Nu is now trading at around a $75 billion market cap. This past quarter it just generated $1.06 billion in net income, which comes out to a $4.24 billion annualized run-rate. To get a sense of what the market is actually pricing in, let’s run a simple DCF off that net income. Keep in mind that because Nu is still growing at a fast clip, net income will overstate true cash flows. Also, because it is a financial institution, it needs to retain extra equity to fund asset expansion and stay ahead of capital requirements. The market is currently implying Nu will grow EPS 18.4 percent annually to achieve a 13 percent annualized return. I assumed a cost of equity of 13 percent, a 2.5 percent perpetual USD growth rate, and a 9.76x exit multiple. You tell me if a business that just reaccelerated revenue growth to 50 percent is capable of achieving this.

Let's take a closer look. If we ignore margin expansion and assume net income margins remain around 20 percent, it largely comes down to increasing the average revenue per active customer and adding more active customers to the platform. If the ARPAC stalls out at $22.00 per month then Nu would need to grow from 116.1 million active users today to 186.5 million to generate $4.10 billion per month. This means that Nu would need to grow from 139 million to 223.4 million total customers if we assume that the activity rate remains at 83.5 percent. Side note: the activity rate in Brazil soared to a record high of 86 percent this past quarter.

"Engagement continued to deepen alongside that growth: our activity rate expanded sequentially to 83.5 percent, while Brazil surpassed 86 percent for the first time." - David Vélez, CEO

The alternative is that the customer base plateaus at around 165 million total users and the ARPAC would need to grow from around $17.10 to $29.76 by 2031. Considering that ARPAC has accelerated to nearly 40 percent year-over-year this past quarter and user growth is still way above the necessary 4 percent growth for this scenario it looks very achievable.

Subscribe now

A Simple Rough Estimate

We already have a good indicator of how much higher ARPAC could get for the current cohorts. Nu shared in the earnings presentation that the 60-month cohorts currently have an ARPAC of $14.80. But more impressively, the 90-month cohorts are showing an acceleration in ARPAC, earning about 31 percent more revenue per user than 84-month cohorts at $30.70. This means that cohorts older than 90 months likely have an ARPAC that skews even higher.

To make it even clearer, we can do a simple math exercise to prove how easily Nu would be able to achieve the necessary ARPAC to reach $4.10 billion in monthly revenue. Nu historically reports a 0.13 percent monthly attrition rate, which comes out to about 1.5 percent attrition per year. There are currently 116.1 million active users, so let’s assume only 107.6 million of them will be left in 2031. We can also assume that active users will reach an upper limit of $35 ARPAC for cohorts older than 90 months, which is only a 14 percent lift from the 90-month cohorts. This leads to a blended ARPAC of $30.24 for the remaining active users in 2031, which already delivers $3.25 billion in monthly revenue. David Vélez, the CEO, directly mentioned on the call that the blended ARPAC for incumbents was in the mid-40s and he does not expect Nu to quite reach that just because it charges fewer fees.

"A lot of the incumbent banks, if you look at the ARPAC, are at 40 to 45. We don't necessarily think we'll get to 40 or 45 because there are a lot of fees that we don't charge. There might be a lot of products that we don't offer." - David Vélez, CEO

Capping the seasoned active users at $35 ARPAC in 2031 when it’s likely that cohorts maturing beyond 90 months could easily break this threshold will help keep this estimate grounded. After accounting for the seasoned users, Nu would still need an additional $850 million per month from newly acquired active users to reach the necessary revenue target. Based on the current reported ARPAC per cohort we can extrapolate that the ARPAC of newly acquired active customers will be at least $10. Even if the ARPAC for newly added cohorts and vintage cohorts puts up underwhelming numbers, Nu would need to add 85 million more active customers or 101.8 million registered customers based on an 83.5 percent activity rate. This would require total customers to grow by 73.2 percent over the next five years or 11.6 percent annualized. I think these are reasonable expectations, especially if Nu hits an inflection point in Mexico, which it is on pace to do.

This becomes even more achievable when we look at the latest initiatives that Nu is taking to further monetize its customer base. Nu just launched the Croma subscription tier for the Super Core segment, targeting customers who earn between three and ten minimum wages. This product will act as a major lever of growth because it functions as a direct bridge between the mass market and the high-income Ultravioleta demographic. This is a very exciting step for the business, but we will talk a lot more about this later. In this exercise we are assuming that Nu will not compound the quality of the business over the next five years and the DCF still looks very reasonable with highly conservative estimates. The underlying cohort math provides a solid margin of safety, but before we take a look at the alpha let’s do some simple accounting.

Behind the Accounting

Just a reminder that Nu recently switched to “Managerial P&L” and this slightly distorts its margins. Under this methodology, income generated from tax-advantaged Brazilian treasury notes inflates Nu’s margins. The actual net income remains untouched but this hits the top-line and tax-line. Similarly, this artificially suppresses the efficiency ratio by 550 basis points to 19.5 percent. Cash generation is strong but headline margins are inflated.

Portfolio Growth and Delinquency Trends

Nu’s credit portfolio grew by 37 percent year-over-year to nearly $40 billion. Credit cards account for 66 percent of the portfolio, whereas personal loans account for the remaining 34 percent. The personal loan mix is about three-quarters unsecured and one-quarter secured.

Non-performing loans were pretty stable for both early-stage and late-stage categories, which is impressive because unsecured lending grew 45 percent year-over-year.

As Nu continues to expand into higher-yielding, downmarket credit segments we can expect NPLs to slightly tick up. More importantly, Nu is appropriately pricing the risk of these new cohorts. It demonstrated just that as Risk-Adjusted NIM jumped by 290 basis points quarter-over-quarter to 12.4 percent and 250 basis points year-over-year. As Nu continues to expand its credit offerings downmarket it should expect to earn a higher yield in the process.

Durable Underwriting Performance in Core Segments

Nu reported a 7.1 percent late-stage NPL for Super Core segment users. This tier of users is reported to have an NPL of 7.6 percent for legacy S1 and S2 Brazilian banks.

Part of the reason for Nu’s success in underwriting has to do with its proprietary nuFormer AI architecture that helps assess credit capacity. NuFormer is supported by the primary banking relationships of over 60 percent of its active users. Its underwriting data stack is clearly more than just an AI buzzword since it is enabling it to underwrite a mid-market segment better than top traditional banks.

Despite demonstrating elite profitability and underwriting, Nu still maintains $6.64 billion in total allowances, which covers 244 percent of late-stage NPLs.

Nu Mexico

Before writing this article, I just finished writing about how Mexico is going to be the primary growth engine for Mercado Libre over the next decade.

For similar reasons, I expect Mexico to be the primary growth engine for Nu over the next decade. In my Mercado Libre article I even shared this same slide from Nu’s earnings because it lays out exactly why Mexico is such a big opportunity for both Nu and MELI. The Mexican population remains largely underbanked compared to Brazil, cash preference is still very high, credit card penetration is nearly one-fourth the size of Brazil’s, and instant payment adoption is nearly nonexistent. More importantly, the ARPAC in Mexico is more than double compared to what Brazil’s was at the same penetration levels of the adult population. Vélez even reiterated that Mexico could be as big of an opportunity as Brazil:

"As we've said, we think Mexico, a base case for Mexico, is a business that could be 60%, 70% of Brazil. If utilization in Mexico happens and a real-time payment system works, it could be as big as Brazil. It's a lower population but has a 30% higher income per capita. And the ARPACs that we're seeing in Mexico are equal or above Brazil." - David Vélez, CEO

Nu just recently secured the Multiple Banking license that MELI is still waiting on from the CNBV. This will allow Nu to upgrade its legacy deposit insurance so that it can secure more money per account. This helps enable Nu to capture upmarket payrolls and new customer segments.

