve for long-term thinking. We were addicted to short-term stimulants. Team coordination and execution had become inefficient.

Few internal innovations have succeeded

Since 2015, few of the group’s internally incubated ventures have achieved real success. The strong growth of Alibaba Cloud was largely driven by the core e-commerce business’s infrastructure needs. Freshippo (盒马）’s momentum was supported by integration with Alibaba’s online user base, and Idle Fish (咸鱼) was essentially a spin-off from Taobao.

Aside from e-commerce-related spin-offs, very few businesses have succeeded from scratch. DingTalk is perhaps the only notable exception, and maybe the emerging browser app Quark could be considered another. The rest have either failed outright or are barely surviving.

Meanwhile, many new competitors have emerged externally, entering the market from different angles and significantly impacting the group’s core businesses. There are still market opportunities out there—but we are no longer able to seize them.

Deep-Rooted Internal Problems

Through conversations with people from different BUs, people are generally agree—since around 2017—on the internal issues and their severity at the company. The consensus: these problems are becoming increasingly widespread and serious.

I group the issues into three categories: people, compensation, and business. The following descriptions do not target any specific individuals, nor does fitting these descriptions necessarily mean someone has a negative impact.

1. People

Blind Faith in Outsiders

There's a tendency to overrate people from popular industries or companies. As soon as a sector gains attention, we assume everyone from there must be exceptional—offering them high salaries and senior titles. In reality, many of these individuals just happened to ride a wave; their actual abilities may not match the hype. Often, only the original core founders are truly outstanding.

Back in the day, Alipay held people from WeChat and Uber in excessively high regard, bringing in a batch of young, high-P-level hires. These newcomers lacked familiarity with our business, aimed for quick wins, and had no emotional connection to our mission. Their misguided decisions prioritized short-term results at the cost of long-term value—essentially killing the goose for the eggs.

At the time, we even joked: “Are these people undercover agents sent by competitors?” These parachuted middle managers often struggled to get supports and ended up forming their own cliques. Their goal, more often than not, was to wait out their vesting periods, and their actions were typically short-sighted. We do need fresh blood—but we must not be blinded by falshy resumes.

Newcomers Hiring Newcomers

There used to be a rule: if you hadn’t been at Alibaba long enough to absorb the company culture, you weren’t allowed to hire others. At some point, this standard disappeared. It’s now common to see fresh hires bringing in their own recruits—even some still in their probationary periods conducting interviews. I’ve always wondered: is the business really so urgent? Do these recruiters themselves meet Alibaba’s standards? And if they haven’t yet grasped what it means to be a Ali’s man, how can they identify and hire people who fit our culture?

Rise of the Wild Dogs

We used to classify employees according to both performance and alignment with company values.

The goal was to cultivate Bulls, eliminate Wild Dogs, and help Rabbits improve. But in recent years, KPI have become everything. The culture has shifted to: Whatever brings fast results with minimal effort—go for it. That naturally favors Wild Dogs, who cut corners, waste resources, and fight over credit. Over time, they rack up performance, win bonuses and promotions, and others either get assimilated or develop strong defenses just to survive. As a result, collaboration costs rise and trust erodes.

Pileup of Rabbits

In recent years, we’ve seen a growing number of both young Rabbits and old Rabbits. Some are demotivated by stalled business growth. Others have chosen to coast due to broken incentive systems or lack of hope. Some have survived because the company was too cautious to lay off staff during times of bad PR. Others are sheltered under internal factions. The bottom line: fewer people are truly getting things done, while procrastination and blame-shifting are becoming more common. There are even many who behave like Wild Dogs but with performance like Rabbits. (Dogs)

2. Compensation system

Pay grade inflation

Our compensation levels have been consistently inflating. On one hand, there’s imported inflation. Our salaries became less competitive due to the poor stock price, especially compared to fast-growing companies. Yet, we still want to attract talent and are overly confident in outsiders. As a result, when offering them a position that matches their level but the salary isn’t competitive, the only option is to raise the P-level (position level).

On the other hand, there’s internal inflation—to retain senior stuff, despite fewer business opportunities, we have no choice but to loosen promotion standards. I’ve seen many such cases. In the end people start to think “If this person can be promoted to this level, and that person too—why not me?” People like this may soon end up leading a business unit, and then influence even more people. Now that job levels are no longer disclosed, these conflicts and problems have simply gone underground.

Opaque performance evaluation

Job performance are directly tied to compensation, but they are are not made public. This gives managers significant room for manipulation. Although there are multiple levels of review and HR assessments, loopholes remain. It’s widely recognized that there’s often a gap between evaluation and actual performance.

This kind of opaque distribution system further encourages a "boss-centric" culture—where loyalty and serving the boss take precedence.

Back in 2014, our department piloted a system where everyone wrote their own performance self-assessment, and all results were made public for mutual evaluation and discussion. This effectively curbed common manipulative tactics, such as claiming credit for others’ work, presenting mere participation as leadership, shifting blame, or redefining performance indicators to one’s advantage.

Short-termism

In contrast to long-term thinking, more and more people now adopt a short-term mindset. The initiative to better motivate employee by decentralizing stock option issuance to individual BUs did not achieve its intended outcome. The company ended up buying back most of the options—often with little to no gain for lower-level employees. As a result, people can only focus on the short term.

Put it another way: we used to treat the company like home—working hard to generate revenue, save costs, and believing that every bit of extra value we created would be rewarded many times over in the future. But for today’s employees, even if they still think that way, it no longer seems to make a difference. It’s a vicious cycle.

3. Business

Unclear Strategy

In recent years, the company's vision for the future and its strategic direction have became less clear, and are gradually drifting away from the user needs. Top-down communication is either overly vague or reduced to simplistic numerical breakdowns, with few truly effective or inspiring initiatives. There’s a disconnect in the middle and upper management.

That said, some clarity and hope have finally begun to emerge with strategies like refocusing on users, returning to an internet-driven mindset, going all-in on AI, and exiting traditional offline businesses.

Wanting it all

In the past, we prided ourselves on being able to have it all, as if it were a testament to our strength and capability. But looking back now, it was actually the strong macro tailwinds at the time that supported all those ambitions.

Once the tailwinds stopped, the irrationality behind all those ambitions began to creep into our operations. In an environment of slower growth, intensified competition, and bloated teams, we still sought projects with low investment, short timelines, and high return.

As a result, we started seeing all kinds of actions that went against common sense. And the people driving these action were often either unprofessional or panderers.

Metric manipulation

How to achieve results quickly with minimal investment? The answer is operations (运营).10 Launch a campaign, and the metrics shoot up; selectively boost data, and the ROI looks perfect; metrics plateau? Just redefine the KPIs. Cross-channel traffic redirection and overlapping user coverage—I've seen all kinds of tricks. Consequently, the value brought by long-term product development ends up being neglected.

I’ve worked in operations for several years—I’m not against it. Different stages of a product’s lifecycle require different operational strategies to help it succeed in the market. But now, we've made operations the primary approach. Too often, the illusion of prosperity created by operations masks the hollowness of the product itself. We are addicted to it — but when the buzz fades, we're left with nothing but a mess.

Bureaucracy

Many level 8, 9, and 10 managers don't get involved on the ground or engage deeply with the industry. Once promoted to senior levels, they tend to talk in abstractions—recounting their past glories and speaking in lofty terms—yet provide little real value to the current business. Many of their achievements were the result of favorable macro trends and would likely be hard to replicate today, though they often attribute them to their own abilities.

Why classify this problem under business? Because this bureaucratic culture most directly impacts the quality of decision-making: it leads to unprofessional judgments, the dismissal of expert advice, and unrealistic performance expectations, all under the guise of slogans like how bold the people are, how rich the land is.

It also brings huge coordination costs—people constantly have to give face and trade favors; even when collaboration is pointless, they still go through the motions to maintain a good impression; people want to say no but don’t want to be direct; teams pretend to be working on B while actually trying to achieve A, dragging junior employees into endless busywork.

Another manifestation of bureaucracy is excessive reporting—up and down the chain, layer after layer. A lot of time is spent organizing data and preparing presentations, but when it’s time to make a decision, leaders hesitate or let things fizzle out.

The Root Cause

Alibaba once had a motto: “A group of people with affection and integrity doing something valuable and meaningful together.” I’m not sure how many people still remember this line, but it was one of the cultural values that originally drew me to Alibaba. When broken down, this sentence actually captures everything a great company should strive for:

• People with affection and integrity: Like-minded individuals brought together and united by shared values.

• Together: This requires systems and structures to ensure alignment and support.

• Doing one thing: Refers to clear strategic focus—essential for efficiency.

• Something valuable and meaningful: Speaks to a sense of mission—like making it easy to do business anywhere, and putting the customer first.

The various problems mentioned earlier all stem from this motto gradually losing its power. We’ve moved from a breakdown in systems, to the erosion of values, to unclear strategy, and finally to a culture where KPIs take priority over any sense of mission.

We are still a large company, but we’ve become mediocre. Values alone don’t guarantee success—but they are what help us get through difficult times together. In recent years, Alibaba has faced many challenges and setbacks. We’ve pulled through, yes—but what’s most unfortunate is that in the process, we abandoned our values. We didn’t overcome adversity by standing united and supporting one another. More often, it was simply because we had deep reserves and a large enough buffer to endure.

Values are a company’s morality. Systems are a company’s laws. Law is the lowest threshold of morality, but morality—when self-imposed—is far more powerful than being forced to comply with rules.

When the culture is lost, people lose their way. And when people lose their way, performance eventually follows.

1. Core Values Exists in Name Only

Since around 2020, it has felt like our values have gradually been weakened, and by now, they’ve become more of a decoration. Although they’re still listed in our OKRs, the detailed reviews are long gone, and evaluations have been minimized. How many newcomers still remember our core values? Do we ourselves still advocate them? My favorite was always the Old Six: Customer First, Teamwork, Embrace Change, Integrity, Passion, and Dedication.

But today, each of these values has gone bad.

Customer first has been replaced by boss first

Because your boss controls your performance ratings, bonuses, stock options, and promotion nominations, prioritizing your boss becomes the optimal strategy for survival—especially when there are few checks on the boss or information is asymmetric. If the boss also truly puts the customer first, then there’s no conflict between the two. But today, many managers focus mainly on carving up existing interests—trading performance for stock, chasing short-term results.

As a result, performance reviews and promotions have become personal tools of reward and punishment. What should have been a system for recognizing individual contributions has become a way for managers to select and train loyalists. Loyalty to the company is now often replaced by loyalty to the boss.

Collaborative teamwork now became internal competition

Where once we applauded the process, now only results count. When business was booming, cooperation meant shared success. But when external growth slowed and resources are limited, grabbing existing value became far more rewarding than creating new long-term value. Competition shifted inward: internal turf wars, poaching performance results, and ugly zero-sum games began to emerge. Over time, bad players crowded out the good.

Embracing change turned into cover up of strategic confusion

In the past, booming markets and endless new opportunities drove a healthy mindset of embracing change. In recent years, however, frequent leadership changes, unclear strategies, shifting policies, and a proliferation of short-term KPIs have made our work fragmented and directionless. No one can stick to long-term goals, and no one wants to "build for others to harvest."

Among senior leadership, there seems to be no accountability—if one project fails, they’re simply reassigned. When a general is incompetent, the entire army suffers. For junior employees, what gets you promoted or well-reviewed? Business innovation? Outstanding results? Trust from your manager? In times of constant change, the only reliable bet is building trust with your boss. And so, “boss culture” has grown stronger. We no longer call each other by nicknames, but by titles like boss, teacher, brother, sister.

Integrity has become a rarity

In a results-only culture, manipulation, deception, and falsified data have become routine. Those who master these tricks gain the upper hand. The company’s response to breaches of integrity has often been too lenient. The infamous “mooncake scandal” may have overcorrected,11 but overall, enforcement has swung to the other extreme—disconnected from what people believe is fair. This lack of bottom-line standards has made us afraid—or unwilling—to seek the truth, since truth can offend, and often reveals things people would rather ignore. People knowingly turn a blind eye to lies. They hype up projects to grab resources, inflate metrics during execution, and shift blame after failure. Those who master these three tricks thrive.

Passion is suffocated by forced, long work hours

Our predecessors proved that the contract responsibility system works better than collective farming.12 Passion comes from within—it can’t be forced. It is unleashed only when there is clear protection of property rights, fair distribution of interests, and a well-designed incentive system.

In recent years, with shrinking markets, overstaffing, and unfavorable capital environments, our formerly functioning systems have broken down. Since 2014’s all-in mobile, we’ve had multiple periods of 99613. From what I’ve seen, 996 brings no real value—it just makes people look busy. But we’re not manual laborers. Creativity is driven by inner motivation, not exernal pressures.

When there are proper incentives—whether it’s a sense of achievement or material rewards—and when the company ensures a just process, people will naturally work hard. In several project and teams I’ve been in, people voluntarily did "9116", either because the project was genuinely interesting, or because the goals and rewards were clearly set. When the work is meaningful, people will naturally work hard; when it's not, forcing overtime only rewards those who are good at office politics and faking.

Dedication gets you punished

Dedication means treating your work as if it were your own—enduring loneliness, working earnestly, and being willing to bear hardship. Sometimes it’s like farming: slow, careful cultivation over time. But today, a nomadic culture dominates. When a business shows promise, crowds rush in; when it no longer does, it’s discarded, and the honest people are left holding the bag. Factions in power run projects based on personal connections, applying soft budget constraints to their own people. As a result, many low-value projects persist.

2. Broken HR system

Why have company values become hollow and meaningless? At the root of it is HR’s failure to fulfill its responsibilities—and its loss of credibility.

Within an organization, HR plays two critical roles: that of a mentor and a supervisor. The mentor’s job is to guide employees to act with integrity; the supervisor’s role is to identify and stop those who harm the organization. HR is supposed to be the guardian and carrier of organizational culture.

Business leaders lead teams to deliver results, while HR manages people and oversees processes. Leaders are responsible for business outcomes; HR is responsible for upholding the culture.

In the past, HR would attend meetings with business teams, pay attention to people and morale, talk to employees, observe whether processes aligned with company values, and check whether employees were facing difficulties or whether managers were competent. When problems arose, employees were relatively willing to talk to HR and give feedback. Together, they could help build a healthier organization. HR was meant to be like a close friend for employees.

But today, HR seems to care only about results, barely showing concern for people. Their focus on values and principles seems to have been forgotten. Ask yourself: how many people today have had meaningful conversations with their HR? How many have even met them a few times? Has HR genuinely helped employees solve any issues? Do employees still see HR as a source of comfort—or more like secret police? How many people feel safe telling HR the truth?

When HR fails in its duty, the business process loses its checks and balances. When HR loses trust, employees lose hope

And what’s the root cause of HR’s failure? I believe it’s that top leadership places too much emphasis on performance results, causing HR’s priorities to shift. But if an organization’s systems begin to rot, and its culture starts to decline, how much creativity can you really expect from the people within?

3. Incompetent middle management

As mentioned earlier, many people happened to catch the right tailwind at the right time, and they were fortunate enough to seize the opportunity and rise to leadership.

Today we are in this new environment, opportunities for young talents are limited, as leadership positions are already saturated. But promotion cannot stop despite stagnating business. The middle and upper management ranks have become increasingly bloated, with a growing proportion of incompetence.

As a result, strategy breakdown and execution becomes difficult. Decisions are passed up the chain, but accountability falls on those below. Are mid-level leaders or managers? No one can say for sure anymore.

The Cure

My access will soon be revoked, and I’ve written all this not to complain. It takes courage to say these things, and I may offend many people. But I sincerely hope the company can face its problems and start to improve. Only by restoring our core values and strengthening our teams can we truly move toward a 102-year future.

Alibaba is a remarkable company. We've done many things that brought real, positive change to society. We’ve helped advance progress. Please continue creating things that are good and meaningful.

Many colleagues who’ve left have posted thoughtful suggestions to the company on the internal forum. Some of them were exceptionally well-written—but years have passed, and nothing seems to have changed. Someone once wrote: “If you sweep dust under the bed, there’s no more dust.” How sad, how laughably accurate.

I hope we can face reality and pursue real change. The internal forum should remain a place for advocation. Suppressing ideas is the same as suppressing innovation.

1. Restore Our Values. Reinstate our culture alignment assessments. The current scoring system is purely formality. We need group-level reviews of each item to encourage reflection and filter out those who don't align with our culture. Once the team is clean, the mess naturally disappears.

2. Rebuild the HR System. 360HR should return to serving employees. Reduce the weight placed on performance metrics—only then can HR act as a true balancing force within the organization. HR must lead by example, acting as a guardian of values. Introduce 360-degree feedback not only for managers but also for HR. HR needs to rebuild its credibility.

3. Reform the Management Structure. Dismantle the "boss culture" and encourage freer movement within the company. When employees have more choices, toxic teams and weak businesses dissolve faster. If strategy and tactics are clear, top-down execution can work; if not, we must foster and encourage bottom-up innovation.

4. Make Pay Grade Levels Transparent. Publicly disclosing compensation levels. Everyone should be able to withstand public evaluation. It would make it easier to spot the underqualified Level 7s, 8s, and 9s.

5. Transparent Performance Evaluation and Promotions. Don’t judge by what people say—judge by what they do. Performance, evaluation, and promotion are the clearest reflections of what we value and what we don’t. These must be transparent and open to public review. This isn’t about making salaries public, but about avoiding politics, favoritism, and credit-stealing.

6. Cut Redundant Personnel and Projects. Having worked in many departments, my honest feeling is that cutting one-third of the people wouldn’t fundamentally affect the business. Even cutting one-third of the projects wouldn’t cause major issues. Much of the work is unnecessary—people are just trying to look busy. When there are too many people and too little to do, meaningless work multiplies. Let’s do proper performance evaluations, identify the Wild Dogs and Rabbits, and part ways with those who need to go.

7. Dial Back Operations and Face the Truth. I’ve worked in operations for years and achieved good results. I’m not saying they are useless—but we need to de-emphasize it. Only when we dare to face the brutal truth beneath the glossy surface created by operations can we begin to take product development seriously and build solid foundations.

AI is here—embrace this era.

I hope the company can once again have blue skies above, solid ground beneath, and clear, transparent air all around.

I’ve said a lot—they are just my personal view. Let me end with a line borrowed from Qin Yan’s post: Written before leaving—hoping someone is willing to read it.

I’ve been with you for 15 years. May you endure for a century. Go, Alibaba!

The holy grail of value investing is buying high-quality companies at bargain prices. Such opportunities occasionally arise during crises unrelated to the companies’ fundamentals. Such crises not only send companies’ stock prices to attractive levels, but also offer valuable tests to their competitive strength.

I believe DiDi presents such a rare opportunity. The great thing about this idea is that as long as I’m directionally correct on the key business drivers, chance of permanent loss is remote and DiDi shares are easily worth multiples of the current stock price. Below is a simple back-of-the-envelope calculation demonstrating the large margin of safety.

DiDi’s share price has ranged between $3/share – $5/share since 2023. At an average cost of $4/share, this is around ¥134B market cap. Backing out around ¥30B in net cash,1 the enterprise value is only ~11x the EBITDA of DiDi’s China segment. Compared to Lyft’s 15x multiple, this is a punitive valuation — probably justifiably only if DiDi were to lose significant market share in China going forward.

And what does the market give us for free? A fast-growing international business with ￥90B GTV, 30% yoy growth and steadily improving margins; a 3% stake in Grab worth around ￥5B (likely undervalued at current market price); and a bike-sharing business with one-third of China’s market share.

My valuation model estimates DiDi’s worth at ¥500B to ¥800B, based on segment-level cash flow projections from 2025 to 2029. This implies a total return of 260% – 500% over the next five years, or 29% – 43% compounded annual return, as summarized in the table below.2

Key Business Drivers

At a high level, I have greater conviction in margin expansion than top-line growth for DiDi’s China segment—and the opposite for its international business. This is reflected in my model, which incorporates a range of assumptions for the lower-conviction drivers. Each of these key assumptions is discussed in detail later.

• In China, DiDi’s massive scale flywheel creates strong competitive advantages and a high barrier to entry for smaller players. I think investors are over-pessimistic about China’s competitive environment. A cut throat price war is unlikely over the next 5 years, given a more rational capital market. Partially this is attributed to a tougher stance taken by the government on excessive competition, which undermines its strategic goal of boosting domestic demand as a key driver of China’s future economic growth.
  
  DiDi’s competitive strength is well-tested by the cyber security investigation during which new user registration was suspended for one and half years. I believe the current 7/3 ~ 8/2 market share split between DiDi and aggregated small players is a stable equilibrium state, serving a solid foundation for DiDi to expand margin and capture future industrial growth.
  

• Given its dominant market share, DiDi’s future top line growth in China will be driven by industrial growth. Given that ride-hailing trip volume in China remains 3 to 4 times lower than in the U.S., I see substantial potential for further penetration in the Chinese market. With fare prices anchored to heavily regulated taxi rates and ride frequency improving on a long time scale, I identify penetration rate as the primary driver of DiDi’s GTV growth over next 5 years. I model GTV CAGR of 3% - 13%, translating to ￥350B - ￥550B GTV in 2029.
  

• Aiming at Uber’s 8% EBITDA margin in 2024, I project that of DiDi China expands to 6% by 2029, factoring in the lower-margin taxi-hailing contribution. With DiDi’s dominant position and expected rational competitive environment, I think this is achievable, say with 1% increase in take rate, 1% reduction in consumer incentives and 1% from operating leverage. Take rate expansion is supported by both the oversupply of drivers due to soft economy and the declining per-trip cost driven by rising EV ownership in China. Margin expansion due to operating leverage might be conservative as I suspect DiDi’s operation in China enjoys a greater local economy of scale than Uber’s global business.

• DiDi’s international business, mainly in Latin America, has scaled rapidly and built a strong multi-vertical ecosystem across ride-hailing, food delivery, and financial services. With fewer entrenched competitors and strong local execution, this segment is well-positioned for continued expansion. I expect its annual growth rate to remain strong, gradually slowing from 30% to 15% over the next five years.
  

  I believe the Latin American market will likely play out like the North America, where Uber and Lyft moved past the prisoner’s dilemma to achieve profitability. DiDi and Uber are expected to do the same—why hurt each other when Uber owns 10% of DiDi? I think Grab is a better near-term reference here, given its recent profitability and a similar market structure in ASEAN. Based on this, I project DiDi’s international segment will reach a 2%–4% EBITDA margin by 2029.
  

• DiDi has scaled back its non-core, cash-burning ventures. Following the sale of its smart auto subsidiary to XPeng in 2023, bike-sharing is now the only alive one in the Other Initiatives segment. There are signs the three major players in this space are nearing profitability, supported by recent price hikes and improving user engagement. To be conservative, I assume DiDi will continue to post a modest EBITDA loss of ¥1B annually for this segment. Combined with ￥2B in capex for bike purchases, I estimate total annual cash burn at ￥3B.
  
  Is there a risk that DiDi could return to wasteful capital allocation in the future? It’s possible, but I think it’s unlikely in the near term—at least not until it relists in Hong Kong and experiences a period of strong stock performance.
  

Why does this opportunity exist?

• DiDi is a Chinese company trading OTC. Most institutional investors stay away from OTC stocks. Not only it’s a Chinese company—DiDi was publicly and severely punished by the CPC. For investors already wary of China exposure, DiDi likely ranks as the poster child for China risk.
  

• DiDi is underfollowed. Since its 'voluntary' delisting from the U.S. in 2021, there has been no analyst coverage for three years. The company also deliberately stayed out of the spotlight following the cybersecurity investigation.
  

• Divergent cash flow streams easily misunderstood by market. On a gross basis, DiDi is trading at 123x PE. However, on a STOP basis, we find a growing China segment, an international business growing at 30% with consistently improving margin profile, and significant cutback on cash-burning ventures. I believe DiDi is at an inflection point where its profitability is about to become visible to the market.

  

Catalysts

• Relist in Hong Kong stock market
  Several signs suggest this process may already be underway. In May 2024, DiDi President Liu Qing — who is believed to be heavily involved in DiDi’s IPO — stepped down as president and was 'promoted' to permanent partner. In October 2024, China Renaissance, one of the underwriters of DiDi’s IPO and the firm that led the DiDi-Kuaidi merger, initiated its first coverage on DiDi. In May 2025, the CSRC Chairman publicly stated that they would create conditions to support the return of high-quality U.S.-listed Chinese companies to the domestic and HK stock markets.
  

• Uber’s continue strong stock performance
  Before 2022, Uber was plagued by negative headlines—clashes with local authorities, conflicts with taxi unions, scandals around CEO, and widespread skepticism about the viability of its business model. Since then, however, Uber’s stock has surged nearly fivefold, and the company generated $10B in earnings last year.
  
  SoftBank clearly exited Uber too early—likely not by choice. It has invested around $12B in Didi and holds a 20% stake, meaning Didi would need a $60B market cap—close to my bear case valuation—just for SoftBank to break even.
  
  In many ways, DiDi’s story mirrors Uber’s: similar regulatory headwinds, the same tendency to burn cash on non-core ventures (not surprising, given Masa’s involvement in both). This resemblance, combined with Uber’s recent success, could help shift investor sentiment and lead to a market rerating for DiDi.
  

• Stock buyback
  DiDi spent approximately $850 M on share buybacks in 2024. In March, management announced an another 24-month $2 billion repurchase plan. If fully executed, this would represent about 10% of DiDi’s market capitalization.

China Mobility I: Competitive Strength

In this section, I’ll explain why I believe DiDi’s competitive moat in China is stronger than perceived, positioning the company to expand its EBITDA margin by 300 basis points. This would effectively double its EBITDA, even assuming a flat top line.

Broken Bear Thesis

Before diving into DiDi, it’s worth briefly mentioning the once popular bear thesis on ride-sharing performs, as it is such a classic example of what Buffett famously said “the rearview mirror is always clearer than the windshield”.

Back in 2017, the year Uber founder and CEO Travis Kalanick resigned amid a series of scandals, Len Sherman published a Forbes article arguing that ride-hailing platforms were inherently unprofitable.3 Over the following years, Sherman and others doubled down on the idea.4 In 2022, a short thesis on Uber appeared on VIC5 —arguably one of the worst-written post in the club—but it reflected the skepticism among investors at the time. Ironically, Uber’s stock price went up 5x since then.

The core of the bear case was that ride-sharing platforms lacked strong network effects, resulting in low switching costs and an inability to raise take rates without losing drivers or riders. But network effects or not, this thesis has been clearly disproven by the consistent rise in take rates—and EBITDA margins—across all major ride-hailing players.6

I believe the critics made a fundamental error: they misattributed the price wars—classic examples of the prisoner’s dilemma that have played out across many industries in history—and the wasteful capital allocation, a byproduct of a time of excessive liquidity, to inherent flaws in the ride-sharing business model.

Massive Scale Flywheel as Moat

With 70% ~ 80% market share, DiDi facilitates 12.4 billion rides annually and has become the default platform for nearly all drivers in China. This dominant position gives DiDi a fundamentally different—and in my view, easier—challenge than its competitors: instead of acquiring new drivers, its primary task is retaining the ones it already has.

To enhance driver engagement, DiDi has increasingly ties drivers’ income ties to their ratings: 7

“If you skip work one day or don’t start driving on time during the morning or evening rush hours, it will affect your rating on the platform, and the number of ride requests you receive will drop significantly.”

— A DiDi driver interviewed by 36Kr

A key component of a driver’s rating is the fulfillment rate, which reflects how reliably a driver accepts and completes rides—a strong indicator of platform loyalty. Multi-platform drivers often cancel rides when receiving simultaneous orders from competing apps, a behavior DiDi now penalizes heavily.:

“If you cancel an order, the service score deducted from that one cancellation will take 500 completed rides to recover.”

— A DiDi driver interviewed by 36Kr

This system makes it extremely difficult for smaller rivals to poach drivers from DiDi. In practice, only low-rated drivers—often with subpar service quality or non-compliant vehicles—opt for alternative platforms. This creates a virtuous cycle for DiDi: high-quality drivers are retained, user satisfaction improves, and the platform's brand reputation continues to strengthen.

The scale brings DiDi’s two more advantages:

First, the traffic and ride data it has accumulated over the past decade is an invaluable asset. This data powers superior algorithms for dynamic pricing, driver-passenger matching, and arrival time predictions—features that directly affect user experience and are difficult for smaller players to replicate.

Recently, Professor Len Sherman proposed that Uber’s profitability is largely driven by its AI’s ability to apply price discrimination on both riders and drivers, maximizing its margins.8 If that’s true, it only underscores the potential value of DiDi’s large dataset. Higher margins mean more resources to reinvest in product improvements and to fend off competition.

This same data also creates a strong foundation for expanding into future value-added services, and DiDi’s accumulated mobility data gives it a strategic edge that autonomous driving companies cannot easily bypass.

Second, ride-hailing is a highly localized network business, where new entrants typically have to build up one region at a time. DiDi, on the other hand, has already reached critical mass in all major urban areas in China. With national scale and resources, it can wage a prolonged competitive battle if necessary—creating significant barriers to entry for new players.

Prisoner's Dilemma Will be Tamed

History has taught us that the ‘prison dilemma’ eventually will be tamed via various structural changes and tactical response strategies. Competition eventually reaches a collaborative equilibrium.9 In fact, I think the current 7/3 ~8/2 market share split between DiDi and other aggregated smaller players is approaching such equilibrium state for the follow reasons:

• The aforementioned reward and punishment mechanism by DiDi effectively creates a structure change that avoid direct competition. Drivers with high quality service and users who would pay a premium for such quality gathers on DiDi while users who is more price sensitive chooses smaller platforms.

• Another structural change is a more rational capital market. After years of aggressive expansion and land grabs, internet companies are now slamming on the brakes. In August 2022, Huawei founder Ren Zhengfei published an internal memo warning that the global economy was heading toward recession and declining consumer spending. He stated that the company would shift its focus from scale to profitability and cash flow, making “survival” the top priority. In 2023, Tencent’s investments had dropped to just 37—its lowest point in a decade and less than half the number in 2014.
  

  This retrenchment isn’t limited to Huawei and Tencent. Since 2022, China’s internet giants have been pulling back. Companies like Alibaba, ByteDance, Baidu, and Meituan all halved their external investments compared to the previous year. Earlier this year, ByteDance even shut down its strategic investment division entirely.
  

• Behind this shift is also guidance at the national policy level. Over the past few years, the Chinese government has repeatedly summoned and advised companies across various industries to curb excessive, cutthroat competition. In 2024, Xi for the first time in a Politburo Standing Committee meeting, explicitly called for preventing involution-style competition. Why is the government placing such emphasis on this issue? Because excessive competition runs counter to China’s strategic imperative to boost domestic demand as a key driver of future growth.

In summary, I believe cutthroat price wars in the ride-sharing sector are unlikely in the next 5 years, the current 7/3 ~ 8/2 market share split is an stable equilibrium state.10 This presents an excellent opportunity for DiDi to further strengthen its position and improve profitability.

Is Aggregator Platform a Big Threat?

As some investors are worries about the aggregator platform, emerged together with small new players during the time of DiDi’s investigation, I will briefly comment on this here.

Essentially, a aggregator platform such as Gaode let user compare and choose offers from different ride-hailing platforms. Why did Gaode get a lot of traction around 2022? On one hand, Gaode was looking for monetizing its huge internet traffic from its map service (~800M MAU). On the other hand, small ride-hailing players desperately need user traffic to grow, eroding DiDi’s market share. It’s a perfect math for both sides.

However, I don’t not think it will pose a big threat to DiDi for the following reasons:

• The business model itself is flawed which inevitably leads to worse user experience, regulation problems and low service quality. Specifically, aggregators don’t directly manage or vet drivers, control driver incentives, dispatch algorithm and customer services. This means they can’t optimize wait times, pricing or issue resolution process as well as DiDi.

• Aggregators taking additional cut means a lower income for drivers given small players cannot afford to increase fare price. This is unattractive for well-trained drivers with better service.

• Gaode is not focused on ride-hailing business— it is just one of many attempts to monetize their map service. Ad revenues and O2O are the other two pillars. Further more, with Tsai and Wu stepped up, Alibaba is refocusing on e-commerce and cloud service. The Local Life segment, which Gaode was part of, is not Alibaba’s core business any more. Therefore, I doubt how focus Gaode is and how much it is supported to take more market shares from DiDi.

A Test of DiDi's Moat

It’s important to have some quantitative data supporting my argument.

In 2021 the Chinese government launched a cybersecurity investigation on DiDi. During that period, DiDi’s apps were removed from app stores and new user registration was suspended for one and half years. Study of how the market reacted to such external event offers a quantitative test to my argument on DiDi’s competitive strength.

During that period, competitors such as Caocao and T3, backed by major Chinese automotive manufacturer, rapidly expanded into new cities and offered substantial subsidies to attract both drivers and riders. Although the Chinese authorities urged everyone to stop “vicious competition”,11 DiDi’s market share was still declined from 90% to 70% by the end of 2022.

However, since user registration recovered in 2023 and the lifting of lockdown, DiDi’s MAU rapidly rebounded and already surpassed its 2011 level. According to various reports, DiDi market share recovered to around 70% ~ 80%. Recall that during that period, Chinese netizens frequently described DiDi as a “traitor” and “walking dog” of the US. How could there be new users when registration reopened, if DiDi’s product and service has no competitive advantages?

China Mobility II: GTV Growth

Many believe DiDi has limited growth potential in China given its dominant market share. While the hypergrowth phase may be behind us, I think the long-term growth of the ride-hailing industry remains underappreciated. In 2023, the traditional taxi-hailing market was still larger (¥323 billion) than ride-hailing (¥282 billion).12 I believe it’s only a matter of time before ride-hailing fully replaces traditional taxis, suggesting the ride-hailing market could at least double from here.

To study DiDi’s GTV growth potential in detail, I first break GTV into trip volumes and fair price.

Fare Price

There is very limited up room for fare price for two reasons: 1) it is anchored to taxi rate, which is strongly regulated. 2) According to Wind, from 2006 to 2021, per mile rate for taxi in 36 cities of China grown from ￥1.66 to ￥1.98, lower than CPI growth. Therefore, I expect DiDi’s average fare price grow at 1% CAGR and is around ￥26 per ride in 2029.

Trip Volume

According to the Ministry of Transport, total taxi and ride-hailing passenger volume in China reached 35.8 billion in 2024.13 DiDi’s China Mobility segment recorded 12.4 billion transactions that year. Assuming DiDi holds a 70%–80% market share, this implies a roughly 1:1 ratio between ride-hailing and taxi trips. In contrast, China Renaissance estimated this number to be 3:1 ~ 4:1 for the US.

If China were to reach a 3:1 ~ 4:1 ratio, assuming total taxi+ride-hailing trip volume remain constant, it would imply a 50% ~ 60% upside potential for ride-hailing. A 10% CAGR suggests this could be achieved within the next five years.

Trip volume can be further breakdown as follows:

Penetration Rate

There are more than one way to define penetration rate. Uber defines it as MAU/18+ population, and publishes this number for different countries in their investor presentations. This provides us valuable benchmark for our analysis. One alternative is to use urban population in the denominator. I think the later is probably more accurate as China has a much lower urbanization rate than developed countries.

I compile the following table based on the penetration rates disclosed for Uber’s international markets:14

The numbers for DiDi should be taken with a grain of salt, as it’s difficult to obtain accurate its MAU data, for several reasons. First, DiDi no longer discloses MAU figures. Second, users in China can access DiDi through both the standalone app and the WeChat Mini Program. I used the DiDi app MAU from QuestMobile, which likely underestimates the true number. Third, QuestMobile reports that DiDi’s WeChat Mini Program reached around 300 million MAU in December 2024 — a figure that seems unusually high. This number may be overstated, as it likely includes users of other services within the Mini Program ecosystem, such as bike-sharing and intra-city freight.

On one hand, If we use DiDi APP’s MAU of ~100M, Didi’s penetration rate in China is roughly equal to that of North America, but is quite low comparing to Australia and Brazil.The actually difference should be slightly larger, as we are comparing Didi’s 2024 number with Uber’s 2023 numbers.

On the other hand, China Renaissance pointed out that DiDi’s MAU reached 200 million around 2023. This figure might be more credible, as it was probably sourced directly from management. Based on a 200 million MAU estimate—and adjusting for DiDi’s and Uber’s respective market shares15—China’s ride-hailing penetration rate already surpass that of North America and be roughly on par with Australia and Brazil.

In any case, the table indicates that the global penetration of ride-hailing platforms remains below 20% — a quite low figure if we believe ride-sharing is fundamentally transforming how people travel.

Monthly Trips per Consumer

Regarding the MTPC, defined as total order number /MAU/12, the number is around 5 rides per user per month over the past two years (using 200M MAU according to China Renansance). I didn’t find any more fine-grained data from DiDi, but a survey from iiMedia back in 2021 shows that 50% of ride-sharing users contribute one or less than one ride per month. This distribution actually largely agrees with what Uber published in the investor presentation.

I expect the MTPC to further improve 1) as the rise of new generations with better adoption of share economy, and 2) as DiDi continue exploring more diverse niche use cases. For instance, DiDi launched a version specifically designed for 60+ users in 2021, and offered the option of riding with pets in 2024.

I don’t observe any consistent pattern suggesting MTPC improved much over the past 5 years (could be affected by COVID), suggesting it might be a long-term driver. I assume MTPC increased by 0.5 ~ 1 rides by 2019, contributing around 2% ~ 4% to the GTV CAGR.

2025 - 2029 GTV Growth Projection

In summary, I think the primary driver for DiDi’s GTV growth in the near term will be increased trip volume or penetration rate. Contribution from the fare price and MTPC growth won’t be more than 3% annually. I model penetration or trip volume grow at a CAGR of 0%, 5% and 10% for my bear, base and bull case, respectively. Assuming flat population growth (which seems to be expected if not decreasing) and negligible urbanization rate growth,16 this implies overall GTV CAGR of 3%, 8%, and 13%, translating to ￥349B, ￥442B, and ￥554B by 2029 in each case.

Is my projection of GTV growth rate too opmtisitc? Probably not, given even my bull case is lower than several quick estimations shown below:

1. GTV growth rate for Didi were 45.4% and 11.2% in 2023 and 2024, respectively. The 2024 slower growth is mostly like distorted by the abnormally strong growth in 2023, due to both the recovery from Covid and lift of government ban. Therefore, low teen might be reasonable for a normalized growth rate over the next few years. In contrast, Didi’s younger international business growing at 30%.

2. Frost & Sullivan projects China’s ride-hailing GTV will climb at 19.5% CAGR from 2024 to 2028, as the segment expands from 41.4% to 52.2% of total car-based passenger-transportation GTV.17

3. Lyft does not have delivery business thus serve a very good comparable. It’s GTV grown at 15% on average over the past two years. With only 30% market share in US and Canada, its growth rate should be a good lower bound for DiDi’s China business.

China Mobility III: UE & EBITDA Margin

Margin projection requires a clear understanding of DiDi’s unit economy and its financial model, as they are quite different from those of Uber, which provides the target for our projection.

Unit Economy

DiDi’s domestic business has two revenue streams which are treated financially different. For Ride Hailing, representing 79% of its GTV, DiDi considers itself as the principal and records the revenue on a Gross Basis. The other smaller contribution is from taxi hailing etc, where DiDi considers itself as an agent, recognizing the revenue on a Net Basis. For Uber, however, there is only one stream, which is the Ride Hailing and is recognized on a Net Basis.