Nu Mexico’s customer growth dropped to roughly 32 percent year-over-year this past quarter after putting up over 50 percent growth the year prior.

One of the main reasons for the sequential deceleration was because Banxico started cutting rates and Nu had to lower its promotional yields. This allowed Nu to keep the cost of funding at an optimized 88 percent of the interbank equilibrium interest rate.

Now for the actual fun part of estimating how big Nu Mexico can be over the next five years. Let’s first estimate how big Nu Mexico is today. The 16 million total customers that Nu has convert to about 12.8 million active customers assuming an 80 percent activity rate. Because Mexico’s ARPAC is $12.30, the estimated monthly revenue that Nu Mexico contributes today is about $157.4 million. That comes out to a ~$472.3 million quarterly run rate, which means that Nu Mexico contributes about 8.5 percent of revenues to the consolidated business today.

Now let’s use Brazil’s historical trajectory to help estimate Mexico’s revenue in five years from now. For comps we will use Q2 2020 where Brazil had nearly an identical 16.8 percent adult population penetration compared to Nu’s 16.5 percent penetration in Mexico today. Nu is clearly monetizing its customers earlier in Mexico, but it is also benefiting from the country having a higher GDP per capita.

Since 16 million users account for 16.5 percent of the adult population, Mexico has an implied adult population of about 97 million. Let’s also assume that Nu can reach 40 percent penetration by 2031, which is well below Brazil’s current 60 percent penetration. Assuming Nu maintains an 80 percent activity rate then it will have 31 million active Mexican users by 2031. This seems pretty grounded considering that Vélez just stated in the Q2 call that:

"Mexico is Brazil's playbook, running faster, and with the benefit of the scale we have today. That's how we broke even in 6 years in Mexico, compared with 8 years in Brazil." - David Vélez, CEO

If we want to be really conservative we can just assume ARPAC will grow in the same absolute dollar amount that it did for Brazil during the equivalent five-year time span, which will lead to $21.80. This implies a much slower growth rate because Mexico is already grossing over twice as much per user. Over the past six years, Brazil ARPAC grew over 20 percent. If we wanted to be a little less conservative we could assume Mexico ARPAC will continue compounding closer to the actual rate that Brazil ARPAC compounded for over the five-year equivalent period, which leads to clearing $30. I will just give Mexico a blended ARPAC of $26.00, which is the average of the two cases, for its 31 million active users in 2031. In this scenario, Nu Mexico would be generating over $800 million per month. That represents a more than 5x increase in Mexico's revenue over the next 5 years. That comes out to nearly a 40 percent CAGR for revenue alone. It’s worth noting that since Nu just reached breakeven profitability in Mexico last quarter the EPS CAGR will likely be significantly larger.

I provided a sensitivity table below to help demonstrate the different possible outcomes. The bearish scenario would lead to a measly 3x over the next five years while the bullish scenario would add over a billion in incremental monthly revenue representing over a 7x.

Nu Colombia

Let’s not forget about Colombia, although most analysts still do. The rollout of the national Bre-B instant payment system may become a tailwind for the business just as Pix was for Brazil. To be clear, Colombia is still early, but it is scaling as fast as Mexico did at the same stage of maturity. Nu Colombia just surpassed 5 million customers, which is about 13 percent of the country’s adult population.

The unique hurdle for Colombia has been the usury rate. Because Colombia’s national interest rate is capped at around 25 percent, Nu was unable to accept the risk of the lower-tier segments. This initially slowed down Nu’s adoption because approval rates for just the standard cards were 15 percent.

Management’s clever workaround for the rate caps was launching NuControl, which is a small credit card that carries a small monthly fee. This fee accounts for the missing interest rate. This allowed Nu to appropriately price risk and increase its approval rates from 15 percent to 40 percent. Nu Colombia also became a top-five deposit base in the country after just one year of launching its deposit product. This base helps fund the growing credit engine.

Now for the hard part: estimating Nu Colombia’s revenue by 2031. At 13 percent penetration, Colombia’s maturity is closest to what Brazil’s was in the first quarter of 2020 when it surpassed 20 million customers. Colombia’s GDP per capita is a little smaller than Brazil’s but Nu is also likely monetizing earlier just like it is doing in Mexico. Because of that I will conservatively model Colombia with the same ARPAC as Brazil’s $5.00 back in the first quarter of 2020. Colombia’s ARPAC is likely higher than this, but since we have no reported data on this, I prefer to play it safe. If Colombia’s ARPAC compounds at a 20 percent CAGR just as Brazil’s did then we can expect it to reach an ARPAC of $12.60 by 2031.

Let’s assume that Nu Colombia’s adoption curve underwhelms and it only reaches 40 percent penetration by 2031. With 15.3 million registered customers an 80 percent activity rate gives us 12.2 million active users. By 2031, Colombia could easily be contributing $154 million per month of revenue, which is how much we just estimated that Mexico is contributing today. This proves that Colombia has the potential to become a material revenue contributor over the next few years. Because its current contribution is so small right now, it’s mainly just upside potential from here.

Brazil Monetization

Now for Brazil. Over the next five years, Brazil’s monetization will have the biggest impact on Nu’s growth. There are several important factors to consider, such as how much revenue seasoned cohorts will generate. In the rough estimate section above, I conservatively assumed that ARPAC for 90+ month cohorts will taper off at around $35 but there is no sign or reason that this has to be the case.

The next important strategy is the recent launch of Croma. I am very bullish on this move. The Croma subscription tier was launched at the end of July to further monetize and lock in the Super Core demographic. This segment of the market sits in between mass market and high-income users. The subscription is currently half the price of Ultravioleta at R$39 per month but the fee is eligible to get waived by spending R$4,000 per month or keeping R$30,000 in a savings or investment account. Croma functionally acts as an ARPAC accelerator and, more importantly, reinforces a user’s reliance on Nu’s sticky ecosystem.

The benefits of Croma include 120-130 percent CDI yields and a free NuCel mobile plan, which locks users into the broader ecosystem. I talked about this in a previous post but the NuCel telecom data also helps reinforce the data flywheel for underwriting.

Croma has an addressable base that is over twice as large as the high-income base that Ultravioleta serves. Introducing this new tier functions as a bridge between the mass market and the high-income market. As Brazil’s middle class continues to grow, this tier’s total addressable market will increase. The interesting part is Nu actually does not want to collect the R$39 monthly fee from Croma because it benefits further from the behavioral lock-in required to waive the fee.

The R$39 fee comes out to about $7.35 but if the user waives the fee by spending R$4,000 per month on credit cards, then Nu collects a 1.2 percent interchange fee. The minimum Nu collects from the interchange fee if the user spends past the threshold is $9.05. Assuming the Croma subscription may influence the user to also utilize revolving credit, finance a purchase or buy something on the Nu shopping marketplace, it is very clear that ARPAC for this segment is going to increase materially.

To get a better idea of Croma’s adoption, let’s use Ultravioleta's rollout as a baseline. Nu acquired about 700,000 Ultravioleta customers by the end of 2024. This implies that Nu converted about 13 percent of the high-income TAM that consists of 5.3 million adults to its premium tier within three years.

The TAM that Croma is targeting consists of 12.4 million adults. The fee waiver threshold is a fraction of UV’s so we can expect the barrier to entry to be even lower. If Nu can achieve a 13 percent conversion rate for Croma then it would acquire about 1.6 million Croma subscriptions. Considering that Nu has already captured three-fifths of the Super Core segment, a 20 percent conversion rate by 2031 does not seem unreasonable. This implies that Croma could scale to 2.5 million subscribers in just five years very easily. Vélez also highlighted how it is in a strong position to monetize the Super Core segment that it already owns:

"I can tell you that we already have three out of five Brazilians in this bracket as customers of Nu. It is not an opportunity, necessarily, we have to go out in the market and acquire these customers and spend a lot of money on marketing. They are already inside our base. We just have not treated them as good as they deserve... There is a huge opportunity to increase the share of wallet within those three to five Brazilians that exist in that base." - David Vélez, CEO

So What?

As we determined in the beginning, for Nu to justify its current valuation and deliver a 13 percent annualized return it would need to generate $4.10 billion in monthly revenue by 2031. We conservatively estimated that Nu’s current active users would generate $3.25 billion per month by 2031. So the remaining $850 million per month is reliant on Mexico and Colombia’s success, new product launches, improved monetization and improved credit modeling.

We already modeled in that Mexico and Colombia can conservatively achieve $505 million in monthly revenue from net-new active users. That means the remaining $345 million in revenue would have to come from the estimated net-new 2.5 million Croma users, S1 Credit expansion via nuFormer model lifts, further acquisition of Ultravioleta members, SME expansion, new Brazilian youths entering the financial system and further penetration into the Brazilian market.

For a mid-teen CAGR, Nu really does not have to do too much over the next five years and it has already historically proved that it should be easily able to exceed our conservative expectations. Based on these projections, Nu appears positioned to outperform current market expectations.

Unsurprisingly, Mercado Libre’s second quarter earnings did not impress Wall Street. Wall Street was more concerned about the 550 basis point compression in operating margins than surpassing $10 billion in quarterly revenue for the first time with 43 percent FX-neutral growth. There was also a fair amount of concern over whether the credit portfolio is beginning to crack as it expanded by 75 percent year-over-year to $16.4 billion.

I largely disagree with Wall Street’s consensus. Mercado Libre is not suffering from competitive erosion. Management has demonstrably continued to execute every quarter with strategic capital allocation, and this quarter is no different. They continue to utilize highly profitable segments to subsidize behavioral change in user habits across Latin America. In this post, I will focus on connecting the dots and examining some major insights that I found throughout this past quarter.

Subscribe now

Re-investing Margins Back into the Ecosystem

Last year it was celebrated that the credit card portfolio was beginning to reach breakeven for cohorts that were 12-18 months old. This quarter, management revealed that the credit cards NIMAL (Net Interest Margin After Losses, which is essentially the profitability of the credit portfolio) compressed from breakeven back down to negative 2.5 percent.

In the first quarter of this year, MELI actually accelerated loan originations by 25% quarter-over-quarter and extended maximum terms for top-tier users from 5 months to 8 months. Martin de los Santos clarified this in the Q1 CFO Perspectives:

“For certain users, particularly those that had paid us well and have behaved very well with our credit, we’re extending the term of the loans... on average we’re going from five to eight months... However, because of the longer terms of the loans, we are required to also provision more in this particular quarter.” - Martin de los Santos, CFO

SVP of Corporate Development, Leandro Cuccioli, also clarified that a fair amount of the credit cards issued are being used solely as a payment tool:

“If you issue a $100 of cards you need to provision for as if you have $100 of loan exposure but in reality a lot of those people will use the card as a payment card and they pay at the end of the month the full balance... in the short term there is a compression in margins as you’re issuing cards.” - Leandro Cuccioli, SVP of Corporate Development

Make no mistake: it is critical for the long term that NIMAL remains positive to offer a margin of safety in the event of a severe credit crisis. But the entire credit portfolio is sitting above 20 percent NIMAL, which means regardless of what the default rates are, they are continuing to appropriately price risk. As for the credit card portfolio, the economics of it naturally have lower margins, so NIMAL will still trend lower over time as it makes up a greater share of the portfolio.

The reason for a year-over-year drop in NIMAL was a calculated acquisition cost. This past quarter, Mercado Libre issued 2.6 million new credit cards, which is a 62 percent year-over-year increase in credit cards issued. This strategy involves intentionally taking an upfront hit on these early cohorts because there is already a clear path to profitability, but more importantly, they know what credit cards do to user habits in the ecosystem. Over time, engagement with both platforms increases for seasoned cohorts.

It is also worth noting the short duration of these loans. Currently, 98 percent of the credit portfolio consists of loans with an average duration of under four months. This is very low compared to its competitors in Latin America and gives Mercado Libre a structural advantage in pivoting when the credit environment begins to sour.

The underwriting continues to look disciplined. Keep in mind that they are doing this during a period of credit friction in Argentina and Brazil in recent months.

The stable NPLs and overall profitable NIMAL demonstrate that Mercado Libre has a lot of leeway before credit could become a major headwind.

Despite credit itself remaining very profitable, management has made it clear that their main strategy for credit is building the “ecosystemic user.” This is a type of user who actively engages with both the marketplace and the fintech platform. Management called out that credit provisioning for bad debt resulted in a 240 basis point drag on operating margins.

To better understand why management views this compression as a re-investment back into the business, here is a direct quote from Martin de los Santos from the Q1 CFO Perspectives:

“The investments that we’re making are bringing more people to our platform, are converting more people, and actually expanding the market… and those users are more sticky, they tend to come and buy in more categories and make use of our ecosystem in a much broader way.” - Martin de los Santos, CFO

Ecosystem users still represent a small portion of Mercado Libre’s customers and an even smaller portion of Latin America. This quarter, ecosystem users grew by 37% year-over-year, which outpaced the steady ~30% growth in monthly active users for Mercado Pago. Management also revealed this quarter that ecosystem users generate 70% more GMV and 90% more TPV than single platform users.

The SVP of Mercado Pago also recently shared why they still don’t utilize customer deposits to fund the credit book:

“We don’t need to because we have plenty of access to funding either through secure debt issuance or securitization... It’s part of our core value prop to yield the account. We don’t see the need and we don’t see the upside [in using deposits for funding].” - Andre Chavez, SVP for Mercado Pago Brazil

The Ad Ecosystem

Mercado Libre’s high-margin ad segment is scaling very nicely. They continue to make key strategic partnerships, such as the integration with HBO Max. This partnership alone added 50% in volume to their video ad inventory, which is a huge step-change for adding a single partner.