Consequently, the comparable segment of DiDi and Uber, aka Ride Hailing, are treated differently in their financials. This is summarized in the table below:

To calculate the take rate for Uber is simple — just dividing its GTV by revenue. For DiDi, however, we need to divide its total GTV by the Platform Sale18 which is disclosed separately. Another issue is that the take rate calculating this way is a mixture of take rates of DiDi’s two revenue streams.

This point is demonstrated by the detailed decomposition of DiDi’s 2020 operating model:

The mixed take rate calculated from aggregated data is dragged down by the low margin of the Taxi Hailing etc business:

My point is that only the take rate of Ride Hailing is comparable to that of Uber. By taking out the taxi-ailing etc contribution, we have an accurate picture of the UE of DiDi’s core business (in 2020). The contribution of taxi-hailing recently is unclear as it was not disclosed, but I expect this part to keep decreasing in the long run.

Margin Expansion

Backing out the contribution from taxi-hailing, by assuming a constant 21.4% GTV contribution and 4.5% gross margin, I estimate the take rate for DiDi’s ride hailing is 24% in 2024.

Can DiDi in China eventually reach Uber’s 30% take rate? It won’t be easy—raising the take rate risks pushback from drivers, especially given DiDi’s dominant market position and the associated antitrust concerns. It is underway though — DiDi has recently introduced various reward cards (such as income boosters and commission waivers), which effectively reduce driver incentives.

Uber’s Mobility segment achieved an 8% EBITDA margin in 2024. For DiDi, I conservatively assume a 6% margin by 2029, factoring in the lower-margin taxi-hailing segment. Given DiDi’s competitive position and an expected more rational competitive environment, I believe a 300-basis-point margin expansion is achievable. For instance, this could come from a 1% increase in take rate, a 1% reduction in consumer incentives (which accounted for 11% of GTV in 2020 per the prospectus), and 1% improvement from operating leverage.

I'll close this section with a question: Given that DiDi's China operations generate roughly the same number of rides as Uber’s global business, should DiDi enjoy greater economies of scale and ultimately achieve a higher EBITDA margin than Uber?

China Mobility IV: 2025 - 2029 EBITDA Bridge

Summarizing previous three sections, for DiDi’s China segment, my model assumes 1) CAGR of GTV is 3%, 8% and 13% from 2025 to 2029 for my bear, base and bull case respectively, and 2) EBITDA margin expands to 6% in 2029 (with simple linear extrapolation).

Combine these together gives 2025-2029 EBITDA bridge. Below is an example for the base case:

International Business

DiDi entered the Latin American market in 2018, first establishing operations in Mexico through the city of Toluca and quickly expanding into major urban centers like Mexico City. Later that same year, it entered Brazil by acquiring the local ride-hailing platform 99 (formerly 99Taxis) for around $1 billion, marking one of its most aggressive international moves.

In both countries, DiDi rapidly grow its user base with its successful local strategy and execution. In Mexico, DiDi reportedly outpaced Uber in trip volume by 2021 in key cities outside of the capital, and in Brazil, 99 maintained strong traction, particularly in tier-2 and tier-3 cities. Since 2018, DiDi’s GTV has grown by more than 10x and shows consistent improving margin profile, suggesting a profit inflection point is around the corner.

DiDi’s Success Recipe for Latin America

DiDi’s success in Latin America stems from its strong local integration and responsiveness to regional issues.19 It formed partnerships with governments, NGOs, and security firms to tackle challenges like public safety and gender-based violence. In countries like Peru, Argentina, and Mexico, DiDi implemented tools such as safety partnerships and data portals for law enforcement. It also invested in community programs, supporting women and education, which helped build trust and strengthen its brand locally.

Additionally, DiDi strengthened its position in Latin America by offering social and financial support tailored to local needs. During the COVID-19 pandemic, it launched a $10 million global fund to support infected drivers and provided disinfection stipends and free rides for healthcare workers.

Beyond crisis response, DiDi introduced financial services like debit cards, microloans, and a no-fee credit card in Mexico and Brazil. Programs like Club DiDi offer discounts, insurance, and financing options, helping build a full-service ecosystem that deepens driver loyalty and sets DiDi apart from competitors. DiDi's financial services segment grew rapidly, with interest income growing by 140% and 79% yoy in 2023 and 2024, respectively.20

High Growth Momentum Ahead

I’m optimistic about DiDi’s future growth in Latin America for two main reasons.

On one hand, Latin America is an ideal market for ride-sharing economy. The total regional population in Latin America has surpassed 660 million, 82% of which will be living in cities. However, public transportation is often unreliable or insufficient, and many cities struggle with traffic congestion—creating natural demand for convenient alternatives like ride-hailing. At the same time, rising smartphone penetration and a flexible labor force make it easier for ride-sharing platforms to attract drivers and couriers.

I estimate that Uber mobility segment has around 30M MAU in Brazil and Mexico combined, implying total penetration rate of ride-sharing in LatinAm is no more than 5%-6%, leaving substantial headroom for growth.

On the other hand, DiDi’s opportunity in Latin America goes beyond mobility. Unlike in China, where sectors like food delivery and financial services are already dominated by giants like Meituan and Alipay, Latin America remains relatively open. DiDi can borrow successful experience from its competitors in China to build a multi-vertical business in the region.

Due to the large under-penetrated addressable market and DiDi’s promising multi-vertical strategy, I assume GTV growth continue its high momentum (30% yoy last year) and gradually slow down to 15% in 2029 in my model. This implies a ￥250B GTV in 2029.

Competitive Landscape and Road to Profitability

There is lack of official data on the market share of DiDi and its competitors in Latin countries. According to what I find from free resources scattered around, my estimate is that DiDi and Uber evenly split the Mexican market, while DiDi (99) holds around 30-40% market share in Brazil.21

I believe that in the end, the local players will either be acquired or merged with DiDi or Uber, as they simply don’t have the resources to compete with global giants.22 Eventually, DiDi and Uber will reach a truce and raise prices together—a win-win outcome. The Uber–Lyft dynamic in the U.S. serves as a clear guidepost: this is the only viable path forward. Let alone it makes little sense for Uber to try to hurt DiDi when it still holds a 10% stake in the company. The timing of this rational equilibrium is uncertain, but it seems to be near the horizon given DiDi’s consistently improving margin profile.

DiDi’s blended take rate (ride-hailing, food delivery and financial service) is 12% in 2024, compared to 24.7% of Uber, suggesting there is large headroom for margin expansion in the long run. However, I think Grab is a better reference in the near-term. Not only Grab just turned profitable in 2023, which I expect to happen very soon for DiDi, Southeast Asia market also shares several structural similarities with LatinAm. In 2024, Grab’s take rate and EBITDA margin is 15% and 4%, respectively.

In my bull case, DiDi achieves profitability in 2026 and reach EBITDA margin 4% in 2029. If the competition is more intensive than expected, EBITDA margin is 3% and 2% respectively for my base and bear case.

2015 - 2019 EBITDA Bridge

In summary, my model assumes 1) GTV growth of DiDi’s international business keeps high momentum and slows down to 15% in 2029, and 2) it turns profitable in next 2 - 3 years and achieve EBITDA margin various from 2% - 4%. Below is the base case 2025-2029 EBITDA bridge:

Other Initiatives ~ Bike Sharing

DiDi’s Other Initiative segment has been where DiDi tucks its wasteful projects. Following the sale of its smart auto business to Xpeng in 2023, there are three business left: Robotaxi (Voyager), intra-city freight (City Puzzel) and bike sharing (Soda).

Robotaxi & intra-city freight

I lump these two together because I believe both business are not very important to my thesis and are unlikely become profitable in the near future.

The robotaxi business focuses on autonomous driving R&D. Like Uber, DiDi’s strategy is to collaborate with car manufacturers. I believe this business is worth maintaining — even if DiDi doesn’t lead the next major breakthrough, staying close to the frontier keeps it well-positioned for future disruption.

As for intra-city freight, it seems like a natural extension of DiDi’s core business. However, I don’t expect it to be easy. Lalamove already holds around 50% market share, while DiDi has just ~5%. And that’s not even accounting for future competition from Full Truck Alliance.

Bike-sharing

Today, 95% of China’s bike-sharing market is dominated by three players: Hello, Meituan, and DiDi, with DiDi holding roughly one-third of the market according to various reports. There is limited publicly available data on this segment, as both Meituan and DiDi do not report bike-sharing results separately, and Hello has remained private since withdrawing its planned IPO in 2021.

Although none of the major bike-sharing players appear to be profitable yet, there are signs that the industry may be approaching a balance, with all three players now incentivized to prioritize profitability through coordinated price increases:

1. All three major players reportedly raised prices in several Chinese cities in 2024. Prices have increased from the original rate of mostly ¥0.25~¥0.5 per 30 minutes to the current rate of ¥1.5 to ¥1.8 per 10 minutes.23

2. Bike-sharing is seeing increasing user adoption. According to the Beijing Municipal Commission of Transport, in the second half of 2023, the total number of bike-sharing rides in the city reached 603 million, with an average of 3.3 million rides per day—an increase of 9.9% year over year. The average daily turnover rate of the bikes was 3.56 times.24

3. In March, YongAnHang, a smaller player in the bike-sharing market, announced that its original shareholder plans to transfer their shares to that of Hello Group. This is seen as the first step for Hello to list in domestic market via reverse merger.25

These developments suggest that (EBITDA) profitability of DiDi’s ride sharing may be within reach in the near future. However, it’s important to consider the capex of this segment, accounting for roughly 75% of the company’s total capex. Unlike ride-hailing, bike sharing is capital-intensive, as DiDi owns and operates the bikes itself.

According to a report I find, DiDi operates 6M - 7M bikes and each bike costs ￥700 -￥1000.26 With a lifespan of 2-4 years,27 I estimate DiDi’s annual maintain capex to be around ￥2B — roughly in line with DiDi’s reported additions to PP&E (bikes and batteries) in 202428.

Cash-burning assumption

Putting things together, I conservatively assume DiDi’s Other Initiatives segment has EBITDA loss no more ￥1B annually over the next five years. With ￥2B capex for bike-sharing, I assume DiDi’s Other Initiatives burns about ¥3 billion in cash per year.

Valuation of DiDi Global

Exit Multiples

With projected 2025 - 2029 EBITDA bridges from previous sections, exit multiples are the last elements to complete my valuation of DiDi. My main reference is the EV/EBITDA and EV/GTV multiples calculated from DiDi’s peers.

I believe a reasonable EV/EBITDA multiple for DiDi’s domestic and international segments should meet two criteria:

1. Lyft’s 15x should serve as a lower bound, as it has lower market share and operates solely in mobility, similar to DiDi’s domestic segment.

2. DiDi’s international business deserves a higher multiple than its domestic segment, as it will remain in an relatively earlier stage with greater growth potential.

Based on this, I assign a 15x exit EV/EBITDA multiple to the domestic segment and 20x to the international segment. As a sanity check, these assumptions imply an EV/GTV of less than 0.9 across both segments and all scenarios.

Sum-of-the-Parts Valuation

Applying the multiples and adding the estimated net cash accumulated over next five years, my model values DiDi Global at ￥500B to ￥812B. This implies a MoM of 3.5x to 5.8x over the five years.

Closing Remarks

In summary, I believe going long DiDi at current levels offers an attractive risk-reward profile. The competitive strength of its domestic ride-hailing business is under appreciated by the market. Combined with its OTC status, the stock is trading at an excessively depressed valuation—effectively giving away its fast-growing international business, bike-sharing operations, and a 3% stake in Grab for free. Based on my segment-level cash flow projections, I estimate a target price range of $14 to $24, implying an IRR of 30% to 43% through 2029.

A few words on DiDi’s founder and CEO, Cheng Wei. An interview conducted by Xiaobo Wu is a great source to get some insight into his character.29 My impression is that Cheng lacks the strong personality or charismatic leadership seen in people like Jack Ma or Richard Liu, suggesting limited independent thinking.

This can be seen by DiDi’s strategic decisions at nearly every stage closely mirror those of Uber. Even the personal anecdote shared in the Founder’s Letter is very similar to that of Uber’s founders.

I still remember that wintery night in Beijing in 2012. It was snowing hard. My jacket was no match for the wind. I wasn't alone. There was along line of freezing people, ahead of and behind me, all waiting with growing frustration for a taxi to take them home. This was a common experience for me since, like most Beijingers, I never had a driver's license. This night was different for me. Unlike the other people in line, I was not frustrated because I had a plan. We launched DiDi that year with the simple goal of making it easier for people to hail a taxi.

— Wei Cheng, DiDi’s Prospectus

However, this background also makes sense to me. Before founding DiDi, Cheng Wei was a manager at Alibaba—a professional manager rather than a grassroots entrepreneur like Jack Ma. From day one, DiDi grew under the heavy influence of capital interests. Cheng’s ability to survive and navigate through this landscape suggests he is skilled at balancing the agendas of various stakeholders. Given the nature of DiDi’s business and the current environment in China, a leader without a strong personality might, in fact, be the best one could hope for.

The author of this write-up owns shares in the companies mentioned and may purchase or sell shares without notice. This write-up represents only the author’s personal opinions and is not a recommendation to buy or sell a security. No information presented in the write-up is designed to be timely and accurate and should be used only for informational purposes. Readers of the write-up should perform their own due diligence before making investment decisions.

Introduction

A Discussion of Modernization is an English translation of Li Lu’s long essay. It was originally written in Chinese and is included as the first half of Li Lu’s book Civilization, Modernization, Value investing and China. The second part is largely a collection of Li Lu’s talks over the past decade, which I won’t focus on in this post. Interested readers can find the english version of these talks on Himalaya Capital’s website.

In A Discussion of Modernization, the author embarks on an exploration of human civilization’s evolution, tracing its journey from 60,000 years ago when Homo sapiens first ventured out of Africa, all the way to the present day, where the world’s economies are growing exponentially with no apparent limit.

So, why does the author explores such a broad topic, which may seem unrelated to investing? As both an intellectual who grew up in China and a value investor committed to intellectual honesty and fiduciary responsibility, the author seeks to comprehend China’s past, present, and future. Specifically, he aims to answer the following questions:

• Why did China fall behind the West in the 1900s?

• Why has China experienced such rapid modernization and economic growth over the past two decades?

• Will this growth continue, and if so, under what conditions?

To answer these questions, one must not only understand the nature and origins of modernization, but also how it differs significantly from earlier stages of human civilization. Only by grasping this distinction can one truly appreciate the forces driving China’s rise and the broader dynamics of global economic growth.

I find studying the magnificent history of mankind is an intellectual feast in itself. Whether you’re an investor or simply curious about the forces shaping our world, this exploration offers insights that can enrich your perspective.

Book summary

Modernization: a phase transition in the evolution of human civilization

Recent advancements in various scientific disciplines have led to the development of new historiography (methodology of historians), allowing us to study human activity far beyond the confines of written records. One of the most significant contributions in this area comes from Ian Morris, who developed quantitative methods to measure the trajectory of human civilization.

His research reveals that human civilization followed a slow and steady path for tens of thousands of years, until a dramatic phase transition occurred around 1800 AD. This shift marked the beginning of modernization and a exponential economic growth that continues to this day. The rest of the book delves into this trajectory, providing a detailed explanation of the forces and events that led to this transformative moment in history.

Civilization 1.0: Humans Venture Out of Africa

Around 150,000 years ago, our ancestors, Homo sapiens, survived the ice age thanks to the relatively warm climate near the equator. When the Milankovitch cycle reversed 70,000 years ago, favorable conditions began to emerge across the globe, prompting humans to venture out of Africa. Over the next 50,000 years, they slowly spread across the world, traveling at an average pace of one mile per year. This period, known as Civilization 1.0, was marked by hunting and gathering as humanity’s primary means of survival and production.

Civilization 2.0: The Rise of Two Centers of Civilization in the East and West

The agriculture civilization, or Civilization 2.0, originated in the Hilly Flanks and the Yellow River basin, regions where many of edible wild plants with large kernels and domesticatable animals were naturally found. These Middle Eastern and Southwest Asian areas gave birth to the two centre of human civilization for the next 13,000 years — what we now recognize as the East and West.

Over time, these advanced civilizations spread to other regions through either colonization or self-imitation. One exception is the America, which remained isolated by oceans, burying the seeds for the eventual rise of Civilization 3.0.

The social development of agricultural civilization faced repeated ceilings due to the inherent limits of energy production from farming. Whenever the population exceeded these limits, famines, epidemics, wars, and subsequent social instability would trigger periods of decline. The intermittent wars between nomadic peoples and agricultural civilizations lasted for thousands of years.

Civilization 2.0 reached its peak with several powerful Chinese dynasties that developed highly efficient and organized governance systems. In an era where production capacity was inherently limited, civilizations with the most effective organizational structures triumphed. However, the merit-based Keju (科举) system, designed to select government officials, was an incomplete solution. It served more as a ruling tool for imperial families than a truly meritocratic system for all.

From Civilization 2.0 to 3.0: The Discovery of a New World

Faced with poorer natural conditions and the perilous Silk Road, Western Europe began its quest for the rich and mysterious Orient by sea. However, around 1500, Europeans accidentally discovered the Americas. With their superior battle experience, advanced weaponry, and the deadly diseases, Europeans quickly conquered this new continent, which had incredibly favorable natural conditions.

The discovery of the New World immediately resolved Europe's land shortage problem. It also led to the development of a Trans-Atlantic trade network, which, in turn, gave birth to merchant-dominant governments and the Age of Enlightenment. During this time, intellectuals began to re-examine all areas of knowledge using scientific methods. These advancements laid the groundwork for the modernization and paved the way for the West to surpass the East.

Civilization 3.0: free market economy plus technology revolution

The merchant-driven Trans-Atlantic economy and the Enlightenment Movement reached their peaks in 1776, marked by two groundbreaking events: Adam Smith’s publication of The Wealth of Nations and James Watt’s invention of the first steam engine. These two innovations laid the foundation for modernization, or Civilization 3.0, combining the principles of a free market economy with advancements in science and technology. Together, they formed a self-reinforcing force that propelled humanity into a new era.

A free market does more than just generate wealth by enabling efficient labor division and the exchange of goods—it also fosters scientific and technological progress through the efficient exchange of knowledge. In Civilization 3.0, these processes are intertwined: technological advancements lower production costs, create new demands, and drive explosive economic growth.

A corollary of this, which the author calls the "Iron Law," is that the largest market will ultimately become the only surviving one. Interestingly, one of the major sources behind a business moat—economies of scale—also applies to trade markets. A bigger market means more participants and more exchanges, leading to greater wealth and more technological advancements, making it increasingly difficult for smaller markets to compete.

The Needham's Grand Question

The spread of Civilization 3.0 in the 19th century was such a traumatic experience for China that countless intellectuals and patriots pondered why modernization didn’t first take root in China despite once being the most advanced civilization. The author explains this paradox through the following observations:

• The chances of China accidentally discovering America were slim. The distance from China to the Americas across the Pacific Ocean is twice as long as the distance from Europe to the Americas across the Atlantic.

• Even if China had possessed the marine technology needed to cross the Pacific, there would have been little incentive to do so. At the time, China was already the most powerful and wealthy nation. In fact, during the Ming Dynasty, China boasted the most powerful fleet, with 240 vessels compared to Columbus’ 30. However, China banned maritime trade in 1492—the very same year Columbus discovered the New World.

• A free market economy was unlikely to develop organically in China, as the emperor's rule, supported by the Keju system, was so dominant both socially and culturally. Business activities in China have traditionally been shaped to serve the interests of the ruling class. This legacy problem continuous to influence China’s economic dynamics to this day.

The Modernization of China

After nearly a century of suffering from incompetent rulers and foreign invasions, China finally had the opportunity to chart its own path in 1949. What followed was another three decades of difficult experimentation and exploration. It wasn't until the Reform and Opening-up in 1978 and China joining the WTO in 2001 that the country experienced remarkable economic growth. Its GDP has multiplied by 60 times since then, with an average annual growth rate of 9%. This modern history of China reaffirms the Iron Law of Civilization 3.0, which states that the only way for any nation to grow its economy is by joining the largest free market.

The Chinese government played a crucial role in the early stages of industrialization by implementing a series of national policies that built essential infrastructure and integrated hundreds of millions of low-cost workers into the global economy. However, as labor cost advantages began to fade and high growth could no longer be sustained by exports alone, the Chinese government had to learn to temper its "visible hand" and allow the free market to more efficiently allocate resources. The future success of China will depend on how well it continues to balance government intervention with market forces.

Thoughts after the read

Be humble

If I’ve learned one thing from this book, it’s that every individual, race, and nation should embrace humility. As Nick’s father wisely said:

"Whenever you feel like criticizing any one," he told me, "just remember that all the people in this world haven't had the advantages that you've had.”

— The Great Gatsby

These 'advantages' are everywhere throughout history. If our ancestors hadn't lived near the equator, they may not have survived the Ice Age. If there hadn’t been an abundance of domesticable plants and animals, powerful empires wouldn’t have been built in those fortunate regions. If Columbus hadn’t accidentally discovered the New World, there would have been no Trans-Atlantic economy, and the dominance of England and later the USA may never have occurred.

I believe that recognizing how much of our perceived superiority in life comes from a series of fortunate accidents could help foster humility and tolerance—qualities that are urgently needed in today’s polarized world.

Do we need to study the entire evolution of human civilization to invest in China?

Not necessarily. And to be honest, even the world expert on such topics cannot predict the future of China, or of any country. However, I observe that every investor needs some sorts of confidence internally to sleep tight during the night. Where that confidence comes varies, depending on the investor’s background, thinking habits, temperament etc. For Warren Buffett, it could simply be his intuitive optimism (based on wisdom and experiences) on US’s economy. For Li Lu, I think he needs a more complex scholar-like probing to build that confidence and intellectual satisfaction.

Is the Risk of Investing in China Too High?

As mentioned in the previous section, China's future success largely depends on whether its government can strike a delicate balance between the 'visible hand' and the 'invisible hand.' This means the government should mostly play a limited role, allowing the free market to function, and only intervene when necessary, such as during a financial crisis.

Achieving this balance is probably not an easy task, given China's political system and its deeply rooted Confucian culture. Pessimists even wonder what would happen if China's leaders suddenly decided to withdraw from the global trade market. The author refutes this possibility in his book, arguing that the cost of such a decision would be too high.

However, this argument assumes that nations, in particular their leaders, always act rationally—a premise with no guarantee. If that were the case, empires throughout history would never have fallen. In fact, we don’t even need to look to historical examples of irrational decisions; just consider the US government’s recent moves to increase tariffs and seemingly pursue a path of isolationism.

In my view, any attempt to answer this question falls outside the scope of investing. Ben Graham taught us to focus on examining the past, not speculating about the future. And the past shows us that China has multiplied its GDP by 60 times, growing at a 9% annual rate over the past 40 years. Whether or not one is a fan of China's political and economic system, it's clear that it has worked quite well so far. Therefore, I believe a rational, contrarian value investor would not exclude Chinese companies from his/her portfolio.

The Chinese Market is Particularly Attractive to Value Investors

Not only do I believe that one should invest in Chinese assets, but I also think the Chinese stock market is actually more attractive to value investors than those of developed countries.

First of all, Chinese stock markets are more volatile, and this is music to a value investor’s ears. I see two main reasons behind this high volatility:

1. A larger portion of Chinese market participants are traders, not long-term investors. Their short-term focus and irrational behavior often drive prices to extremes—either overly inflated or excessively depressed.

2. Partially due to its political system and culture being so different, and partly due to its rising influence, China is heavily mistrusted by the West. As a result, foreign capital often overreacts to economic glitches, adding an additional layer of volatility.

The recent bear market is a great example of such volatility. Domestic retail investors began to believe that any government intervention in the market was simply a way to trick them out of their money. At the same time, foreign investors started to worry about the potential risks of war or the unsustainability of China’s growth—topics that pessimists have raised every few years.

Secondly, the Chinese government has strong incentives to build a healthy stock market that delivers satisfying long-term returns. On one hand, since China’s economic growth can no longer rely solely on exports, stimulating domestic consumption has become a top priority. On the other hand, China has the highest savings rate in the world (40%), and these savings are supporting a debt-dominant financing structure (with 70% of financing coming from bank loans). Therefore, encouraging equity financing is crucial to unlocking savings and stimulating consumption. Indeed, increasing the equity financing ratio is part of the 14th Five-Year Plan.

However, the success of this financial reform depends on the public's confidence in Chinese stock markets. This is why the government will do everything in its power to ensure the long-term health of stock markets. The rollout of registration-based IPOs, directing long-term funds (such as pensions and insurance) into the capital market, and encouraging state-owned companies to improve shareholder returns are just the beginning of these efforts.

More discussion about value investing and China can be found in Li Lu’s talks given in Peking University.

Wrapping up

Thank you for reading this far. I’d love to hear your thoughts on this book. Let me conclude with the epigraph of Security Analysis, which I feel is particularly relevant to the topics discussed in this post:

H World (HTHT) is one of the largest multi-brand hotel groups in China, with 257 million members and 912,000 rooms. Its three most popular and widespread mid-scale brands—HanTing, JI, and Orange Crystal—have allowed H World to ride the tailwind of China’s consumption upgrade. Such tailwind will persist as China moves beyond the Lewis Turning Point and transitions into a middle-class-dominated society.

The market has underestimated H World's leading position, largely because investors and analysts often rely on the total number of hotel rooms to measure market share. However, I believe this metric is incomplete. A better measure could be Total Room Revenue, defined as

By that standard, H World has overtaken its main competitors since the COVID, and is now the NO.1 hotel group in China.

H World enjoys a RevPAR premium over its main competitors, which translates into both higher operating margin and asset turnover. This premium ultimately drives its superior ROIC—20% compared to the industry average of 10%. Since 2017, H World’s ROE has consistently exceeded 20%. If not for disruptions due to the pandemic, 2024 would have marked the company's eighth consecutive year maintaining an ROE above 20%.12

H World's RevPAR premium is the result of a strong brand built over the past two decades, driven by a quality-over-quantity business philosophy. This approach is reflected in the company’s emphasis on self-operated hotels, its preference for developing in-house brands rather than pursuing large, inefficient mergers, and its focus on upgrading hotels—particularly within the midscale segment. Ultimately, this philosophy stems from its founder Qi Ji, who has founded three public companies in the travel and hospitality industry.

H World is currently trading at around 11x EV/EBITDA, roughly in line with its domestic peers. However, this valuation is low given the company’s stronger brand and higher ROE. It is also trading at a significant discount to well-known U.S. hotel brands—even after their recent 30% decline. In my view, part of this discount is justified, as U.S. hotel groups typically have a higher franchise ratio, a proven track record in international markets, and a much longer operating history.

However, I believe the remaining discount is largely driven by broader headwinds, including the ongoing weakness in the Chinese equity market and rising U.S.-China tensions. This is evident in H World's stock price, which has remained largely flat since 2017, despite the company tripling its revenue over that period and delivering ROE above 30% since 2023. As market sentiment improves, this valuation gap is likely to narrow.

That said, H World is also a growth play. A mid-teen growth rate over the next decade seems realistic, as the company continues to gain market share from competitors, driven by its RevPAR premium, and the hotel chain penetration in China keeps rising.3 Meanwhile, its international expansion is just beginning—offering a promising new growth avenue.

Last year, H World announced to accelerate its transformation to an asset-light model—this should further expand its margins. Centered around a strong brand, the franchise model requires minimal capital expenditure for growth, making it an ideal business model for value investors. In line with this, management recently unveiled a three-year Shareholder Return Plan, aiming to return up to $2 billion through dividends and share buybacks. That’s 20% of the company’s current $10 billion market cap.

H World — Best hotel group in China

Since the Reform and Opening-up in the 1980s, driven by rising household incomes and improved transportation infrastructure, the number of hotel rooms in China reached 16.9 million by 2023. At the same time, the penetration rate of hotel chains increased to 40%, which is on par with Europe’s average but still below that of developed countries like the UK and US (around 70%).

H World was founded by Qi Ji (the former CEO of Trip.com and HomeInn) in 2005. Ten years ago, H World had 217,000 hotel rooms; by the end of 2023, this number had grown to 885,600 (excluding overseas operations). Their impressive growth, both in absolute terms and relative to the industry, is summarized below.

I’ll explain why H World has enjoyed such superior growth later, but the table above gives a snapshot of the competitive landscape in China’s hotel chain industry. Over the past decade, Jinjiang (600754.SH), H World, and BTG (600258.SH) have been the top three players in terms of the number of operating hotel rooms. The rest of the market is fragmented with clustered shares, except for GreenTree (GHG), which has managed to hold the 4th position fairly consistently (though I believe GHG will soon be overtaken by Atour).

It’s important to note that several major players have merged or acquired others over the past 10 years. In 2016, Jinjiang spent ¥1.7 billion and ¥10 billion to acquire Vienna Hotel and Planteno (the most well-known brand being 7 Days Inn). In the same year, Beijing Travel Group acquired HomeInn (also founded by Qi Ji) for ¥1.1 billion. H World also acquired Orange Crystal in 2017, but it was a relatively small deal (¥3.6 billion) compared to those of its competitors. Therefore, be mindful that numbers in Figure 2 for Jinjiang and BTG before 2016 are synthesized.

An astute reader might ask: if Jinjiang holds the largest market share, why is this post focused on H World? In my view, the number of hotels or hotel rooms alone is an incomplete measure of a hotel company’s market share and competitive strength. Before I explain why, let me show you the ROIC4 for the top 3 players in China, as well as Marriott International:5

Exclude the impact of Covid, H World consistently achieves 20% ROIC since 2016, a level that is twice as much as its domestic competitors. This ROIC premium clearly demonstrates that the size of a hotel network alone is not enough. Other factors need to be considered to explain H World’s superior performance.

It’s interesting to note that H World’s ROIC has already reached the level of Marriott, one of the world leaders in the industry. However, H World’s business is much more asset-heavy compared to Marriott due to a higher percentage of leased and owned hotels. In fact, if lease assets were included in the calculation of invested capital, H World’s ROIC would be much lower than Marriott’s. This asset-heavy model also contributes to greater volatility in H World’s earnings, as shown in Figure 3.

That said, it’s important to remember that H World has only been around for 14 years, while Marriott opened its first hotel in 1957! As I’ll explain later, quality is far more important than quantity, especially for a franchise business in its early stage. Those who overlook this often find themselves on a path to self-destruction.

RevPAR premium — source of high financial return

The income statements tell a similar story: H World enjoys a 4%-5% operating margin premium over its competitors.

While H World’s gross margin is lower than that of Jinjiang and BTG—mainly due to a higher proportion of low-margin leased and owned hotels—its SG&A costs per dollar of revenue are much lower than those of its competitors. Does this mean H World is operating more efficiently? The unit economics shown below suggest that’s not the case:

H World spends 20% more on sales and marketing per room compared to its competitors. In fact, H World likely spends more G&A costs per room as well, as the numbers for Jinjiang and BTG also include a small portion from other businesses.6

That said, the unit economics do reveal the source of H World’s superior profitability: much higher revenue per hotel room. A higher proportion of leased and owned hotels has certainly helped, but a 2% greater share of leased and owned rooms can’t explain the 100% revenue premium per room over Jinjiang. The difference is even more striking when comparing H World to BTG: despite having 6% fewer leased and owned rooms, H World generates 60% more revenue per room than BTG!

Indeed, H World has been enjoying a RevPAR premium over its main competitors, for both leased & owned and franchised hotels:7

The RevPAR premium is also not a result of one company operating more high-scale hotels than another, as shown below:

H world has enjoyed a RevPAR premium for both its economy brands and mid-to-high scale hotels, suggesting its RevPAR premium is truly a brand premium. One problem with this comparison is that the definitions of economy and mid-scale hotels could be arbitrary. For example, Hanting started as an economy brand, but
with continuous upgrades, its rates might rise from 100 to 200per
night. Then one could argue that Hanting is now a mid-scale brand, not
comparable to Jinjiang’s economy hotels at lower price points.

To remedy this ‘problem’, I also show the overall occupancy rate of each company in Fig. 5, which clearly demonstrates the brand premium of H World — customers choose to stay at H World’s hotels despite they need to pay a higher price. Fig. 5 also indicates that the brand premium of H World over its main competitors is expanding.

Our quantitative analysis has identified the root of H World’s superior performance: the RevPAR premium drives both high asset turnover and operating margin, which together result in a high return on capital.

The RevPAR premium also creates a flywheel effect: more franchisees are attracted because the premium signals a high ROI. A larger hotel network not only boosts brand awareness and accessibility but also reduces owners' investment costs, further enhancing their ROI due to increased bargaining power with suppliers. This flywheel effect explain the faster growth rate enjoyed by H World, as mentioned in the first section.

A hotel brand with a RevPAR premium—reflecting both high average daily rates and occupancy—is only achievable with a strong brand reputation, built over years of delivering excellent guest experiences. Next, I’ll show you how H World achieved this by consistently prioritizing quality over quantity.

How H World built its Brand

Slow is smooth, smooth is fast. This Navy SEAL motto also holds the key to building a great hotel franchise. In this section, I will explain how H World has built a stronger brand than its competitors.

Self-operated Hotels as Flagships

So far, we’ve presented H World’s high mix of leased & owned hotels as a potential downside. After all, who wouldn’t want an asset-light business model with strong cash flow and low financial volatility? But this only works if the franchise model is sustainable. Just as a RevPAR premium drives growth, a declining or inferior RevPAR can reverse the flywheel, leading to a loss of market share and a lower return on capital.

RevPAR premium is the result of a strong brand reputation, built over years of delivering excellent guest experiences. Leased & owned hotels, fully managed by the company, play a vital role in that brand-building process. The CEO summed up this strategy well during the 2017 Q4 earning calls:

Actually, we see leased hotels still as an essential part of our overall strategy. They play a few roles:

1. By occupying the very good location, that these hotels continue to generate very good profit.

2. When we innovate on our different brands, the leased hotels is a good way to serve as the flagship for the franchisees to follow.

So overall speaking, we continue to invest in leased hotels. And we are very cautious in the site selection.

This strategy8 is clearly reflected in Figure 7 below.

Notice that the RevPAR of leased hotels by Jinjiang is consistently lower than that of its franchised hotels, suggesting a poor attitude toward brand quality control.

Economy Hotel Upgrades

Another key initiative contributing to H World’s RevPAR premium is the upgrade of its legacy economy hotel brand, HanTing. As household incomes in China rise, economy hotels with outdated and aging facilities have fallen out of favor. Owners' margins are squeezed due to increasing operating costs and low daily rates. In response, H World launched HanTing 2.0, 3.0, and 3.5:

In addition to a modern, high-end aesthetic, H World introduced the concept of modularization.9 Cabinets, desks, bed frames, etc., are all standard modules that can be selected and combined by the owners. This standardization significantly improves supply chain efficiency, reducing both renovation costs and timelines. HanTing 3.0 and above also integrate AI technologies, such as self-check-in, self-service laundry rooms, and robot deliveries.

Owners are highly motivated to upgrade their hotels, as newly renovated HanTing 3.5 properties see a RevPAR increase of ¥50 (a 30%-50% boost compared to HanTing 1.0), with occupancy rates rising to 90%-100%.10 By the end of 2023, HanTing 1.0 properties accounted for only 4.4% of the brand’s total.

In contrast, H World’s competitors, who have been trying to replicate this strategy, are moving much slower in upgrading their economy hotels. For example, around 100 7 Days Inn 3.0 properties have been completed, with 240 under construction—just 10% of its 3,000 hotels as of 2023.11 According to BTG’s 2023 annual report, 58% of Home Inn hotels were upgraded to NEO3.0.

Prudent Acquisition

One of the best ways to evaluate management’s capabilities is by looking at the deals they make. This is especially important in the hotel industry, where brand acquisitions are a common way for hotel groups to grow. The acquisitions and mergers around 2016 by the top three players provide a great case study of their management strategies.

H World spent ¥3B (12x EBITDA) to acquire Crystal Orange, the largest designer hotel chain in China’s mid and upscale segments. By the end of 2016, Crystal Orange had just 126 hotels — a small number compared to H World’s then network of 3,269 hotels. This demonstrates that H World’s acquisitions are not about expanding size but about enhancing brand quality and diversification.

In fact, H World prefers brands that haven’t yet grown too large, allowing them to be incorporated into H World’s culture and platform before further expansion. As the CEO mentioned in the 2016 Q4 earnings call:

…this brand still hasn’t expanded very fast recently through the franchise model. Currently the leased hotels still come for more than half of their portfolio.

In our more established brands like HanTing and JI Hotel, we only have 20% to 30% of the hotels are under the lease model. So believe Orange Crystal has a lot of room of growth through the manachised model going forward.

And our strong platforms are going to support through that extension process. …after they merge into our existing platforms, through more efficient back office operation as well as higher occupancies through the loyalty program, we have identified various areas that we can add value to the existing operation.

Today, the Orange brand has 1,030 hotels and, together with JI and HanTing, comprises the ‘Iron Triangle’ of H World.

Acquisitions made by H World’s competitors were far less prudent. For example, Jinjiang acquired Planteno and Vienna for ¥12B. The number of rooms doubled from 350K to 800K after the deal, but the largest brand under Planteno is 7 Days Inn, an economy hotel chain. Considering the trend of mid-scale hotel growth and the consumption upgrade over the past decade, I doubt the money was well-spent.

A major issue with large deals is the management challenges they create. Jinjiang is a state-owned company, which has a culture that differs greatly from that of Planteno. Even after the acquisition, internal consolidation didn’t occur until 2019 when Planteno’s CEO, Zheng Nanyan, left the company.12 Losing three years to a capable competitor is a permanent disadvantage! The merger between BTG and HomeInn suffered similar problem, so I won’t repeat here.

No Distractions

Last but not least, I’d like to point out that H World’s management is the most focused one among its main competitors. Jinjiang has a food and restaurant business. BTG operates a 5A-level tourist attraction in Hainan. GreenTree acquired several restaurant brands, including Daniang Dumpling, but all have underperformed. Managements like to argue that these non-hotel ventures create synergies with their hotel operations. However, the poor RevPAR of their hotels suggests otherwise. That said, Atour’s recent retail business seems to be working well—at least for now.

Growth prospect

I’ve spent enough time discussing H World’s competitive strengths, which are rooted in its philosophy of prioritizing quality over quantity. Once a strong brand is built and maintained, a flywheel effect drives organic growth, contingent on the industry’s growth potential. In this final section, I will explore how H World is aligning itself with the macro trends of China’s hotel industry and briefly touch on its international expansion.

Domestic Market

I took the figure above from Atour’s prospectus. The projection works surprisingly well—the hotel chain penetration rate in China indeed reached 40% in 2024. Assuming the rate gradually rises to match developed countries like the US, where it’s around 70%, the current low-teen CAGR will soon become unsustainable. A growth rate of 8%-9% would likely give the industry about five more years of runway. This serves as a lower bound for H World’s growth in the next few years. With its RevPAR premium, I’m confident H World will continue to take market share from its competitors, giving a higher-than-industry-average growth rate.

As I mentioned earlier, the mid-scale hotel segment has seen the highest growth over the past decade, and this trend will continue as China transitions to a middle-class-dominated society. A report from Mckinsey back in 201213 this societal transformation nicely:

The precise numbers from the charts don’t matter too much, as they are projections and depend on subjective definitions of social classes. However, the rise of the middle class in China is inevitable as long as China’s economy continues to grow. The report also points out that a significant portion of the growing middle class will come from lower-tier cities. In fact, this is where the hotel chain penetration rate is the lowest—around 20% in 2020, the latest data I could find online.