The high-converting display ad landscape was historically dominated by brands with sufficient capital and ad agencies. This was not something that small and medium-sized businesses had resources for. Last year, Mercado Libre partnered with AWS and Stability AI to build GenAds to help integrate the necessary generative AI tooling into the seller platform for these smaller brands. This helped empower SMBs on Mercado Libre’s platform to participate in Mercado Libre’s integrated ad network. At the time of the launch, the VP of Mercado Libre’s marketplace said that Amazon Bedrock was pivotal in improving data processing and creative capabilities. The initial rollout resulted in over 90,000 product banners generated across Latin American countries, which drove a 45 percent increase in display impressions and a 25 percent higher click-through rate compared to non-GenAds campaigns.

CMO Sean Summers pointed out in a recent interview that AI allows Mercado Libre to move beyond local advertisers to the long tail. By lowering the barrier to entry for ad spend, MELI unlocked a new total addressable market of high-margin revenue. This has been complemented by the marketplace’s recent deployment of AI-powered search across its 5 largest regional sites, which generated uplifts in both conversion and ad click-through rates that are offsetting the cost of 3P LLM queries.

Social Commerce

This past May, Mercado Libre launched the MeliConnect affiliate program. This integration allows them to seamlessly inject their catalog into major social platforms and bypass paying for impressions on auctions. They just pay for guaranteed performance conversions. Ariel, the CEO, revealed during the call:

“The affiliates channel is actually scaling very, very nicely and becoming more efficient simultaneously. So, we think it’s actually working. The affiliate GMV share grew across every market in Q2, including Mexico where we dialed down some of the couponing that we’re deploying in the past.” - Ariel Szarfsztejn, CEO

Ariel also made it clear that the affiliate program is not just driving up one-time engagement metrics but actually bringing in ‘high-quality’ users:

“The channel is not only driving one-time purchases, it’s also acquiring users who come back later to our platform, and that’s really exciting as well.” - Ariel Szarfsztejn, CEO

This creates another tailwind for the ecosystem where Ariel further explained how the economics are improving:

“Economics are improving. Seller co-founding for affiliate is growing. So, extremely excited and positive on the outlook for the different markets.” - Ariel Szarfsztejn, CEO

This strategy has also been strengthened by a deepening alliance with TikTok. MELI’s strategy is to turn the social media giant into a discovery engine for Mercado Libre products. A TikTok rep at the Cannes Lions 2026 festival shared their view on the value of this integration:

“In addition to Discovery Commerce, we also talk about entertainment commerce or shoppertainment. In shoppertainment, we see a very clear hierarchy: first entertainment, then conversion. I believe that with the Mercado Libre alliance, given the robustness of Mercado Libre’s product, that conversion comes very fluidly.” - Dani Craim, Business Director at TikTok

The Physical Moat

At a recent logistics summit, Mercado Libre shared details on some of the R&D they have been doing with automotive manufacturers to customize vehicles specifically for urban Latin American last-mile delivery. They partnered with Ford Pro Latin America and their representative shared some of the latest features they developed such as equipping the vehicles with integrated camera systems and ensuring the rear doors only open with specialized proximity wristbands.

Their newly launched $140 million XEM3 cross-docking facility in Mexico adds the capacity to process 1 million more packages per day and deliver more same and next-day deliveries across 28 cities. Given that Mexico is where management sees their biggest opportunity, this is critical for scaling and capturing more of the market.

Something that got less attention during earnings was the quarter-over-quarter acceleration in same and next-day shipments while operating margins remained relatively stable. The new scale and density that Mercado Libre achieved over the last year is beginning to make previously unprofitable segments profitable:

“The unit economics of this decision are also evolving as we expected. Free and slow shipments are now variable contribution-positive across half of the ASP ranges between R$19 and R$79, as technology, scale, and the use of previously unused capacity continue to bring costs down.” - Mercado Libre Q2 2026 Letter

Cross-Border Trade

Mercado Libre is actively building a cross-border pipeline that connects the most efficient manufacturing hubs in the world directly to Latin American buyers. The VP of Cross-Border Trade shared on the management podcast:

“It goes beyond just connecting supply and demand. We want to provide sellers in China and in the US with a full suite of technological and logistical solutions they need to unlock these markets… Increasingly we think about China and the US first when we build our tools to make sure they are scalable.” - Karen Bruck, VP of CBT

This represents a big strategic pivot beyond local merchants that will allow Mercado Libre to strengthen its positioning around product assortment and pricing. Management is focused on easing friction with international sellers and building relationships with Chinese factories. This will allow the business to lower its cross-border delivery speeds by injecting supply directly into its proprietary logistics network. Volume at MELI’s dedicated China fulfillment center surged 170 percent quarter-over-quarter. This comes back to strengthening the ecosystem offerings for the user and increasing buyer frequency:

“CBT is another lever for expanding supply… in this case by giving buyers access to a far broader range of products and price points than domestic sellers alone could offer.” - Karen Bruck, VP of CBT

What about AI?

Mercado Libre is decoupling its revenue growth from headcount. They spent an incremental $80 million on AI computing this quarter and this is already resulting in saved costs. Product Development expenses fell from 8.4 percent to 7.2 percent while productivity increased. This operating leverage was a result of unprecedented developer productivity where overall code submissions increased 110 percent year-over-year, and deployments increased around 75 percent year-over-year.

COO Daniel Rabinovich shared some compelling data on this front last year. Back then, he mentioned how AI was already writing 20 percent of their new code, and management has alluded to this ratio being a lot higher this year. Rabinovich noted that based on their growth trajectory, they should require half as many customer service representatives while continuing to achieve record-high Net Promoter Scores. They are seeing AI is contributing to an elevated deflection rate:

“Using LLM models, for example, to resolve customer service... today we have almost a 90% deflection rate and record NPS. We grow double digits every year, and we are lowering the absolute number of reps because our robots understand problems better and resolve them faster.” - Daniel Rabinovich, COO

Andre Chavez recently confirmed that AI is rewiring their core financial engine. He said:

“We are investing in better models for credit underwriting, better models for fraud prevention… We are investing in a lot of automation of back offices like customer experience, customer service, fraud etc... So we are all in on AI in general and I’d say it’s going to be transformative on all three fronts.” - Andre Chavez, SVP for Mercado Pago Brazil

The AI Seller Assistant resolves over 50 percent of Help Portal requests entirely autonomously without any human intervention.

The Stablecoin Pivot

Earlier this year, Mercado Libre sunsetted Mercado Coin to focus on its stablecoin, Meli Dolar. This is more than a consumer crypto play because Meli Dolar will serve a critical role in B2B financial infrastructure. Meli Dolar allows merchants to hedge working capital against local fiat volatility, such as the Argentine Peso, and facilitates frictionless cross-border settlements. Mercado Libre is deploying blockchain rails to lower operational costs and retain merchant liquidity entirely within its ecosystem.

Beyond E-Commerce

One concept the broader market struggles to grasp is Mercado Libre’s endgame. Management views it as the entirety of Latin American commerce, banking, and media. In a recent interview, CMO Sean Summers detailed the strategy.

When evaluating their runway, experts focus on the reference regional e-commerce penetration rates. This metric is actually flawed because Mercado Libre is also targeting offline incumbents. When asked about their trajectory, Summers said:

“27 years on, I still think that the next 10 years are going to be fantastic. I think that this is a company that could be the biggest retailer in every market where we operate. Retailer, not e-tailer.” - Sean Summers, CMO

Summers projected that Mercado Pago will be:

“a top five bank in every market where we operate over the next five years, and we have a fair chance of being a top one over the following three to five years.” - Sean Summers, CMO

Summers also provided a glimpse into their efficient marketing scaling strategy. Mercado Libre actively takes their proprietary performance marketing capabilities built for internal operations and packages them as high-conversion external products for retail media clients.