If you read H World’s earnings call transcripts over the past few years, the two strategic priorities they emphasize are mid-scale hotels and lower-tier cities. I believe H World is well-positioned to capture the macro tailwinds of the next decade in the domestic market.

Oversea Market

From the very beginning, founder Qi Ji envisioned H World as a world-leading hotel group. If you can read Chinese, I highly recommend his book, Founder’s Note:

It’s not the best book in the world, but it provides valuable insight into Qi’s impressive entrepreneurial track record, his aesthetic taste, and business philosophies. In his book, published in 2018, he shared the following:

So, there's no rush to open stores in the U.S. or Europe, or even in Southeast Asia. The priority is to first establish a strong presence in China. By diligently cultivating the richest land in China and reaping abundant rewards, we can achieve full market penetration. Only by becoming the leader in the Chinese market can we have the opportunity to lead globally.

— Translated by ChatGTP

In 2019, Qi likely felt it was the right time, and H World announced the acquisition of Deutsche Hospitality for €700 million. Two years later, the company’s name changed from Huazhu to H World—underscoring its ambition to expand internationally.

It’s still early to draw any conclusions, as H World’s overseas expansion is just beginning. Expanding into international markets won’t be easy—different local markets, cultures, and the need to build a global supply chain are all challenges. However, as long as H World continues to follow the same philosophy—quality over quantity, slow and steady growth—that led to its success in China, I’m confident the results will be great.

It appears that this is exactly what the European team has been doing: 61% of Deutsche Hospitality’s hotels are leased and owned, with only 10% growth over the past four years. The management is taking its time to experiment and understand the local market, without rushing to expand. This seems like a prudent strategy to me, especially in the wake of the disruptions caused by COVID.

Will Atour be a big threat to H World?

As mentioned before, I expect other major players in the industry, such as Jinjiang and BTG, to continue losing market share. A notable new competitor is Atour (whose founder was a former VP at H World), which has seen impressive growth recently. However, I believe its threat to H World is limited (at least for now) for several reasons:

1. Its recent profit growth has been largely driven by its retail business (selling neck pillows). While creative, the long-term sustainability of this business is uncertain.

2. Currently Atour focuses exclusively on the upper-mid scale segment, while H World has a much more diversified brands to contact consumers at multiple price points.

3. One key advantage of large-scale hotel chains is that prime locations in each city are limited. This will be a major problem for players late to the come such as Atour.

4. Atour has operated on an asset-light model from the start, with only 3% of its hotel rooms leased or owned. With a 30% annual growth rate, I see a risk of poor quality control, which could harm its brand

The author of this write-up owns shares in the companies mentioned and may purchase or sell shares without notice. This write-up represents only the author’s personal opinions and is not a recommendation to buy or sell a security. No information presented in the write-up is designed to be timely and accurate and should be used only for informational purposes. Readers of the write-up should perform their own due diligence before making investment decisions.

Founded in 2015, Pinduoduo (PDD) reached the trillion-yuan GMV club in under five years—a remarkable achievement in its own right and nearly miraculous given the duopoly in China’s e-commerce market at that time. For comparison, it took Alibaba 10 years (reaching ￥1.1 trillion GMV in 2013) and JD 13 years (￥1.29 trillion GMV in 2017) to reach the same milestone.

Pinduoduo’s (PDD) core business is an online B2C platform, aimed to offer value for money to its users. A widespread misconception, held by both consumers and even some of its competitors, is that PDD’s low prices are sustained by cash-burning subsidies or a compromise in product quality. However, this is far from the truth. PDD keeps prices low through two main strategies:

1. Directly connecting manufacturers and small, unbranded merchants with consumers, eliminating costs generated by intermediaries;

2. Optimizing unit economics through various ‘pin’ (aggregation/group) formats, including gamification, limited SKU offering, content recommendation etc.

Once PDD establishes itself in consumers’ minds as a value-for-money platform, its exceptionally high operational efficiency and low costs create a nearly impenetrable market position. This strategic playbook has been successfully applied not only in its main app but also in Duoduo Grocery and Temu.

On the surface, Pinduoduo’s (PDD) success can be traced to its well-known business strategies, such as gamification, large subside plans, and low cost operation. However, fundamentally, I believe PDD’s success is rooted in its core value — 本分, which I translate as be who you are. This guiding principle enables PDD to continually develop strategies that align with its own strengths, while left competitors wasting significant resources chasing tactics that don’t suit their own business models.

Concerns

While Pinduoduo (PDD) has a great company culture and continuously demonstrates impressive growth, I have several concerns:

• Unfavourable competition landscape

  In China, it competes with major players like BABA and JD, new forces like Douyin and Kuaishou. Internationally, it faces competition with Shein, Tiktok, Amazon and several e-commerce platforms in Southeast Asian market. The competition from some of these rivals is, in the words of PDD’s founder Colin Huang, ‘distorted’. In contrast, companies like BEKE, YMM have virtually no matching rivals and enjoy a long runway for growth (the only risk for them is self-explosion).
  

• Lack of diversification

  PDD’s businesses in my opinion is too vertical. Both DuoDuo grocery and TEMU are variations of its domestic B2C business. On the contrary, BABA possess a variety of high quality businesses across difference sectors. People might see PDD’s lack of diversification as laser focus, but to me, it suggests they simply don’t have many options.
  

• Plenty of Negative Sentiment Ahead

  First, the high-growth phase of PDD’s domestic business is behind it. As we all know, the curse of high-growth stocks is that any performance falling short of expectations can trigger a sharp price correction.

  
  Second, PDD's overseas venture, Temu, faces significant regulatory and geopolitical risks. The ongoing U.S.-China tensions hang over PDD investors like a ticking time bomb. Moreover, I believe these risks aren’t limited to the U.S. or the EU. Temu’s ultra-low pricing inevitably disrupts local businesses, and I see no reason why governments wouldn’t step in to protect domestic interests. As a result, I’m quite pessimistic about PDD’s international prospects.
  

• PDD’s financial reports are like black box

  Management did not discloses figures for PDD’s three segments separately, making it unclear whether Duoduo Grocery and Temu are profitable or how close they might be. Protecting sensitive information from competitors may be a reason, but given the company’s fiduciary responsibility to shareholders, it isn’t a strong excuse.
  

• No shareholder return so far

  The 39B of cash siting on PDD’s balance sheet also worth some discussion. There are only two reasons I can think of why the manage refuses to use that cash to enhance shareholder return. 1) They know they need it, either through subsidies or marketing, to remain competitive. If this is the case, that 39B is not all free cash. 2) They believe they can invest the cash on new ventures generating better return. Like I said before, I don’t see there much vertical opportunities out there, so most likely the cash will spend on other verticals, which is a huge risk to me.
  
  Some investors use Buffett as an example to defend DiDi’s no-dividend-no-buyback ‘strategy’. First of all, PDD has no proven record as an investor. Secondly, PDD is not an investment company like Berkshire and Constellation Software, for which there are in theory unlimited opportunities out there. Is it a common mistake for management to believe investing excess earnings will yield better returns than returning capital to shareholders?

Summary

At time of this writing, PDD is trading around $170B. With $16B forward earnings and 39B cash & investment, PPD is effectively trading at merely 8x multiple. This is obviously a bargain even if we assign zero value to Temu. However, I prefer to wait (to establish a large position) until management discloses official numbers for each business segment, pay dividends or buyback stocks, and until there is more clarity on risks associated with PDD’s overseas operations.

This is my one-page thesis on PDD. In the rest of this post, I will delve deeper into each of PDD’s business segments and elaborate on several points I’ve mentioned above.

Core business I — E-commerce platform built on SKU economics

An e-commerce platform connects merchants with consumers and generates revenue by charging service fees. This interaction drives a flywheel effect that can either grow or shrink over time. What sets one platform apart from another is the type of merchants and consumers it attracts. PDD’s flywheel is powered by SKU economics, where aggregated bulk demand from consumers meets merchants with substantial manufacturing capabilities.

In this section, I’ll explore this unique design by reviewing PDD’s early history. Analyzing the strategies that initially propelled the flywheel offers valuable insights into PDD’s recent ventures, where similar approaches are being applied.

Kick Off the Flywheel — Acquiring Users By Gamification

One key observation by Colin Huang, founder of Pinduoduo (PDD), was the highly fragmented nature of farms in China:

China has relatively less arable land per capita given its population and landscape. This is different from countries like the United States, where large-scale farms are prevalent, and the production and transportation of agricultural products could be highly industrialized.

— 2018 Letter to Shareholders

These small agricultural producers face high costs due to limited economies of scale and experience volatile prices driven by inefficient supply-demand dynamics across long distribution chains. By creating an online marketplace that connects small farmers directly with consumers, PDD aimed to address these issues. This vision formed the foundation of Pinduoduo.

The biggest challenge for any online marketplace at launch is getting the flywheel turning. Even if farmers were willing to sell on PDD, it wouldn't work without a critical mass of users.

To build its user base, PDD introduced an idea that transformed the e-commerce industry. Inspired by Groupon, PDD allowed users to secure large discounts by assembling a team to complete an order. The only way to grow a team was by inviting friends on WeChat. In fact, when PDD started, it didn’t have its own app and operated as a mini-program on WeChat.

What a powerful three-birds-one-stone move! By tapping into WeChat’s vast network of 1.34 billion users, PDD rapidly grew its user base at no cost.1 At the same time, the bulk orders generated through team purchases enabled merchants to lower their costs. Finally, a portion of these savings could be used to offer better prices, attracting even more users

Team purchases are just one of the many gamified features on PDD. For example, in Duoduo Orchard, users can ‘water’ their virtual fruit tree. Once the tree grows to the point of bearing fruit, they receive a free box of real fruit delivered to them. To earn more ‘water drops,’ users complete various tasks, during which they’re introduced to popular products.

By September 2016, PDD’s annual active users had surpassed 100 million, and its monthly GMV reached ¥100 million. Today, you can still find these games within PDD’s app, though they serve more as legacy features. They’re not as critical as they once were, now that the user acquisition phase is behind PDD. However, as we’ll see, similar tactics are being applied to TEMU to attract new users in overseas markets.

Alibaba’s Gift — A Defining Moment for PDD

In 2012, e-commerce giant Alibaba made the strategic decision to launch Tmall, a new B2C platform focused on well-known brands. This move allowed Alibaba to diversify from Taobao, which had faced criticism for selling low-quality and IP-infringing products. Additionally, large, established companies were better positioned to pay higher commissions and marketing fees, providing Alibaba with a more profitable revenue stream.

Tmall proved to be a major success for Alibaba. On November 11, 2015—often referred to as China’s version of Black Friday—Tmall achieved 91.2 billion RMB in sales within 24 hours, setting a new Guinness World Record. Tmall’s success significantly boosted Alibaba’s ad revenue, reaching ¥155 billion in 2018, as the company began charging higher fees.

This wasn’t good news for the small merchants who had once thrived on Taobao. Even more challenging, the platform’s algorithms began directing free traffic toward larger brands. As prime ranking and ad positions were increasingly dominated by big brands, small merchants saw their revenues decline, making them less likely to invest in advertising. This vicious cycle eroded small merchants' loyalty to Taobao, opening a door for PDD to step in and capture their business.

However, I believe the significance of this exodus is far greater than simply adding millions of new merchants to PDD’s platform. It allowed PDD to realize that its business model could be applied to a much broader range of merchants beyond small farmers. Many of these small, no-brand merchants own their own factories, giving them substantial manufacturing capabilities. They are an ideal match for the large demand that PDD excels at aggregating from its users.

Essentially, PDD gained a deeper understanding of its identity and established itself as the first choice for price-sensitive consumers and no-brand merchants with manufacturing capabilities. In doing so, PDD uncovered a market segment that was underserved by both Alibaba and JD. This competitive differentiation helped establish PDD’s moat. By the end of 2018, PDD had reached a GMV of ¥470 billion and 270 million monthly active users.

A SKU-Centred Platform

Since PDD’s flywheel is built on SKU economics, its app is designed to be SKU-centered, both on the front-end and back-end. Here are few examples:

1. Content recommendation/promotion as the main entry point. Unlike other B2C platforms where the shopping experience starts with a search bar, PDD focuses on continuously recommending and promoting products to users

2. No shopping cart. While most platforms today offer one-click ordering, this is the default setting for PDD’s customers. The shopping cart is a consumer-centered concept, and therefore, it doesn’t align with PDD’s business model.

3. Low-price preferred algorithm. On PDD, free traffic is directed to products with the lowest price. This contrasts with Alibaba, which directs traffic on large brands.

Traditional e-commerce platforms simulate a shopping experience by offering a large number of SKUs, including many similar products, allowing consumers to compare details and prices. However, PDD takes the opposite approach by hosting a much smaller selection of SKUs compared to platforms like Taobao and Tmall.2

Why? The answer is SKU economics, again! fewer SKU selection increases user spending per SKU, effectively aggregating the demands. Essentially, I view this as a hidden format of Pin, a strategy similar to Costco.3 A limited SKU selections also benefits small merchants, as they have limited resources thus are better off focusing on just 1-2 products.

Therefore, a search bar and shopping cart are not suitable for PDD’s limited SKU selection. Additionally, similar to TikTok, content recommendations make PDD’s app even more addictive, which in turn helps increase spending per user.

Core business II — ’Competition’ and growth

E-commerce platform can organically grow its revenue thanks to the flywheel effect. As more and more consumers and merchants join the platform, GMV increases and monetization rates can be also raised. However, such growth is possible only if the flywheel is not eroded by its competitors. Therefore, I will first address the competition PDD is facing in China.

A Competition PDD Won’t Lose

As PDD grows, it has come to be seen by Alibaba and JD as their main rival. Both companies decided to enter the low-tier city market, which PDD has dominated. JD launched its group shopping business in 2016 and introduced the JD Pingou app in 2019. Similarly, Alibaba followed suit by launching Taobao Discount in 2020. Both have also copied PDD’s promotions, such as the 10 Billion Subsidy Plan, on their own platforms.

Neither of these efforts have been successful. As of 2024, reports indicate that merchants on Taobao Discount were instructed to migrate back to Taobao.4 JD Pingou was rebranded as Jingxi and shifted focus to the Community Group Purchase sector in 2020. However, this business was eventually exited, and the team was dismissed in 2022.5 Recently, there have been reports that JD is attempting to revive its efforts in the low-tier city market, but I suspect it may face a similar fate.

These efforts failed because they overlooked the fact that e-commerce is a commodity business without differentiation. Both Alibaba and JD have their own unique strengths or moats: Taobao and Tmall are the go-to platforms for women seeking a shopping experience, while JD is known for its high-quality electronics, home appliances, and excellent services. However, by attempting to compete with PDD in low-tier cities or offering the same low prices, Alibaba and JD have dragged themselves into a battle where they have no distinct advantage. Specifically,

1. Alibaba and JD do not necessarily have deep resources or experience with supply chains and manufacturers, which is a key part of PDD’s business model

2. PDD’s market and strategy may not align with the established perceptions of these companies. This is especially true for JD, which is known for selling high-priced, high-quality products. Suddenly adopting a low-price strategy confuses its customers and frustrates its merchants.

3. PDD operates with much lower costs due to its flat organizational structure. Its revenue per employee is 3-4 times higher than that of its competitors:

Therefore, I believe there is no real competition between PDD, Alibaba, and JD. If there is any competition, it is one that has been forced upon PDD by Alibaba and JD. The rational approach would be to take the high road—work together to expand the overall market. Even if Alibaba and JD refuse to see this and continue to compete with PDD, they will ultimately be the losers, not PDD.

A quick comment on Douyin and Kuaishou’s e-commerce business: My understanding of this newly emerged format is limited. From what I gather, it seems that these platforms focus more on monetizing the relationship between influencers and their fans, rather than functioning as traditional retail businesses. I believe Yiming Zhang is a smart guy and appears to align with the win-win theory, showing little interest in directly competing with PDD.

Limited Upside in Raising Monetization Rate

Revenue in e-commerce is derived by multiplying GMV with the monetization rate, which includes transaction commissions and marketing service fees.:

As we can see, marketing services have been the primary income source, experiencing rapid growth as PDD established itself as one of China’s largest e-commerce platforms. PDD is known for maintaining a low commission rate, under 0.3% (compared to Taobao & Tmall’s ~1%).6 The increase in commission rate in 2021 was overstated by two factors: 1) PDD introduced a 1-3% commission for merchants participating in the 10 Billion Subsidy plan, and 2) Duoduo Grocery's revenue was incorporated into this category, distorting the figure.

Using the 2021 rate of 3.27% for marketing services and the 2019 commission rate of 0.33% (to exclude the effects of Duoduo Grocery and Temu since 2019), PDD’s overall monetization rate was around 3.5%. When compared to the current total monetization rate of Taobao & Tmall being around 4.3%, most likely PDD has limited upward room (at least for short-term).

This is especially true given that 1) items sold on PDD are generally low-priced, 2) non-branded merchants have lower spending power, and 3) the competitive landscape remains intense

Spending Per User Is the Main Driver of GMV Growth

With limited potential for raising monetization rate, PDD’s top line growth will hinge on its GMV. PDD’s user base has nearly saturated after years of rapid growth. In 2021, its annual active users (AAU) reached 0.9 B (last time they disclose this number). This number was 0.9B and 0.5B for BABA and JD respectively. Given that China’s total online user base stands at 1.1 B in 2024,7 it leaves only one growth pathway to PDD: increasing user spending.

Spending per user can be further broken down into two components: the number of orders per active user (#OPAU) and spending per order (SPO). Due to PDD’s lack of a shopping cart and its focus on low-priced items, #OPAU tends to be high while SPO remains low. In 2021, #OPAU reached 70 and could still increase driven by continued improvements in customer stickiness. I think there is a lot potential to raise SPO (which was around ￥40 in 2021) by optimizing the product mix.

Assuming PDD can’t increase SPO due to its focus on low prices is a misconception. Unlike a dollar store that caps prices on all items, PDD’s strategy is to offer the lowest available price for each SKU, regardless of its category. For example, an AC and a bag of apples might both be the cheapest options online, but the AC is much more expensive than the apples.

The 10 Billion RMB Subside plan, launched around 2019, was partially designed to address this problem. the plan not only enhances order values but also serves two additional strategic purposes:

1. Attracting Users in First-Tier Cities: Initially, PDD’s user base was concentrated in lower-tier cities due to its focus on low prices. By bringing more established brands to the platform, PDD could attract a broader audience in first-tier cities, intensifying competition with rivals like BABA and JD. This aligns with PDD’s philosophy of catering to a specific demand across various consumer segments, rather than focusing exclusively on a particular group.

2. Improving brand image. Like Taobao in its early days, PDD struggled with issues of low-quality and counterfeit products. By incorporating more reputable brands, PDD aimed to counter this image problem, helping to position itself as a reliable platform while maintaining its value-focused approach.

The 10 Billion RMB Subsidy Plan worked well with spending per order rising by 20% in 2019. This figure declined again from 2020 onward, likely due to the launch of Duoduo Grocery, which focuses on high-frequency, low-cost purchases. While recent trends are unclear, we can estimate the GMV of PDD’s B2C platform by assuming a constant marketing service rate from 2021 onward, as this rate reflects contributions from its B2C business alone. The result is shown below:

As shown, PDD’s high-growth phase in its B2C business may come to an end (likely a reason for stopping GMV disclosure since 2021). Going forward, we can expect steadier, moderate growth aligned with China’s e-commerce sector and economic trends. This explains PDD's pivot from a high-growth strategy toward one focusing on high-quality :

For the first half of 2024, we have remained committed to our high quality development strategy and consistently invested in our platform ecosystem … At this moment, the competition among e-commerce platform is quickly escalating, and therefore, it's even more important that we go back to the basics of supply chain improvements and invest firmly in supply chain efficiencies.

— 2024 Q2 earnings call

In the next two sections, I will discuss PDD's more recent ventures. As these businesses are still in their early stages, the competitive landscape remains highly dynamic. And the official data is also limited. Therefore, these sections will be relatively concise, focusing on explaining their business models and providing a qualitative analysis of their prospects.

Duoduo Grocery — Another Format of "Pin"

Duoduo Grocery operates as a Community Group Purchase (CGP) business, a model uniquely suited to China's local dynamics. I will begin with an introduction to the background that gave rise to CGP, followed by an explanation of its business model, and conclude with an overview of the competitive landscape and future prospects.

Background

China's high population density has given rise to a complex municipal organization. The country is divided into 56 provinces, 590 cities, and 1,800 counties. At the lowest level is the "community," a concept that differs from its counterpart in the U.S. In China, a community usually refers to a cluster of buildings in close proximity, typically managed by the same property management company.

Residents within a community are often familiar with one another and share common spaces for social interaction, such as convenience stores, grocery shops, barbershops, and small breakfast spots. These unique dynamics have laid the foundation for the emergence of the CGP model, which is designed to aggregate demand in the unit of such community.

Business model

The figure above provides a visual explanation of the CGP platform's operation. Here’s a breakdown of the process:

1. Order Placement: Users place orders daily, typically before a cutoff time, such as 11 PM. These orders can be made directly via the app or through links shared by the community lead (a role I’ll elaborate on shortly) in a WeChat group.

2. Order Processing: After the cutoff, the platform consolidates all orders and notifies the respective suppliers.

3. Logistics and Delivery: Products are transported through the platform's logistics network and delivered to each Community Lead, usually by 4 PM the next day.

4. Order Pickup: Users collect their purchases from the community lead.

The platform generates revenue from the price difference between selling to users and buying from suppliers or distributors. Essentially, CGP is a first-party (1P) business rather than a third-party (3P) one, making it more asset-heavy. This model is chosen for two reasons. First, delivering orders within a short timeframe, often less than a day, necessitates a logistics network specifically tailored for speed and efficiency. Second, the nature of the products sold, such as fresh produce, makes quality control crucial. By handling these items directly, the platform ensures they meet the standards expected by customers.

CGP is another form of "pin," thus consistent with PDD’s characteristic focus on SKU economics. Grouping within each community allows the platform to aggregate demand and place bulk orders, enabling it to source products further upstream in the supply chain at lower costs.

A central pillar of the CGP model is the role played by community leads, who are often local business owners such as shopkeepers, barbers, or breakfast stall operators. These individuals are crucial to the platform’s success, particularly in the early stages. By leveraging their social ties within the community, they act as an effective channel for acquiring new users. Last but not least, they provide logistical support by offering space to unload and organize groceries for customer pick-up. community leads are compensated with commissions for each order placed.

A Fast and Cheap Option

The concept of CGP was formed around 2015 when some offline supermarket chains started to explore this business model. It later entered a high-growth phase, especially driven by the pandemic lockdown. So, what demand does it fill, considering there are already fresh food providers, both online and offline?

The pros and cons of different fresh food providers can be best summarized in the Venn diagram above. E-commerce platforms like Freshippo (BABA) and JD offer same-day delivery, but due to the high costs of last-mile delivery and self-managed warehouses, they focus on selling high-quality items (such as organic and imported products) to remain profitable. Their customers are mostly from first-tier cities. On the other hand, CGP provides low-priced groceries with relatively fast delivery. This is the niche market CGP. Supermarket or farmers markets sell at a higher price comparing to CGP8 with no delivery services. They are the go-to places when freshness and variety are the priority for consumers.

So how does CGP platforms manage one-day delivery but keep their costs low?

1. SKU economics. By aggregating large demand through community-level grouping and limiting the number of SKUs available, the platform can pass on the savings from lower purchase prices to consumers.

2. Pre-sale model. Unlike PDD’s B2C platform, CGP operates on a true pre-sale model. Orders are placed only after the number of orders is finalized (usually by midnight), reducing warehouse costs and inventory risks. This fast turnover allows the platform to pay suppliers within 1-3 days, while supermarkets may take up to 2 months. Of course, this requires a highly efficient logistics network.

3. Self pick-up. A major cost for 1P platforms is fulfillment. By having users pick up their orders directly from Community Leads' stores, CGP platforms save significantly on last-mile delivery costs, which account for a large portion of fulfillment expenses (e.g., 70% of JD’s employees are couriers). The trucks only need to make a single bulk delivery to the local store each day.

Competitive Landscape

The first high-growth period for CGP began in 2017, with over 20 fundraising rounds raising 4B RMB for CGP companies. 2019 saw a phase of consolidation, highlighted by the first company bankruptcy (Dailuobo). The market was dominated by the "Big Three" at the time: NiceTuan, Fresh Orange, and Xingsheng Selected.

This landscape quickly shifted as internet giants recognized the opportunity presented by the pandemic, which altered consumer behavior. BABA, JD, PDD, Meituan, and Didi all sought a piece of the CGP market, with 15B RMB raised. As a result, the old Big Three were replaced by a new one: Meituan Selected, Duoduo Grocery, and Taobao Grocery.

The situation worsened as China lifted its lockdown and the government publicly criticized the cutthroat price war. Jingxi (JD) lost 8B RMB in 2021, and Richard Liu admitted the failure in internal meetings. Fresh Orange filed for bankruptcy in July 2021. Nice Tuan, once backed by BABA, shut down all of its CGP operations in March 2022.

It is widely agreed that today's CGP market is a duopoly between PDD and Meituan. This is not too surprising, considering that group purchasing was how both Meituan and PDD started. In fact, CGP is just a slight variation of PDD’s core business, as fresh food makes up more than 20% of the SKUs sold on PDD’s platform, and both businesses target price-sensitive consumers.

It seems that BABA and JD has not given up on the CGP market yet, as both are them reentering the space.9 However, similar to my views on the competition PDD faces in its B2C business, I’m quite pessimistic about these efforts.

Can Duoduo Grocery Support PDD’s Future Growth?

A more pertinent question is how much Duoduo Grocery can contribute to PDD’s bottom line. While Duoduo Grocery’s financial data has not been disclosed separately, we can observe that its main competitor, Meituan Select, despite some progress, has yet to achieve profitability:

Furthermore, it has been reported that Meituan Select closed its operations in Beijing and some remote areas,10 which suggests that the narrowing of losses is partly due to shutting down businesses in regions with little hope of becoming profitable.

Indeed, from various sources I’ve gathered, the profit margins for CGP platforms are very thin. The gross margin is around 20%, with operating costs including a ~10% commission paid to Community Managers and about ￥1 in fulfillment costs per item. With an average order price around ￥10, this leaves little room for profitability, and we haven’t even accounted for perishability costs, among others.

One way to further reduce costs is by paying less to community leads, which is already happening. Their income has dropped from above one thousand yuan per month to a few hundred.11 This is because CGP platforms are becoming less reliant on Community Managers to acquire new users. However, I’m not sure this is a sustainable approach. Recall that another key role of Community Managers is replacing last-mile delivery, which is vital to the CGP business model. What will happen if more and more Community Managers quit due to lower income?

Another classic strategy to improve profitability is adjusting the product mix to increase the price per order. Reducing fresh food offerings also helps lower perishability costs. This is likely already implemented, but I doubt there's much room for further adjustment, as fresh produce is a key demand from CGP platforms.

Back to PDD, I came across reports suggesting that Duoduo Grocery became profitable in some cities in 2022 and is expected to be fully profitable by 2023. I’m skeptical about this claim—if it were true, why hasn’t PDD disclosed it in their reports? The business has been around for nearly five years, so it not like there is not enough data to disclose.

One possibility I suspect is that PDD is no longer prioritizing its CGP business. It’s reported that the manager of Duoduo Grocery is now leading Temu. Even if the CGP business is, or soon will be, profitable, the margins are likely to remain very thin, as we’ve discussed. However, the logistics experience gained from CGP could be applied to more profitable opportunities.

TEMU — Oversea expansion

PDD began expanding into overseas markets in September 2022. As we’ll explore next, Temu’s business model is, in some ways, an asset-heavier version of Duoduo Grocery. The strategies that worked in China are proving successful in international markets as well. However, despite its rapid growth, I believe Temu faces significant risks from local government regulations

Business Model

Similar to Duoduo Grocery, Temu operates a 1P model. The platform selects and acquires merchandise from merchants and manufacturers across China. The inventory is stored in its own warehouse in Guangdong, and international fulfillment is fully managed and covered by the platform.

The 1P model is the preferred approach, at least initially, because:

1. Users in developed countries are more sensitive to product quality and authenticity. Having the platform in charge of selection and restocking makes quality control much easier.

2. International shipping is not economically viable for individual merchants. The cost can only be reduced to a manageable level if all inventory is aggregated and shipped at the platform level.

One major difference from Duoduo Grocery is that Temu also needs to cover the cost of last-mile delivery, which is very costly as I mentioned before. It seems like the only way to further reduce fulfillment cost in the future is to build its own logistic infrastructure like JD and Amazon.

Domestic Playbook Applied To Oversea Markets

Temu grows at a stunning rate in overseas markets. According to Sensor Tower data, it has been the most downloaded app in the U.S. every quarter since 2023 and has held the #1 spot for free apps on both the Apple Store and Google Play Store since 2024.12 o date, Temu has expanded into 49 countries worldwide. Remarkbly, Temu’s strategy is not much different from its playbook used early in China:

• Tapping into existing networks like Facebook and WhatsApp to fuel exponential user growth. This social fission effect is driven by encouraging current users to invite friends in exchange for discounts or free items.

• Heavy advertising expenditure to build brand awareness. Temu has spent aggressively on social media platforms, app stores, and even high-profile events like the Super Bowl. Reports indicate that Temu spent $2 billion on Facebook alone in 2023,13 and its Super Bowl ad cost about $10 million.14

• Consumer-friendly policies such as 90-day free returns and refunds, making it an attractive platform for shoppers.

Several articles, many from LatePost, report key operating data for Temu, including daily shipment tonnage, estimated shipping costs per item, HR expenses, and more. Since these figures are unofficial, I won’t place too much emphasis on them. It’s also easy to find sensitivity tables in research reports showing how variations in key drivers could lead Temu to profitability. The strategy to achieve this is fairly standard:

• Increase average price per item by adjusting the product mix.

• Reduce cash-burning advertising once the user acquisition phase is past.

• Shipping costs will decrease as more orders are placed.

• Shipping efficiency will improve with a more streamlined logistics system, including the construction of Temu's own infrastructure.

• Revenue diversification by blending 1P and 3P models.

I believe Temu can be profitable if it follows this playbook, but my main concern is whether it can execute this plan smoothly, without interruptions from competitors or other external factors.

Competition and Risks: Emphasis on Risks

Globally, Temu’s main competitors are Shein and Amazon. Similar to my earlier analysis of PDD’s domestic competition, I’m not overly concerned about the competition Temu is facing.

PDD’s management has consistently found ways to differentiate itself from competitors, and I believe Temu can do the same. Interestingly, the dynamics between Temu, Shein, and Amazon resemble the competitive landscape in China among PDD, Taobao/Tmall, and JD. Both Taobao/Tmall and Shein cater to a large female customer base, while JD and Amazon mainly operate a 1P model, selling books and high-priced appliances, while also building their own fulfillment infrastructure.

Therefore, if PDD can survive and thrive in competition with BABA and JD in China, maybe Temu can do the same in overseas markets.The situation in Southeast Asian with the fastest e-commerce growth in the world,15 is more chaotic. Some of the top players are Lazada, Shopee and Tokopedia. And let’s not forget about TikTok Shop.

However, if local governments prevent Temu from operating or competing freely, all of our previous analysis becomes irrelevant. Temu has attracted significant attention from local authorities as it continues to grow in popularity. The Issue Brief by the U.S.-China Economic and Security Review Commission summarizes the various concerns raised by local governments worldwide,16 including issues such as data privacy, product safety, forced labor, and intellectual property infringement.

I’m not overly concerned about issues related to product quality, safety, or intellectual property. These problems are quite common for e-commerce platforms in their early stages and tend to decrease as the platform enforces stricter policies and measures. Of course, if you believe that Temu is offering low prices solely because it’s dumping low-quality or problematic products, then you probably shouldn’t consider PDD as an investment either, and definitely wouldn’t be reading this far..

Some might argue that these accusations are driven by US-China tensions and are therefore US-specific, and that Temu’s fate could depend on the outcome of the upcoming election. Maybe. But I doubt this is fundamentally a political issue. Why? Look at the similar concerns raised by governments in Thailand17 and Vietnam.18 How can these be explained by geopolitical conflicts alone?

Therefore, I believe the primary reason behind these government interferences—and Temu’s biggest risk—is that its low-price model will harm local merchants and distributors. In China, this model can even be supported by the government because it lowers consumer costs and addresses the overcapacity issues faced by many local manufacturers. However, when the same model is applied in foreign markets, local governments see their own merchants and distributors suffer while Chinese producers and platforms profit. Given this dynamic, I see no reason why local governments, particularly in developed countries, would not intervene to protect their own interests. They will likely use all available tools, such as raising tariffs, launching propaganda campaigns, or even imposing outright bans.

Core values — true source of PDD’s success

I have reviewed all segments of PDD’s business, and it's clear that the company has achieved success across these areas in a relatively short period of time. On the surface, these successes can be attributed to several well-known strategies such as gamification and the pin model. However, I believe that at a fundamental level, these successes are a result of PDD's core values, which were instilled by its founder, Colin Huang. In this section, I will explain how PDD’s core values continue to drive the company to develop new strategies, make the right decisions, and ultimately achieve success.

本分 — To Be Who You Are

In his first shareholder letter, Colin laid down PDD’s core value as 本分.19 Like he said, it is difficult to translate it into English and there could be several layers of meanings. My translation of this word is to be who you are. More specifically, it means making decisions based on independent thinking, one’s own principles and responsibilities, rather than simply following others.

As a consequence of being who you are, PDD often leads the way, while its competitors end up following. A few examples include:

• Features like the pin-format and gamification, has been copied by Taobao Discount, JD Pingou, and Xiaoe Pinpin.

• PDD’s 10 Billion RMB Subside plan was later adopted by its competitors.

• All e-commerce companies had made low-tier cities a priority at some point, following PDD's success in these markets.

• Shein began selling grocery products, similar to Temu's offerings.

• Amazon launched Amazon Haul, a discount store, in an attempt to compete with Temu.

Making PDD's competitors followers puts them at a significant disadvantage for several reasons:

1. It gives PDD a head start, and sometimes that's all it takes to set the flywheel effect in motion and establish a competitive moat.

2. The best defence is a strong offence. The companies copying PDD are not on the attack but merely reacting, which puts them on the defensive.

3. Most attempts to copy PDD's strategies are doomed to fail because they tend to be superficial and based on a misunderstanding of the underlying principles driving PDD's success.

I’d like to expand on the last point with examples. For instance, competitors observing PDD selling products at very low prices and implementing eye-catching subsidy plans assumed they could replicate the same strategy. They started burning cash to offer discounts, only to find that it became a black hole, draining resources without meaningful returns, while PDD remained unaffected. What they failed to understand is that PDD's low prices were not achieved through cash burn but were a result of its pin format, direct access to manufacturers, and highly efficient operations.

Another example is the concept of the low-tier cities market, which gained popularity after PDD’s success. Competitors began to claim this market as their priority. What they don’t understand is that PDD probably never set out to specifically target low-tier cities in the first place. The large customer base in these markets for PDD is a consequence of its low-price products, which are in turn a result of its SKU-economics-based model. What PDD is actually targeting is the money-for-value demand of all consumers.

If we look at PDD’s strategic moves, they are all very logical and consistent. For example, even before entering the CGP business, PDD already had years of experience selling fresh food to people in low-tier cities and had mastered the pin format. Duoduo Grocery, in some ways, is a 1P variation of PDD’s 3P business. None of the other players have that advantage (maybe except Meituan). The same can be said for its decision to expand overseas while its competitors are still focused on battling in the CGP space. Temu, in essence, is another variation of Duoduo Grocery, applying a similar 1P model to a global scale.

Colin Huang Is a Disciple of Value Investing

I believe the core values laid down by Colin are closely tied to his strong influence from value investing. If you think about it, having 本分 (to be who you are) often aligns with being a contrarian. While I haven’t found much information on this in English, it’s widely known among Chinese investors. The influence of value investing on him is evident, as can been seen when he mentions intrinsic value several times in his first shareholder letter:

The pursuit and focus of our long-term vision and intrinsic value may not always translate into near-term profits.

you should expect a team with passion that is trustworthy and always focuses on serving users and our company's intrinsic value.

…it always follows the basic and simple principle—growing its long term intrinsic value.

— 2018 Shareholder letter

Colin’s exposure to value investing began through his connection with Yongping Duan, a prominent figure in the Chinese value investing community. Duan, the founder of BBK Electronics Group, is also a mentor to CEOs of Oppo, Vivo, and others. He gained recognition for his investment in Netease back in 2000, when it was at risk of being delisted from Nasdaq, yielding several hundred percent returns. Duan paid $620K for a lunch with Warren Buffett in 2006,20 and one of the guests he brought is Colin Huang.

Perhaps I'm biased, but I do believe this background gives me more confidence on Colin and PDD. Colin stepped down as CEO of PDD in 2020, and later as Chairman in 2021, to pursue research in food and life sciences. This was a rather unusual decision, especially considering PDD's market value had just surpassed that of BABA. But exceptional individuals often make moves that seem unconventional to the public eye.

One reason of stepping down given by Colin is:

First, the intensifying, and to a certain extent, distorted competition [emphasis mine] in some area has made me realize that the traditional approach of competing through scale and efficiency has its inevitable limitations. To change this, we must take action on deeper and more fundamental issues and seek answers from core technologies and the fundamental theories that underpin them.

— 2021 Letter to Shareholders

I think the judgment here aligns well with the main message of this post: the high-growth phase of PDD is in the past, and the company is now facing intensive competition and regulatory risks.

Am I worried that Colin's departure will negatively affect PDD's operations? Not much. To quote Jim Collins, founders of visionary companies are clock builders, not time tellers. Colin's willingness to step down at the peak suggests that he sees himself more as a builder of the company than someone who needs to be at the helm forever. Additionally, Colin remains the largest shareholder, so it's highly likely that he still has influence at the highest level. Finally, remember that PDD has a relatively flat management structure, which means fast promotion within the company is crucial for keeping employees motivated and engaged.

Summary

PDD is undoubtedly a remarkable company with a history of impressive success and a great corporate culture. At its current stock price, it trades at just 8x after-cash P/E multiples, which is undervalued even if we were to assume Temu contributes zero to its valuation.

That said, the company operates in a challenging environment with fierce competition and faces significant regulatory risks overseas. Given these uncertainties, I prefer to wait for more clarity regarding Temu's situation and increased transparency in PDD’s financial reporting before considering a large position.

This write-up represents only the author’s personal opinions and is not a recommendation to buy or sell a security. No information presented in the write-up is designed to be timely and accurate and should be used only for informational purposes. Readers of the write-up should perform their own due diligence before making investment decisions.

To be fearful when others are greedy and to be greedy only when others are fearful

And you have probably heard a more concrete version of it by Joe Kennedy:

If shoe-shine boys are giving stock tips, then it’s time to get out of the market.

— the Shoeshine Boy Theory

Reading these lines in books is one thing, but experiencing them firsthand is quite another. I personally witnessed such irrational behavior of market participants over the past few years.

When I returned to China in 2021, everyone I knew was buzzing about ETFs. One of the most popular one was called Sino-EU Healthcare. Its manager had become something of a celebrity, with a shining background including a bachelor’s degree from Tsinghua University and a PhD from Northwestern University, etc.