Summers noted that this operational dominance is the direct byproduct of Amazon entering Mexico in 2015. Rather than ceding market share, this competitive threat forced Mercado Libre’s leadership to completely pivot strategies. They had to aggressively learn how to:

“…out-innovate them, how to move faster, how to be more aggressive on investment, how to be smart on capital allocation.” - Sean Summers, CMO

What’s Next?

For the last year, Mercado Libre’s growth has consistently outpaced analysts’ expectations. This raises the question of when their growth will actually begin to slow down.

Despite Brazil accounting for a majority of revenues, it is also serving as the main growth engine for Mercado Libre. Brazil saw its FX-neutral revenue accelerate from 39 percent in Q1 to 42 percent in Q2. The law of large numbers dictates that Brazil’s hyper-growth will eventually moderate. While the ecosystem multiplier will likely sustain strong growth in Brazil for years to come, the long-term focus is shifting to Mexico.

Mexico represents the growth engine for the next decade because it offers a completely open runway. The country has been navigating short-term macroeconomic friction this year with a slower consumption environment and recent tax reform. Aside from these temporary headwinds, Mexico’s underlying fundamentals remain strong. Mercado Libre is continuing to capture market share from legacy retail and outpaced its main competitor this quarter, which I assume is Amazon.

This comes to me as no surprise as I have been closely monitoring the app usage stats this past quarter and Mercado Libre has been crushing it. This is partially due to their success in cross-border trade, which is allowing MELI to offer a vast depth of low-cost Chinese products.

The biggest differentiator for Mexico is fintech. Unlike Brazil, where Mercado Libre fights several fintechs and entrenched banks, Mexico remains largely unbanked. Over half the population relies on informal credit, and a majority of small purchases are made through cash. Fintech monthly active users in Mexico surged 45 percent year-over-year in Q2. Acquiring TPV grew 47 percent year-over-year, driven by a near 70 percent jump in their in-store channel. Mercado Pago is also demonstrating more robust app usage than Nubank in Mexico.

Management is deliberately accepting higher device costs from the spike in memory chips and spending heavily on user acquisition because the long-term paybacks are very attractive. Ariel brought up some promising insights on the credit cards too:

“…paybacks are even better than in Brazil because the economics of the industry works better than in Brazil.” - Ariel Szarfsztejn, CEO

Ariel also noted that Mexico is the market where their ecosystem could play out most favorably because of the lack of access to financial services. As Brazil matures, Mexico is well positioned to become a leader in both e-commerce and digital banking.

Andre Chavez also reminded me how their investment products are still very early and can provide another tailwind of growth as they develop them and tailor them for the Latin American public.

“You shouldn’t expect us to move into sophisticated securities trading. Instead, you should expect us to add more products that are aimed at middle-class customers... mutual funds and ETFs, products that are easy to understand, easy to invest.” - Andre Chavez, SVP for Mercado Pago Brazil

What’s the Catch?

While MELI’s tailwinds are huge, any objective analysis must address the risks. As a long-term investor, I would say the biggest risk for the business would be government intervention because ultimately what makes this business so special is its flywheel business model. Without the marketplace and fintech segments adding value to each other, the ecosystem thesis pretty much dies. At the end of the day, Mercado Libre is just another story of Nick Sleep’s ‘Scaled Economies Shared’. The marketplace-fintech flywheel is their competitive advantage and is what allows them to provide more value to their customers at a lower internal cost.

Regarding credit, I don’t believe it poses an existential threat to the business, but I do think there is potential for it to become a multi-quarter headwind. A credit shock would likely tank profitability, but it would also hurt mid-term growth because the credit business has been a big source of outsized profits that continue to get re-invested back into user acquisition and ecosystem growth. My bigger worry is that any slowdown in credit would largely hurt MELI’s ability to compete for growth. There was nothing unusual about credit this quarter. NIMAL continues to trend down year-over-year because credit cards are less profitable than the remaining mix. NPLs remained relatively stable, and the average duration for loans is still under four months.

Management does not get into details about how they mitigate credit risk on the call but Andre Chavez shared further insights on the latest management podcast. They monitor collections on a daily, granular basis, which allows them to throttle credit card issuance or personal loans instantly if a specific segment deteriorates. Since Mercado Pago is the preferred financial institution in the biggest regions that it operates in it actually helps shield them in times of credit friction:

“The level of NPS actually moves us up in the order of collections. So if a crisis hits, everything else equal, we are able to collect more than we did in the past just because we are the preferred financial institution of a lot of our customers.” - Andre Chavez, SVP for Mercado Pago Brazil

The competitive landscape varies a lot for each market. In Mexico, Amazon and Nubank are MELI’s biggest competition for their respective markets. Despite this, Mercado Libre is uniquely positioned to lead both markets because of its flywheel. In Brazil, Shopee is continuing its upmarket push and racing MELI to build out its fulfillment network. Mercado Libre successfully widened its gap this past year by lowering its free-shipping threshold but Shopee just recently made another push in Q2 by launching VIP in Brazil.

Sea Limited CEO, Forrest Li, reiterated during their Q2 earnings call that Brazil was Shopee’s fastest-growing market. They even managed to cut down on buyer wait times by 15 percent year-over-year while doubling their fulfillment order penetration. Shopee made further progress moving upmarket by onboarding 500 new official brands this past quarter. Shopee VIP also crossed a million users in the region after only launching in April, and they have their sight set on building a standalone ShopeePay app. For the first time, MELI is staring down a competitor that is attempting to build the exact same flywheel.

The CFO of Sea Limited even took a direct jab at Mercado Libre during the earnings call:

“We were launching an app which is similar to ShopeePay app in Brazil, but with the SCFI license, which means we will be similar to what Mercado Pago or other players in the market can offer in Brazil... If you compare what we offer and what the other player offer, there are many low-hanging fruits that we believe that we can capture just by doing the right product structures...” - Tony Hou, CFO

This goes back to why Mercado Libre dropped the free-shipping threshold a year ago to compete in the low-end market that at the time was economically infeasible. The goal was to rewire user habits and it worked. The VP of Marketplace Brazil brought up how the previous threshold acted as a friction-point:

“We had abandoned carts below 79 reais we put it at 19 reais. It was investments from MELI that leveraged a lot our traffic conversion... full funded by MELI.” - Roberta Donato, VP Marketplace Brazil

Martin de los Santos shared during the earnings call how the conversion rates in Brazil jumped by 110 basis points year-over-year. That is huge uptick in volume at MELI’s scale and items per buyer in Brazil jumped 19 percent. They are at record highs for buyers in more than 3 categories.