One friend of mine bought this ETF and mentioned he discovered it through a promotional page on Alipay. (Yes, you can purchase ETFs directly on Alipay without the need to open a brokerage account) To see just how wild things had gotten, consider one of the ETF's top holdings — Aier Eye Hospital Group (300015). After plummeting 90% from its peak in 2021, it is now still trading at a PE ratio of around 40.

Last year, in April, I traveled to Dali, a beautiful city in Yunnan Province renowned for its stunning mountains, lakes, and Buddhist traditions. Coming from Northern China, I struggled to adapt to the local cuisine, so I ended up dining at a small noodle shop every day.

This tiny eatery was a one-man operation. The owner, also the chef, would prepare my noodles and bring them to me slowly with his limp leg. He was clearly a kind and hardworking individual. Among other things, he shared that his son was getting married the following month, and he planned to close his shop for a week to attend the ceremony.

After a few days, I noticed that each time after he served my food (there weren't many customers in Dali at that time due to the strict quarantine policy), he would return to this desk to watch some TikTok videos on his phone and take notes carefully. It turned out he had been following an ‘expert’ teaching about investing in ETFs for several months.

It was meant to be a heartwarming story—a kind, hardworking man trying to learn how to become wealthy. Unfortunately, there wasn’t a happy ending: the stock market began to spiral downward soon after.

Fast forward to this January, I was catching up with the same friend who bought the healthcare ETF back in 2021. He told me (obviously after some reflections) that he now is certain that the Chinese stock market is a scam. Little did he know that was probably the best time to buy.

Indeed the human nature does not change. Recently, Chinese government announced a comprehensive economic stimulus package. Right after the press release, I conducted a quick poll, randomly selecting 10 posts on Xiaohongshu promoting the stimulus package. Out of 643 comments under these 10 posts, only 39 were positive. The rest were saying 'I won’t fall for this again.' in various ways. Just a few days later, as the stock market continued to rally, sentiment seemed turn around completely. According to the news, brokerages were bustling with retail clients rushing to open new accounts.

Background

While the US stock market continues to set new highs, the Chinese stock market is experiencing a significant correction. The world's second-largest economy has been severely impacted by its stringent COVID-19 quarantine policies. Consumer confidence is low, new home prices have plummeted at the fastest rate in nearly nine years, and the population has declined for the first time since 1960.

The situation in Hong Kong is also bleak. The region's economic and financial standing is weakening amid the rapid rise of mainland China. Coupled with the US interest rate hikes, the HSI has been on a downward trend for five years, following its peak in 2018. Major e-commerce companies such as Tencent, Alibaba, and JD.com are now trading at PE ratios in the low teens.

However, these pessimism factors may create unique opportunities for value investors. Property management (PM) companies operating in mainland China but listed on HKEX have seen their prices falling over 80% since 2021. This decline is largely due to their affiliations with real estate developers, many of which, like Evergrande, Country Garden, and Vanke, have faced bankruptcy or severe liquidity crises.

On the other hand, regardless of the future prospect of China’s real estate market, the hundreds of millions of existing buildings in China will always require management. Moreover, PM companies enjoy a great business model with light assets and stable cash flow. Consequently, it is highly probable that some of these PM companies are currently undervalued only due to their association with struggling developers, representing excellent investment opportunities.

In this post, I will present a screening and ranking process to narrow down approximately 50 candidates to a select group of relative high-quality, undervalued PM companies. Rather than recommending any specific stock—which would necessitate more intensive research—the aim here is to provide the readers with the necessary knowledge, including major risks and common pitfalls unique to this sector, to understand this sector and to begin their own research.

Screening: Filtering out those in trouble

Since 2019, there has been a surge of property management companies going public on the HKEX. Many of these companies are spin-offs from real estate developers. Although the official rationale behind these moves is often described as strategic transformation, I suspect that a hidden motivation was to salvage a profitable subsidiary from its faltering parent company before it was too late.

A cursory glance at some popular online investment forums reveals that many retail investors suspect these companies serve as vehicles for developers to legally move their money abroad. While conspiracy theories often captivate the public, this perception partially explains the depressed stock prices of these firms. Additionally, it is not an overstatement to say that conflicts of interest between the property management companies and their affiliated developers have existed from the outset. Therefore, assessing various risks posed by these related developers (or other parties within the same group) will be a central focus of the remainder of this post.

The outcome of aforementioned IPO surge is that we now have a list of over fifty property management companies to begin with. Clearly, we must narrow this list down before conducting any meaningful research. I employ the following screening criteria and provide explanations for each:

Remove those with negative earnings or operating cash flows since 2021

The crisis in China’s real estate market began when Evergrande defaulted on its promissory note in November 2020. Since then, numerous developers have encountered severe liquidity issues, leading to delays or defaults in payments to PM companies. Given the typically stable cash flows of the PM business, any company experiencing negative earnings or operating cash flows likely faced increasing accounts receivable or substantial write-offs. Therefore, companies that pass this filter are those less impacted by the developers' financial troubles.

Remove those who with significant declines of earnings or operating cash flows since 2021

PM companies in this category exhibit less severe symptoms. It is possible that these companies experience declining earnings due to a combination of 1) reduced revenue from affiliated developers and 2) a stagnating business. Given the large number of candidates, we can afford to apply stringent criteria to ensure we are left with the most resilient companies.

Remove those that did not pay dividends

PM companies that fall into this category either cannot pay dividends due to financial troubles or choose not to despite being profitable. In the first case, these companies are clearly not desirable, assuming they haven't already been filtered out by the previous criteria. In the second case, while opting not to pay dividends is not always a significant issue, it raises concerns in an industry not lack financial shenanigans, especially when many peers are paying generous dividends.

After applying these three filters, we are left with 10 companies:

I have also listed their PB ratios, Debt ratios and Payout ratios for references. Here are three observations:

1. All remaining companies either have zero or minimal debt, aligning with the light-asset model typical of PM companies.

2. They all have relatively high dividend payout ratios, indicating they are likely not cash-strapped. This lack of debt and substantial dividends are common characteristics of high quality companies.

3. With the exception of E-Start CM, the PB ratios of these PM companies are not particularly low. This implies that the market values them not as business in liquidation, but as ongoing businesses. In fact, most of these companies are experiencing year-over-year growth. This suggests that earning power is more appropriate choice in assessing their undervaluation.

After narrowing down our targets, the next step is to evaluate the relative attractiveness of the companies made to our final list. To accomplish this, I design a score system, which I will walk through in details. This scoring process is an excellent tool for exposing us to the nitty-gritty of this industry.

A five-factor score system

The idea is to identify several factors that comprehensively capture the characteristics of each PM company. Here is my thought process:

To start with, profitability and management quality are two mush-have factors for evaluating any business. Given that I’m examining PM companies with strong ties to their affiliated developers, I prefer those with lower risks related to their developers and with less distorted financial conditions due to connected transactions. Finally, I need an earnings multiple to assess the degree of undervaluation.

I then selected five quantitative factors to address these considerations and compiled the following table:

If any parts seem unclear, don’t worry—I will explain each factor in detail, including the rationale behind each choice and examples taken from this table.

Gross margin

This is a common measure of a business’s profitability. In the PM industry, a high gross margin might indicate that a company charges higher management fees, reflecting its brand reputation and service quality. Alternatively, a high gross margin could result from offering high-margin services beyond traditional PM functions, such as security, cleaning, and maintenance. This is where one PM company can distinguish itself from another.

For residential buildings, there are value-added services for both owners and non-owners. For owners, these services include leasing, parking, storage, and home renovation. For non-owners, services may include pre-delivery assistance, consulting etc. for developers. Such added-value services are also a common area where developers enhance profits for their affiliated PM companies, a topic addressed in the next factor.

Another example of value-added services is 'life services,' which can include catering for commercial buildings and operating community grocery stores. A notable example in this category is New Hope Service. Its related party, New Hope Group, is a Fortune 500 company specializing in agriculture, livestock, and food, with a particular focus on dairy products.

For commercial buildings, such as malls and office buildings, the company provides various services to the owners—in this case the developers, as these properties are usually leased rather than sold to businesses. These services include tenant sourcing, opening preparation, marketing, branding etc., and they generally offer much higher margins compared to traditional PM services. Consequently, companies that specialize in commercial PM, like E-Star CM, or those with a significant focus on commercial PM, tend to enjoy higher gross margins.

Revenue generated from related parties (%)

As I have mentioned several times, connected transactions are common practice in this industry. Let’s be fair: since many of these PM companies are spin-offs from their affiliated developers, some connected transactions are likely legitimate and unavoidable.

However, many of these transactions can be questionable, and investors should be cautious. For instance, there is a substantial risk that profitability might be inflated through these transactions, making it difficult to assess whether the prices paid are excessive or in line with industry standards. Therefore, I consider a higher percentage of revenue derived from connected transactions to be a negative factor.

A list of connected transactions can be found in the annual reports under the "Report of the Directors." More detailed information about each transaction is available in public filings, though my experience indicates that the annual reports usually provide the most useful details. To calculate this ratio, identify the total amount paid by related parties for the fiscal year, sum these amounts, and divide by the total revenue.

A review of the factor table shows that this percentage appears reasonable for most companies on the final list, with E-Star EM having the highest at 36%. However, this ratio could be as high as 90% for some companies three years ago.

To provide readers with an idea of what these transactions look like, let’s examine those of ONEWO:

During the year ended December 31, 2023, the total revenue from property services provided by our Group to members of the Retained China Vanke Group and their respective associates was RMB1,074.4 million, which did not exceed the annual cap of RMB1,078 million.

Framework Property Services Agreement — 2023 annual report

… the total revenue from value-added services provided by our Group to members of the Retained China Vanke Group and their respective associates was RMB2,382.8 million …

Framework Value-added Services Agreement — 2023 annual report

… the total revenue from AIoT and BPaaS solution services provided by our Group to members of the Retained China Vanke Group and their respective associates was RMB676.9 million …

Framework AIoT and BPaaS Solution Services Agreement — 2023 annual report

Together, these transactions account for 12.5% of ONEWO’s total revenue of 3.3 billion RMB in 2023. Notably, Greentown Service stands out as an exception, with almost no revenue coming from related parties. The specifics of Greentown’s situation are beyond the scope of this post, but it is probably the only PM company without a related developer and may warrant a separate, in-depth analysis.

Finally, I want to emphasize that this ratio alone cannot fully capture the distortions caused by connected transactions. For example, many connected transactions involve PM companies agreeing to purchase goods or services from related parties. If these sellers offer their goods or services at below-average prices, it could artificially inflate the profits of the related PM companies. However, I believe the current choice is sufficient since this post aims to be illustrative, not comprehensive.

Receivables from and prepayment to related parties (%)

Both this factor and the previous one aim to address the potential negative impact of connected transactions between PM companies and their related developers. However, this factor specifically highlights the risk associated with the financial health of related developers. If these developers encounter liquidity problems, PM companies with significant receivables from and prepayments to them may face substantial write-offs.

Management

A shareholder-friendly management can directly benefit investors through generous dividends or share buybacks. It can also indirectly benefit investors with lower chance of encountering financial shenanigans, which is crucial in this industry. Since all companies on our final list have substantial dividend payout ratios, dividends alone are not a useful factor for further differentiation. To keep things simple, I use the presence or absence of a history of stock buybacks as a proxy for the quality of management.

EV/EBIT with adjusted cash

Ultimately, we want to determine if a company is undervalued at its current price. Since the PB ratios for most companies on the final list are not low and they are quite profitable, earning ratios are likely the better choice. Why use EV/EBIT instead of P/E, especially when most companies have minimal debt? The answer to this question touches on an important issue that has not yet been addressed.

It turns out that all ten of these PM companies hold substantial cash on their balance sheets, which can obscure their true earning power. An extreme example is E-Star EM, which has more cash on hand than its market value. Therefore, only by using enterprise value can we accurately assess the earning power of these companies.

However, this is not the end of the story. We have already addressed one risk associated with connected transactions—the inflation of PM companies’ profits. It’s difficult to believe that developers would not seek something in return. What better target than those sizeable cash reserves?

So far, I have observed three common ways developers or other related parties exploit to access a PM company’s cash. Many of these are easy to overlook without a detailed examination of each connected transaction:

1. Loan agreements. This one is easy to understand. Let’s look at one example from China Res MIXC.
   
   ”According to the Offshore (and Onshore) Framework Loan Agreement, the Offshore (Onshore) Lenders may extend loans to the Offshore (Onshore) Borrowers in accordance with the terms of such agreement…
   — 2023 Annual Report“
   
   The annual cap of such loans was set to at 1,000 M RMB, and the actual maximum outstanding during the year was reported as zero. However, the annual report lacked one crucial detail, which is available in the public filing: “repayment shall be no later than six months after the date of advance”. This raises the possibility that the funds were borrowed but repaid before the end of the year. To err on the side of caution and to penalize such practice, I deducted 1,000M from China Res MIXC’s cash when calculating its EV/EBIT.

2. Financial services agreement. This method can be used when the PM company belong to a large group with its own licensed financial institution. An example is from Yuexiu:
   
   ”…the Company entered into the 2023 bank deposits agreement with CHB, pursuant to which the Group may, in its ordinary and usual course of business, place and maintain bank deposits with CHB and its subsidiaries on normal commercial terms from time to time. The daily caps … for the years ending 31 December 2023 … are RMB2,293,500,000… For the Year, the highest daily outstanding balance (including accrued interests) of the Bank Deposits placed by the Group with CHB Group on any given day amounted to approximately RMB985,366,000. — 2023 Annual Report”
   

   The actual deposit amount represented a quater of total cash on Yuexiu’s account. Similar agreements are found for Jinmao, Poly and China Res MIXC. To penalize such malpractice, I deduct the daily caps amount, rather than just the actual deposit amount.

3. Deposit in parking lot agreement. Parking lot service is a common value-added service offered by nearly all PM companies. Since the parking lots are owned by the developers, transactions related to parking spaces can be a convenient means of extracting cash from PM companies. Despite being suspicious, many of these transactions are beyond reproach, involving an exchange of parking spaces for cash. One might even argue that these funds are better utilized than sitting in a bank earning minimal interest.
   
   However, some of these transactions can serve as a front to conceal their true purpose. Poly Property Service provides an excellent example of such shenanigans:
   
   ”Pursuant to the Parking Space Leasing and Sales Agency Services Framework Agreement (Phase I), the Group will provide exclusive parking space sales and leasing agency services… to Poly Developments and Holdings and its associates… The Group is required to pay refundable deposits to Poly Developments and Holdings and its associates, and, at the same time, to receive agency service fees....

   
   For the year… 2023, the maximum daily balance of the deposit paid by the Group … pursuant to … Agreement (Phase I) amounted to RMB3,000 million, which did not exceed RMB3,000 million, the annual cap for 2023. — 2023 Annual Report”
   
   Guess what? There is indeed a Phase II Agreement with another deposit of RMB 2,000 M. Similar tricks was pulled by ONEWO for RMB 1,643M.

In summary, rather than creating a new factor to penalize these malpractices, I deduct a PM company’s cash by the amount (potentially) occupied by its related parties. This correction results in a higher EV/EBIT, achieving the same effect.

Ranking: Combine things together

The final step is to rank the 10 PM companies based on the factors calculated. There are various paths to proceed from here, but since this post is not focused on selecting the single best company, I choose the following simple procedure:

1. For each factor, rank the companies based on their attractiveness according to that factor.

2. Award +1 point to the top 5 ranked companies and -1 point to the rest.

3. Repeat step 1 and 2 for all five factors. Sum up the points each company receives to determine its final score.

4. Rank the companies based on their final scores.

When summing up the points, weaknesses offset strengths, effectively taking all five factors into account. For example, a company might receive +1 point for a very high gross margin but also get -1 point if a significant portion of its revenue comes from related party transactions. These opposing points cancel each other out, indicating that the company’s overall quality is not better than others, as its profitability may be inflated.

Here are the final results:

I highlight the factors worth +1 point in green and those worth -1 point in red. The final scores separate the companies into three groups: those scoring 3 points, those scoring 1 point, and those with negative points. In cases of ties among several companies, I rank them further based on their EV/EBIT ratios.

There is no single best way to use these results. One could construct a portfolio with the top three or the top five companies, or delve into the top choices and select a few based on personal criteria.

Limitations

I will conclude this post by addressing some limitations of the process I presented.

First, my initial screening criteria may exclude some potentially attractive companies. For example, several firms specializing in commercial building management are trading at very low multiples. While these companies recorded negative earnings in recent years, if there are signs of recovery, they could be promising opportunities.

Second, the five factors I selected are not exhaustive. For instance, Jinmao, Poly, and China Overseas are associated with large state-owned conglomerate. This affiliation could be beneficial, as the risk of these related developers going bankrupt is minimal.

Lastly, there is no definitive method for synthesizing these factors. Different weights could be assigned to various factors based on individual priorities. As an extreme example, for those who believe that one cannot make a quantitative deduction to allow for an unscrupulous management, companies pulled financial shenanigans should be avoided regardless of their scores on other aspects. Additionally, if a company surpass its peers in certain factor with a wide margin, such as Greentown, E-Star CM and New Hope, it might warrant more than the standard +1 point.

The author of this write-up owns shares in the companies mentioned and may purchase or sell shares without notice. This write-up represents only the author’s personal opinions and is not a recommendation to buy or sell a security. No information presented in the write-up is designed to be timely and accurate and should be used only for informational purposes. Readers of the write-up should perform their own due diligence before making investment decisions.

Thesis

CYD is a holding company and its main business is a 76.4% stake in Yuchai, China’s No.2 (mainly diesel) engine manufacturer for trucks, buses, construction, and agricultural machinery. Yuchai is a Sino-foreign business meaning the rest 23.6% shares are state-owned. CYD also has a 48.9% stake in a hotel operation in Malaysia, which will be ignored in this write-up due to its negligible (0.2%) revenue contribution.

Currently, CYD has net current assets of $611 million, and its stock trades for $339 million. This is a Ben Graham net-net stock with a large margin of safety. If the stock trades up to the net current asset value, the total return will be 78%.

Many net-net stocks today are problematic: either they are heavily in debt, or their businesses are unprofitable and declining. Neither of these issues apply here. Founded in 1951, Yuchai has built a reputation for high product performance and excellent customer service in China. It has generated profits and distributed dividends for the past 20 years. Its long-term debt ratio is merely 3.5%, which is very healthy for a capital-intensive manufacturing business. In fact, CYD’s cash is twice its total debt.

Over the past 10 years, CYD’s PB ratio never surpassed 0.5 — a figure that is too putative even for a cyclic business. I believe the main reason behind such suppressing multiple are two fold: 1) it’s Sino-Foreign structuremakes it unlikely to actively return value to shareholders, and 2) Chinese investors, who are familiar with Yuchai's business, are unable to invest due to its listing on the US stock exchange.

What happened for the past three years

In 2021 and 2022, Yuchai’s net income shrank by 50%, and CYD’s stock price dropped to $7 per share from its 2020 peak of $16. However, as I will detail below, Yuchai’s poor performance was a result of a series of industry-wide one-time event and national macroeconomic difficulties, not a fundamental deterioration of its business. In fact, Yuchai maintained its No.2 market share during this challenging time.

National VI(a) emission standards were enforced in July 2021. A pre-buy of cheaper and less strict National V vehicles in the first half of the year led to significant inventory build-up and much lower demand for National VI trucks in the second half of 2021. National VI engines not attaining economy-of-scale production contributed to a lower gross margin. To make things worse, in the second half of 2021, commercial vehicle unit sales declined by 36.5% nationwide. Meanwhile, Yuchai’s R&D spending kept increasing, preparing for the implementation of National VI(b) standards in 2023. As a result, Yuchai’s net income dropped by 47.6% in 2021.

In 2022, although the gross margin recovered to its 2020 level, national sales of diesel engines declined by 26%, back to its 2016 level. More strict area lockdowns and travel restrictions led to supply chain interruptions, construction declines, and weakened consumer confidence. Yuchai recorded a negative operating cash flow for the first time in 20 years, as did all other competitors. Things are slowly recovering: in 2023, national diesel engine sales increased by 23%, and Yuchai’s CFO was almost back to its 2020 level.