Back in April, MELI rolled out PIX discounts for buyers and cut take rates for sellers who kept their prices competitive on certain items. This obviously hurt margins but as Ariel explained, every time they trim take rates, seller success accelerates and this creates a virtuous feedback loop. This strategy is already showing strong results:

“The result was a meaningful improvement in our price competitiveness index, and a sharp acceleration in active sellers, which grew 29% YoY in Q2’26. A larger, more competitive seller base strengthens supply, and stronger supply is one of the clearest ways to support demand. This is a powerful flywheel that should continue to support growth and engagement in Brazil – where we continue to grow ahead of the market – and beyond.” - Mercado Libre Q2 2026 Letter

Leandro Cuccioli summed up leadership’s mindset in a recent interview:

“You need to keep the culture, the humility, the paranoia that one day another competitor may come and be better than you and you don’t need to rest on your laurels... retail is a very intense game and you need to be top of that game, but I think we have the right team and the opportunities out there.” - Leandro Cuccioli, SVP of Corporate Development

The Price of Dominance

To end the Q2 call, Ariel asserted that investment decisions are made in a disciplined manner with the goal of strengthening the flywheel of the business:

“These are not two businesses running in parallel. This is one flywheel, each side making the other one more valuable. As we said in the letter, we invest in it carefully, with discipline, and only when we see that the economics are working. If we get that flywheel right, the result is structurally higher engagement, loyalty, and scale and that is what will maximize our long-term profitability.” - Ariel Szarfsztejn, CEO

This is exactly what I like to hear from management. Leadership is fully focused on influencing consumer habits across the region, and unlocking long-term value in their ecosystem.

The Disconnect Between Market Sentiment and the Business

The market entered 2026 aggressively punishing enterprise software, driven by a severe macroeconomic rotation. Chief Information Officers actively shifted capital toward hardware, GPUs, and data center infrastructure at the expense of application-layer software budgets. This dynamic fostered the SaaS apocalypse narrative, which peaked this month following IBM’s weak Q2 earnings warning. IBM cited clients sharply reprioritizing capital expenditure toward hardware, sparking a broad sector selloff and marking IBM’s worst single-day equity drop in the company’s 115-year recorded history.

Against this volatile backdrop, ServiceNow experienced significant multiple compression. The stock declined over 30 percent year-to-date heading into the earnings as investors assumed generative AI and cheap, custom-built solutions would cannibalize traditional IT budgets and hurt seat-based licensing.

ServiceNow’s Q2 print challenged this bearish sentiment, and improved the company’s overall outlook. ServiceNow crossed $15 billion in scale and raised its forward guidance, posting 23 percent constant currency subscription revenue growth. Beyond the top-line revenue beat, ServiceNow is evolving its fundamental business model. It is actively engineering a transition into a consumption-based Agentic OS, building the AI control tower that governs human and agent interactions across the enterprise.

Subscribe now

The SaaS Apocalypse vs Token Efficiency

The core bear argument is that language models will replace structural workflow software. Bears argue code generation makes building bespoke applications cheaper than buying enterprise software. ServiceNow is actively exploiting the flaws in this logic. Enterprise executives are realizing deploying raw language models creates security and cost liabilities. Unmanaged AI creates significant operational risk. To run an enterprise securely, models require deep context, workflow orchestration, and strict governance. Foundational models lack these capabilities natively.

ServiceNow is building a token efficiency moat, positioning itself as the optimization layer between the enterprise and the model providers. The platform dynamically switches models underneath the hood to save enterprises money on compute costs. ServiceNow routes complex problems to expensive frontier models and simpler tasks to cheaper, open-weight models. Amit Zavery, President and Chief Product Officer, explained recently that everything does not have to be on the latest and greatest frontier models. You need small models and cheaper ones too. He noted enterprises must look at tiering while building proprietary intellectual property on top of the models.

Zavery expanded on this commoditization of foundational models, stating:

“A lot of these models, in some cases, are interchangeable. Whenever the pricing changes, we look at which is the best pricing... Just like I think nobody cares what chip you use underneath your cloud most of the time. Same thing will happen to the models.”

This routing architecture protects enterprise margins. Bill McDermott, Chairman and CEO, quantified the risk of abandoning enterprise platforms. He noted the cost to replace an enterprise platform in this supposed SaaS apocalypse would be ten times greater to try to replicate it with a language model. He added business leaders understand people make mistakes, but they will never forgive software for making a mistake. Despite market fears of AI killing software headcount, ServiceNow grew active users on its platform. The seat-based software model is evolving to include both human and machine identities.

The Margin Protection Mechanics

The margin protection ServiceNow is demonstrating is driven by aggressive hiring avoidance, and not mass layoffs. ServiceNow is utilizing its own AI products internally through a program called Now on Now. Despite facing gross margin pressures from hyperscaler partnerships they were able to demonstrate strength in operating margins.

This internal leverage is highly quantifiable. CFO Gina Mastantuono confirmed the exact mechanics of this margin defense, stating she saved $100 million in planned hiring with no layoffs. She explained she does not have to hire as many people even when growing above 20 percent a year, defining this as real productivity gains. Now Assist helped contribute to ServiceNow saving half a billion dollars internally. AI now supports 91 percent of L1 HR and routine IT case service requests. They are trading external computing costs for internal human capital efficiency. This structural shift helped lift operating margins to 29.5 percent, beating guidance by 300 basis points. Note management also mentioned that part of this jump was attributed to timing of spend.

The M&A Spree

The market has been debating ServiceNow’s recent acquisitions. Investors viewed the purchases of Armis, Veza, and Moveworks as reckless, desperate expansion designed to buy top-line revenue growth. McDermott addressed this bear thesis directly:

“The market might have said, are they doing M&A to cover up for any lack of growth in the core? How could that be true when it wasn’t even in the number?”

True autonomous AI is presenting itself as a liability on its own to an enterprise. If an AI agent has the power to act within a corporate network, it must be governed. Ungoverned agents introduce substantial risk. During a recent keynote, executives shared an example of an ungoverned AI agent at a startup that found an API key and made a single call wiping out the customer database and production system in just nine seconds.

ServiceNow recognizes AI autonomy cannot exist without strict governance. The acquisitions specifically address this vulnerability. Armis secures the physical operational technology and IoT perimeter. Veza secures the identity perimeter for both humans and non-human agents. ServiceNow cannot safely deploy its autonomous AI specialists without these rigorous guardrails built into the platform. McDermott addressed this directly when discussing cyber crime, stating the biggest growth total addressable market he sees in the next decade is cyber crime. The goal is to manage the identities of humans and agents, treating the agents exactly like humans.

Moveworks, however, is an offensive capability. It was acquired to provide an “AI-native front door” and an agentic reasoning engine for enterprise search and workflows. Moveworks is the autonomous AI capability that actually requires governance and is not a governance tool itself.

Redefining CRM

ServiceNow is also focused on moving beyond its traditional back-office dominance in IT and human resources, directly attacking legacy Customer Relationship Management vendors. ServiceNow is leveraging its control over backend systems of action to provide autonomous customer resolution. Legacy CRM represents a large expense line for corporations. The enterprise demand for an AI alternative that executes work and resolves issues is immediate.

McDermott outlined this pivot at a recent technology conference, stating the net present value of a very satisfied customer is the greatest asset of any corporation. He noted ServiceNow changed the paradigm from customer relationship management to the comprehensive resolution and full service of AI to a customer. The financial data supports this strategy. Customer Service Management is a business exceeding $2 billion in annual contract value for ServiceNow. Growth is accelerating because they are pairing customer service with quote orchestration, allowing clients to cut pricing workflows from weeks to hours without adding human headcount.

ServiceNow is proving this unit economics model internally. Zavery detailed their own operational success:

“We have done today around 76 percent deflection rate on customer service. We have a large case volume go up because our business has gone up very fast, but our customer support team is not growing anymore. We don’t need to.”

The Path to 2030

Despite the recent re-rating, ServiceNow is not objectively cheap. The public markets are beginning to increasingly price ServiceNow as a less certain business compared to the sky-high multiples it used to give it. On the other hand, management is executing a blueprint to become the mandatory governance layer sitting between foundational AI models and proprietary enterprise data. They are building the neutral AI control tower that manages the entire enterprise ecosystem. Jensen Huang, CEO of NVIDIA, validated this architecture, stating ServiceNow is destined to be the best platform and the operating system of enterprise AI agents.

ServiceNow has committed to crossing $30 billion in revenue by 2030. The true test of this target will depend on how effectively they can enforce the new hybrid pricing model. The model combines base subscriptions with consumption-based token pools for agentic workflows. If data over the next couple of years show enterprises are willing to pay a premium consumption rate for safe, agentic workflows, ServiceNow will have executed a massive evolution in enterprise software architecture.

During a recent interview, a journalist pointed out to Co-CEO Greg Peters that engagement time has been relatively flat over the last couple of years. Peters admitted he is not happy that engagement is not growing faster and stated Netflix needs to grow it more. But he also clarified a critical shift in how Netflix views attention. Not all viewing hours provide the same business value. For example, live events account for about 5 percent of the content budget but only 1 percent of view hours. Yet, live events have driven six of the top ten new member sign up days over the last five years. On the other hand, animation takes the same 5 percent of the budget but drives 8 percent of view hours. Netflix is not chasing raw hours. They are optimizing for the specific value that different types of content bring to the business.

Subscribe now

This strategy extends to how they handle legacy television. Instead of buying declining traditional media networks, Netflix is absorbing them through partnerships. In France, Netflix fully integrated the national broadcaster TF1 into its application. Members get TF1 linear channels and on demand content at no extra cost. Peters noted that these types of partnerships will become increasingly important. He did, however, push back on the idea of creating an internal Netflix news division. Netflix lacks real added value in news production. Instead, they will rely on partners like TF1 to provide that specific content to their subscribers. Reminds me of Mercado Libre management’s stance on making entertainment: “It’s not in our DNA.”

There is constant speculation about whether Netflix will acquire major studios. Executive leadership continues to shut this down. Co-CEO Ted Sarandos stated that Netflix is primarily a builder, not a buyer, and has a very high bar for large acquisitions. I thought they made this clear when they walked away from Warner Bros. when they increased their ask. Instead of buying legacy studios, CFO Spence Neumann noted Netflix is using its cash to reward its own investors. In the second quarter, they executed a 4.7 billion dollar share repurchase. This was the largest single quarter of buybacks in their history. As of July 2026, they still have 27.1 billion dollars in capacity remaining for future buybacks.

To capture new viewing occasions, Netflix is pivoting toward the creator economy and alternative formats. They are actively expanding into video podcasts. Management noted that podcasts over index during daytime hours and on mobile devices. This creates entirely incremental engagement outside of traditional prime time viewing. They are also bringing native social media creators directly onto their platform. In the shareholder letter, Netflix highlighted creators like Danny Go and Salish and Jordan Matter. These creators have spent weeks in the Global Top 10. Netflix is using this talent to drive highly retentive engagement without the massive budgets required for prestige television.

Netflix is also applying this audience expansion strategy to video games. Rather than trying to compete directly with traditional consoles like PlayStation or Xbox, Netflix is building accessible games specifically for its existing base of non gamers. Sean Crankle, the General Manager for Narrative Games at Netflix, detailed this strategy during a recent interview on The Game Business podcast. Netflix is breaking down the barrier to entry by developing short 30 to 40 minute narrative experiences that use the subscriber’s phone as a controller. For their recent horror game, Unhinged, they utilized the exact same internal casting department as their television shows to secure A-list talent like Sadie Sink and Zoe Kravitz. By matching interactive game genres directly to the viewing data of their audience, and removing complex gaming mechanics, Netflix is successfully converting passive viewers into interactive cloud gaming users.

Netflix is also rethinking how it monetizes its original intellectual property. When asked if Netflix would syndicate its programming to third party networks in the next few years, Peters doubted it. He stated there is not a lot of value in syndication for them. Instead, Netflix is moving into physical real estate. They are actively building Netflix Houses to host in person immersive experiences based on their shows. Peters compared this to theme park attractions for franchises like K pop Demon Hunters. This allows them to extract premium out of home spending from their existing fan base.

On the production side, Netflix is taking advantage of regional talent pools to efficiently produce global hits. Paco Ramos, the head of Latin American content, recently explained their strategy in the Spanish language market. Instead of running everything out of one central hub, Netflix empowers regional infrastructure. They utilize the shared and fluid talent pools of filmmakers, directors, and actors across different Spanish speaking countries. Ramos compared this structural advantage to the English language film industry. It allows them to produce high quality content, like the Colombian produced series 100 Years of Solitude, at a fraction of Hollywood costs.

Finally, Netflix is continuing to overhaul its advertising technology to extract higher yield from its ad tier. I went into a lot more detail about Netflix’s ad stack in my previous post, but Netflix is still on track to double its advertising revenue to approximately 3 billion dollars in 2026. To achieve this, they are expanding their artificial intelligence tools across the entire advertising lifecycle. This includes planning, creative production, campaign management, and optimization. They are also extending programmatic access to Pause Ads and live inventory. This automation reduces manual friction and opens Netflix up to a much broader range of smaller ad buyers. In 2027, Netflix plans on opening its ad tier to 15 new markets.

Netflix executives know they are operating in the most competitive entertainment environment that has ever existed. Peters admitted that complacency is their biggest vulnerability. But their current strategy shows Netflix relentlessly focused on execution. They are defending their capital through massive buybacks and avoiding expensive studio acquisitions. They are building physical retail experiences, integrating local broadcasters, and rolling out programmatic AI ad technology. Netflix is proving it can grow revenue and expand its profit margins even as management optimizes for the value of engagement rather than just raw viewing hours.

While some corners of Wall Street may underappreciate the long-term monetization capacity of the Netflix Ads Suite and particularly its programmatic AI layer and the scheduled 2027 expansion into 15 new international markets, this focus on high-yield monetization yields a highly visible cash-generation flywheel. With Netflix currently penetrated into less than 45% of its 800 million addressable households and capturing just 7% of a $670 billion addressable revenue market, Netflix’s long-term trajectory offers a highly asymmetric opportunity with limited structural downside and clear secular upside.

Investors basically punished MELI for missing earnings-per-share expectations. This happened largely because management deliberately decided to squeeze operating margins by 5 to 6% to invest in long-term value initiatives like first-party services, “Full China”, the credit portfolio buildout and subsidizing shipping. The general consensus blames this margin drop on heavy competition from Asian rivals like Shopee and Temu, but I disagree. Management isn’t bleeding margins just to stay afloat. Instead, they are systematically sacrificing short-term profits to dismantle competitors’ unit economics and improve their own long-term customer value. To really comprehend MercadoLibre’s true strategy, you have to look at the mechanics of the China-LatAm corridor. This margin squeeze is actually a highly calculated investment across four key battlefronts: dominating local logistics, controlling the China-LatAm route, smoothing out friction, and offering a superior ecosystem.