Yuchai’s industrial position is stable

Construction and agriculture are essential components of China’s economy, leading to a high demand for trucks, tractors, cranes, and other machinery, all of which run on diesel engines. The EV transition has had little impact on the heavy-duty engine market, at least for the foreseeable future. Additionally, there is further growth potential in exporting to new emerging markets. Overall, I believe it’s safe to assume that the demand for diesel engines in China will remain stable, if not grow, in the near future.

Based on the market share data from the past five years, Weichai (2338) has significantly outpaced its competitors and firmly held the top position in the market. On average, Yuchai has maintained the second position with a slight advantage. Weichai and Yuchai are the only two companies that have consistently remained in the top three in the market over the past six years.

With its brand name, financial prudence, 50 years of R&D investment, and rising technical barriers due to increasingly strict national emission standards, I have little concern that Yuchai’s can maintain its around 10% market share.

Increasing operating margin is expected

With increasingly strict national emission standard, diesel engine production is a difficult business. To meet the National VI(b) standard, Yuchai’s R&D increased at a CAGR of 37.4% from 2019 to 2021 . Since 2022, Yuchai may final be able to take a break:

Total research and development (R&D) expenditures, including capitalized costs, were lowered to RMB 1.0 billion (US$ 146.4 million) in 2022 from RMB 1.2 billion in 2021, as most National VI and Tier-4 engine development programs reached commercial production. — from 2022 Annual Report

In 2023, Yuchai's R&D expenses increased by only 1.9%. To predict how long this trend can continue, we need to consider the potential requirements of the National VII standards.

The China VI emission standards have already surpassed the Euro 6 standards in many aspects. Meanwhile, the development of the Euro 7 standards remains controversial. Under the current Euro 6 standards, vehicle emissions have been reduced to negligible levels. Many research institutions argue that Euro 7 is not the right approach: its environmental impact is minimal while the costs are extremely high. As a result, some experts predict that when China sets its National VII emission standards, it will take a more cautious and moderate approach. Therefore, I expect Yuchai will not need to significantly increase its R&D to meet the National VII emission standards, which are expected to be announced around 2026.

Lastly, Yuchai's gross margin increased by 20% in 2022 and by 3.7% in 2023. This improvement is due to achieving economies of scale and the reduction in unit production costs as National VI engines became widely adopted in China. Overall, with increasing revenue, stable R&D expenses, and further cost reductions, I expect to see a promising upward trend in Yuchai's operating margin.

Valuation

Balance sheet analysis

As I mentioned at the beginning, CYD is trading at a large discount to its net current asset value (NCAV). If the stock trades up to the net current asset value, the total return will be 78%. Below is the calculation with a bit of more details:

2025 forecast

Putting together my analysis of Yuchai’s industrial position and operating margin, I project Yuchai’s 2025 earning based upon following assumptions:

• Assuming its revenue recovers at a CAGR of 9.4%, the same rate as in 2023, Yuchai's revenue is expected to return to its FY'20 level by 2025.

• Using 10y average of 4.8% for operating margin. As analyzed above, Yuchai’s operating margin is expected to improve.

• Tax rate of 15%. Yuchai has been enjoying this reduced corporate tax rate under the Western Development Tax Incentive Scheme or the High Technology Incentive Scheme since 2008.

• Using a EV/NOPAT multiple of 8, which I think it’s fair for a cyclic business. (The median PE ratio for CYD for the past 10 years is 7.6x)

• Applying an 80% discount to its cash, as it seems unlikely it will be distributed.

Here is how the business would look at 2025 by using above assumptions:

At current market cap of $339 M, the upside is 173%. Of course, whether CYD’s stock will trade to its fair value or not is a big question mark, considering its depressing multiples over the past decade.

Conclusion

In summary, I think Yuchai's recent poor performance is largely due to several one-time events, such as the adoption of National VI standards, the impact of COVID-19, and the slowdown of China's economy. With Yuchai's stable industrial position and the large R&D expenses for National VI standards coming to an end, I expect Yuchai's profits to continue improving as the Chinese economy recovers.

A quick note on insider ownership: Hong Leong Asia Ltd holds 44.7% of the shares, and the Guangxi Government holds 18.2% (effectively 37% of Yuchai is state-owned). I didn’t find any directors or executives with significant ownership, so I don’t expect management to actively address the undervalued stock price. CYD has consistently paid high dividends, with an average payout ratio of 63% over the past five years, as this is how Hong Leong Asia collect cash from the company.

Since only 37.8% of the shares are publicly held, the stock is quite illiquid. Being a US-listed local Chinese business certainly doesn’t help, as Chinese investors who are familiar with Yuchai's business cannot invest. The illiquidity leads to significant volatility in its stock price. Looking at CYD's historical price, returns have been quite substantial (over 100%) as Yuchai's performance recovers each time.

Finally, I’d like to point out that value investing funds such as Brandes Investment Partners and Chou Associates are among the institutional investors.

The author of this write-up owns shares in the company mentioned and may purchase or sell shares without notice. This write-up represents only the author’s personal opinions and is not a recommendation to buy or sell a security. No information presented in the write-up is designed to be timely and accurate and should be used only for informational purposes. Readers of the write-up should perform their own due diligence before making investment decisions.

BEKE is a leading integrated online and offline platform for housing transactions and services in China. Despite being the 4th largest online platform in the country, BEKE doesn't get much attention compared to the Big 3. Value investors betting on China seem more focused on Alibaba, JD, or the unstoppable PDD. This lack of attention piqued my interest. Plus, I frequently saw BEKE's realtor shops while living in Beijing, and I've always been curious about its business model. This curiosity led me to conduct an in-depth study, the results of which I’m sharing here.

A successful business possesses a strong moat to fend off competitors and ample room for growth. In 2023, BEKE achieved a 65% ROIC, showing its competitive advantage. Its low ROE is masked by the significant cash reserves it holds, which make up 68% of its total assets. The fragmented realtor industry and the to-be-developed home service market offer BEKE substantial growth opportunities.

Good management is just as crucial for a great company. No matter how much analysis is done, no one can predict all the problems a company might face in the future. So it’s up to leadership to steer the company through challenges and uncover new growth opportunities. In this regard, as I will discuss in detail, BEKE has room for improvement.

With prevailing pessimism surrounding the Chinese economy, particularly its real estate sector, BEKE have been trading around 9~15x EV/UFCF — looks like a bargain price for a company with franchise value and ample room of growth. I use RMB throughout this writing unless stated otherwise.

History: from a local shop to the largest online platform

BEKE’s founder Zuo Hui once said “BEKE is 18 years of Lianjia + 2 years of Beike”. To fully grasp BEKE’s business model, it is essential to comprehend China’s local market history and how BEKE evolved from a local realtor shop two decades ago to China’s largest housing transaction platform, with a GTV of 3.2 trillion. For readers who are familiar with China’s local market, feel free to skip this section.

In 2001, when Lianjia launched in Beijing, China’s house market was chaotic due to a lack of regulation. Websites like Kjj were filled with fake listings to lure customers. Realtors frequently raise asking prices covertly and sabotage each other’s deals. The list goes on. These issues, along with failed attempts to enforce transparent transactions, led to Lianjia’s agents leaving in 2004, prompting the company to reform at a fundamental level.

To combat fake listings and create a reliable industry database, Lianjia started developing its “housing dictionary” in 2008. They established a dedicated team to verify landlord identities and property paperworks and conduct detailed physical property inspections. Additionally, they systematically surveyed buildings to determine the number of units in each city. (a bit of context: in China, most people reside in a condo in a 30-floor building. Just in Beijing there are 7 million such condos. Public trading of properties was only allowed after 1998, thus properties with missing registration or title certifications are not uncommon at that time.) Essentially, Lianjia undertook the task of building the Chinese MLS. After 10 years of efforts and investment of 5 billion RMB, the dictionary now encompasses 2 billion authentic properties across 326 cities. Each entry includes meticulous details with 500 data fields. Allegedly it is more accurate and complete than the database of the National Bureau of Statistics.

Meanwhile, as Lianjian expanded, it began testing the Agent Collaboration Network (AGN) concept to foster collaboration between agents. AGN standardizes agent collaboration by breaking down complex home sales/rental transactions into standard steps. Agents, even from different branches, can collaborate, each handling one or more steps. Lianjian’s algorithm automatically distributes commissions based on everyone’s contribution. With system and data set up, Lianjia also needed people who can share its values. To achieve this, they prioritize hiring recent college graduates and train them in-house. Today, 50% of BEKE’s agents possess college degrees or higher.

In 2018, Lianjia became the largest brokerage brand in China, boasting 6000 offices and 120,000 agents. Staying true to its mission of reinventing the industry, Zuo poached Yongdong Peng, the current CEO of BEKE, to lead the effort of digitalizing what Lianjian has built. The result was the Beike platform, which made Lianjia’s “housing dictionary” and AGN available to all brokerage brands willing to join. Additionally, customers can browse listings and agent profiles for free on Beike’s app.

This is the Ke Holding we know today. As of 2023, BEKE generates 77B in revenue, sourced from commissions under its Lianjia brand and platform fees from 268 brokerage brands on Beike. 40% of its revenue comes from existing home services, including home sales and rentals. 47% stems from new home transactions, where BEKE acts as a sales channel for real estate developers. The other two income streams, home renovation and emerging services, are relatively small but growing fast.

BEKE’s Moat: an online-offline integrated fly wheel

In 2023, BEKE generated 4.7B NOPAT (excluding SBC) by using 6B average invested capital. ROIC of 65% indicates a strong moat around BEKE’s business. In this section, I’ll delve into BEKE’s competitive strengths and provide two pieces of evidence at the end.

(1) BEKE distinguishes itself as an integrated online and offline platform. Unlike typical e-commerce platforms, such as Alibaba or Amazon, that facilitate direct transactions between sellers and consumers, BEKE provides a platform for realtors in serving both sellers and buyers in the market. This distinction arises from the unique nature of the real estate industry, where residential properties are not click-and-buy products. Each property is unique in terms of location, condition, and other factors, requiring local insights and professional knowledge of realtors. Additionally, the significant financial investment and complex paperworks involved further underscores the need for personalized service. Therefore, digital solutions cannot fully replace the offline aspects of the process, making the agent network indispensable in housing transactions and services. This deters those e-commerce giants from breaking into the housing industry.

(2) The ACN and authentic housing data are two pillars of BEKE’s competitive advantage. The ACN fosters collaboration between brokerage brands and agents, standardizing and streamlining the transaction process to improve service efficiency and quality. Its “housing dictionary” database and the associated SaaS service serve as the backbone for the ACN operation. Their seamless integration improves service quality and builds brand name, empowering the network effect (Fig.1) and forms an impenetrable barrier for its peer competitors.

(3) The 3100 TB database compiled by BEKE, developed over billions of RMB and decades of effort, is invaluable by itself. It provides exclusive insights into industrial trends and consumer behaviours, enabling better matching between buyers and sellers, strategy optimization, and the growth of new business.

(4) In 2020, Alibaba and Yiju, a real estate service company, formed a joint venture to sell new properties online. The idea was to leverage Alibaba’s vast online traffic and huge discount offerings. As I pointed out, traditional e-commerce strategies won’t work because residential property is not a click-and-buy product. Just one year later, Alibaba exited the venture by selling 85% of its share to Yiju. Yiju defaulted on its debts in 2022.

(5) Since 2021, BEKE has operated with negative working capital. As current liabilities rose with revenue recovery, current assets slightly decreased. Fig.2 shows the trends of BEKE’s revenue, receivables and contract liabilities for different segments in the past 5 years. When real estate developers faced liquidity issues, BEKE implemented various risk-control measures, including upfront commission payments from developers before sales. As a result, receivables from new home transaction services, comprising over 95% of BEKE’s total receivables, decreased by 74% since 2020. Meanwhile, its contract liabilities tripled. This prudent management contrasts with Yiju’s bankruptcy due to developers’ insolvency. Moreover, BEKE’s ability to implement the “commission in advance” model underscores its strong bargaining power, a hallmark of businesses with competitive advantages. A similar trend is observed in BEKE’s home renovation and emerging services, where customers are required to pay (partially) in advance. As depicted in Fig. 2, contract liabilities for both segments increased significantly faster than receivables.

Growth: a long hill to roll the snow ball

A great company generates high return on invested capital and has sustainable room for growth and compounding. I’ve already discussed BEKE’s ability to generate high returns; now, let’s look into its growth prospects.

(1) Before COVID-19, BEKE experienced annual revenue growth of 30%. However, the strict lockdown measures significantly affected its operations as offline work is essential for agents. Despite this, the pandemic’s impact is waning. The rapid recovery of the existing house segment and the growth of renovation and emerging services, as indicated in Fig.2, demonstrate this. In fact, BEKE’s overall revenue has already recovered to 90% of its 2021 level.

(2) The lag in the new home segment is linked to the recent turmoil of China’s real estate market, a result of a decade-long overheated market and high leverage among major developers. However, this downturn has a lesser impact on BEKE’s existing home segment. The rental market in China’s major cities, where BEKE has a significant market share, is supported by demand of foreign workers. With the highest gross margin of 47.2%, compared to 26.6% in the new home segment, the impact is further reduced on the bottom line. Finally, the rapid growth of emerging services like renovations and property management will provide further diversification in the future.

(3) BEKE still has ample room to expand its market share by leveraging its network effects. The realtor industry remains highly fragmented, with BEKE currently holding a market share of 9.7%, while the nearest competitor only has 1.4%.

(4) In the long term, people are increasingly viewing properties as places to live rather than for speculation. This shift presents new opportunities for growth in the housing service industry. BEKE, with its extensive data on housing consumers, is well-positioned to capitalize on these opportunities. For example, in 2021, BEKE launched home renovation and furnishing services, allowing users to customize and preview designs in the app using VR technology. Revenue from this segment grew from 197 million to 10 billion in three years, making it BEKE’s third-largest business. Interestingly, the home decor and renovation industry in China right now is highly fragmented and lacks standardization, resembling the realtor business 20 years ago. BEKE is aiming to replicate its past success. Just a lot faster this time.

Management I: Succession problem

I believe Zuo's passing, due to cancer, in 2021 left BEKE with a complex issue of succession and equity control. Successfully succeeding a visionary founder, like Tim Cook did with Steve Jobs, is rare. Great companies need great leaders—individuals who not only build the company but also establish its culture and install its soul. Founders like those of Alibaba, JD, and PDD continue to influence their companies, whether openly or behind the scenes.

Zuo, started as a real estate agent, was deeply respected by BEKE’s old employees. The current CEO, Peng, is a professional manager personally chosen by Zuo to digitalize Lianjia, paving the way for today's BEKE. BEKE’s unique business model, which integrates online and offline operations, creates potential conflicts between the veteran offline agents, represented by Zuo, and the ‘new’ tech team, represented by Peng. Such tension was under control when Zuo was alive. But now with Zuo’s passing and the tough macro environment in the past few years, this could be a risk. In fact, there were employees complained online that Peng seems more concerned with the company's financials than with the welfare of agents.

This tension is also evident from BEKE's recent equity changes. In May 2021, Zuo passed away. The 2022 financial report shows that Zuo's wife converted all inherited B shares with ten votes per share into A class ordinary shares. As a result, Peng, the CEO and Chairman, holds 22% voting power, co-founder and executive director Shan holds 10%, and the Zuo family holds 17%.

Despite Zuo had publicly claimed Peng is the one leading BEKE, this equity structure tells a different story: Peng does not have absolute control over BEKE. Shan, a veteran of Beijing Lianjia and a real estate professional, obviously represent the interests of the realtor faction. Together with Zuo’s wife, they have more voting power than Peng. I’m concerned that the succession planning focus more on checks and balances rather than prioritizing the company's best interests.

That said, given BEKE’s unique business model, this equity setup might be understandable. For BEKE to function smoothly, both the offline agents and the online tech team are indispensable. Not giving Shan, representing the realtor force, a large share of control could lead to discontent. Therefore, I believe that one of the biggest challenges for BEKE's management is how to harmonize the online and offline segments into a cohesive whole, rather than letting them fall apart. Right now, it appears that everything is under control and Peng is leading without much hindrance. But maybe I’m just paranoid; I worry this could be a hidden time bomb.

Management II: Stock-based compensation

I think BEKE’s stock-based compensation over the past three years has been exorbitant, making me question if the management's interests are aligned with those of the shareholders.

In 2022, the company adopted a Share Incentive Plan in which Peng and Shan was awarded 71.8M and 53.8M shares, respectively. At the beginning, I thought this was necessary to grant the new leaders enough power of control. But a simple calculation shows this is not the case:

Converting all B shares from Zuo’s wife was sufficient and the Share Incentive Plan did not help or hurt any party’s voting power. As a result, the shares hold by directors and all executive officers in 2022 increased by over 100% compared those of 2021, while BEKE’s revenue dropped by 25% in 2022. At a time when leadership needed to set an example and be frugal, the management chose to line their own pockets. Without the SBE due to this incentive plan, operating income in 2021 would be positive and that of 2022 would increase by 25%:

Management III: Capital allocation

Recall the Point 2 of Fisher’s 15-point checklist:

Does the management have a determination to continue to develop products or processes that will still further increase total sale potentials when the growth potentials of currently attractive product lines have largely exploited?

I think BEKE’s new strategy of developing its furnishing and renovation business demonstrates such determination of the new CEO. In 2023, BEKE acquired Shengdu, a home renovation and decoration company, for 7.8B , paid half in cash and half in stocks. After just a year, this segment contributed 7.7B in revenue and 3.1B in gross profit. Crudely assuming that the operating expenses and D&A of each segment is proportional to its corresponding gross profit contribution, the EBITDA for the furnishing and renovation segment would be 1.3B. So, BEKE essentially paid 1x revenue and 6x EBITDA for Shengdu, which I don’t find expensive.

My issue with management lies in how they funded the acquisition. They issued 44M shares to pay 3.9B RMB, implying they valued BEKE at 89 RMB per share, or $38 per ADS. This might seem like a good deal, considering BEKE was trading between as low as $7 in later 2022. However, I see it as a significant loss of shareholder value. A good management team with ambition and a long-term perspective should firmly believe that the company they manage is worth an infinite value. This decision is particularly unfathomable considering BEKE has at least 55B in cash and short-term investments.

BEKE has spent $1.1 billion on share buybacks since 2022 and will likely spend another $1 billion by the end of 2024. While it's always reassuring to see large buybacks from management, and I'm sure these actions will help the stock price and the ROE, this can only be deemed adequate, not exceptional. Other major internet companies are also aggressively buying back their stocks, and BEKE's total outstanding shares are still 30% more than those in 2021 due to dilution from shares issued to 'incentivize' management and fund its acquisitions.

Valuation

Many investors tend to shy away from Chinese companies for many reasons that I won’t repeat here, perhaps with the real estate sector being particularly unpopular right now. However, BEKE’s business differs significantly from that of highly leveraged real estate developers. The realtor business is light on assets and lucrative. BEKE holds a robust balance sheet with 77 billion in cash and investments and only 41M short-term debt. (Although I include lease liability in my calculation of EV for consistency, I do not think that leasing liabilities are real debts).

At time of this writing, BEKE’s stock price is fluctuating between 12-15 USD/share, with an EV/UFCF ratio ranging from 9-15x (the large cash it holds acts like leverage making the ratio extremely sensitive to its price change). The price is not expensive if we consider BEKE’s franchise value and growth potential.

The author of this write-up owns shares in the company mentioned and may purchase or sell shares without notice. This write-up represents only the author’s personal opinions and is not a recommendation to buy or sell a security. No information presented in the write-up is designed to be timely and accurate and should be used only for informational purposes. Readers of the write-up should perform their own due diligence before making investment decisions.