Subscribe now

Battlefront 1: Dominating Local Logistics & Low-Ticket Goods

The R$19 Threshold

Think about how Shopee beat Lazada in Southeast Asia, or how PDD disrupted Alibaba in China. In both cases, the big incumbents made the same mistake. They chased short-term profits, went upscale, and treated shipping like a luxury service. That left the door wide open for challengers to grab the everyday mass market by subsidizing shipping and focusing on lower-income shoppers. Shopee is trying to run this exact same play in Brazil right now. But MercadoLibre is fighting back with a different strategy. When Shopee first entered Brazil, MELI first focused on investing billions of dollars to improve their logistics network to support great capacity. Then, they slashed their free shipping minimum in Brazil from 79 reais all the way down to 19. By taking a short-term hit on these cheap items, MELI is aggressively taking back the exact low-cost, everyday purchases that Shopee and Temu rely on. And it’s working. In Q3 2025, the first full quarter with this new threshold, items sold in Brazil shot up by 42% year-over-year, and new listings in that R$19 to R$79 sweet spot tripled, with the continued momentum rolling into Q4.

Decentralized vs. Centralized Networks

A lot of global investors mistakenly think Mercado Envios is just a Latin American Amazon clone. Sure, both cost a ton of money to build, but they are wired completely differently. Amazon built its system for the smooth roads and mature infrastructure of the US and Europe, focusing on bigger, more expensive orders and Prime memberships. But Amazon’s model struggles to make money on a small ~$4 purchase. MercadoLibre, on the other hand, custom-built its network for Latin America. It’s very decentralized. Yes, they have massive warehouses, but their strength is also how deep they reach into local neighborhoods. They also use crowdsourced drivers and turn thousands of local family shops into mini-delivery hubs. This setup lets them bypass bad roads and get packages into remote or tricky areas where a standard delivery van just can’t go.

The Slow Logistics Layer

The trick here is figuring out how to compete on cheap items without ruining long-term profits. MercadoLibre solved this by building their own slow logistics layer. They constantly have delivery trucks on the road so when someone buys a cheap item just to hit that new R$19 free shipping minimum, MELI doesn’t rush it for next-day delivery. Instead, they toss it into the slow layer, batching these cheap items into any leftover, empty space on their existing trucks. Even though they heavily subsidized millions of cheaper orders in late 2025, their actual shipping cost per item in Brazil still dropped by 11% year-over-year.

The Trap of Asset-Light Models

Ultimately, MELI can handle the squeeze on their profit margins because their physical infrastructure is so dominant. Their managed network handles ~96% of the stuff sold on their platform. Because MELI handles so many packages in-house, it will continue to get internally cheaper for them to deliver each box over time. Historically, Shopee kept prices low in Brazil by being completely asset-light, meaning they didn’t own the trucks or warehouses, but this is quickly changing. That works at first, but it breaks down at a massive scale. When you outsource your shipping, every single package costs you a set fee, and the outside courier is the one pocketing the extra profit as the delivery routes get busier. The third-party logistics can help complement the network but the ideal state is internalizing as much of the costs as possible.

While Shopee has recently accelerated its shift toward investing in its own local distribution infrastructure and drivers, they still face significant capital expenditure hurdles to match MELI’s scale. Historically, relying on third-party couriers kept their upfront costs low, but moving forward, bridging the gap to a fully internalized network will require them to burn heavy capital while simultaneously subsidizing cheap shipping. MercadoLibre did the exact opposite by spending heavily upfront, opening 16 new fulfillment centers in 2025 alone. When your network is already moving 752 million items a quarter, tossing one more package on an owned truck that’s already driving the route costs practically nothing. This leaves competitors trapped. The asset-light competitors are stuck paying rising fees to middlemen, Amazon is stuck with a rigid system that can’t easily handle cheap orders, and anybody else that wants to compete has to pivot and spend exponentially more.

The physical scale of this local moat is staggering. According to MercadoLibre’s 2025 Impact Report, the company processed over 2.4 billion products last year. By leveraging over 12,458 local establishments as decentralized “MELI Places” for pickup and drop-off, they have engineered a localized network where 75% of all express shipments are delivered in under 48 hours. This infrastructure is a massive economic engine, directly generating over 150,000 jobs linked to SME operations in 2025 alone.

Battlefront 2: Controlling the China-LatAm Corridor

Upstream Vertical Integration

The route from Asia to Latin America used to be a mess of fragmented drop-shipping and slow local mail, with packages sitting in customs for weeks. Competitors tried to make up for this experience by heavily subsidizing prices. MercadoLibre has a different perspective of this $10 billion cross-border market and they treated it like an engineering problem. They knew dominating local deliveries wasn’t enough. They had to control the middle-mile. This is the trans-Pacific air freight. Since Asian rivals naturally control the starting point in China, MELI is focusing on building their own consolidation centers right in China. They cut out the middlemen and feed the cargo straight into their own Latin American air fleet. That squeeze on their profit margins is the cost of building this massive bridge from China. In December 2025, MELI opened its first fulfillment center inside mainland China, letting sellers drop off goods to be batched and packed before they leave Asia. They didn’t do this alone, though. They partnered deeply with Cainiao Network, Alibaba’s global logistics partner, securing shipping routes to the Americas at a discount to the industry average cost. To navigate China’s complex export rules, they hired Mah Soon Sin, the former Legal Director for Cainiao and AliExpress, as their China Head of Legal in early 2026.

Automation

To protect their profits from rising wages, MELI is spending heavily on automation. In mid-2025, they signed a $43.4 million deal with Damon Technology to install smart sorting systems in Mexico, Brazil, and Argentina to handle the massive flood of Asian packages. They started testing bipedal robots from Agility Robotics at their Texas facility to help handle heavy lifting and high warehouse turnover. But the real game-changer happens when a plane lands in São Paulo, Mexico City, or Santiago. Instead of handing packages off to slow national postal services, MELI uses those new Damon Technology sorters to instantly process the cross-border parcels and drop them into their own delivery fleet. Because of this automation, shipping times from China have dropped to as low as five days on the best routes. MELI effectively turned a slow, low-margin process into a very reliable extension of its own network.

Insights from Management

The demand from Asian manufacturers to tap into this ecosystem is strong. During a recent seller expo event in China, MercadoLibre hosted 1,500 physical attendees alongside 70,000 monitoring the livestream, according to VP of Cross-Border Trade Karen Bruck. To convert this massive attention into actual inventory, MELI is deploying a highly capital-intensive “Fully Entrusted” consignment model. Chinese merchants ship their inventory to a domestic Chinese warehouse, and MercadoLibre absorbs the entirety of the operational friction. They are handling the pricing, local listings, multi-currency compliance across six countries, and customer service. While this aggressively locks in high-quality suppliers, it carries more operational overhead. Taking over localized customer service and returns for thousands of foreign merchants is exactly the kind of margin-squeezing investment that spooked Wall Street, but it is a necessary expenditure to definitively win the corridor.

Assuming this “Fully Entrusted” model is purely a margin-draining cost center misses the unit economics. By sourcing inventory directly from small and medium-sized manufacturers in China, MercadoLibre is actively removing middlemen from the equation. Because MELI takes on the operation, they gain greater control over the final consumer pricing, making this specific consignment model particularly attractive from a margin perspective. They are essentially capturing first-party retail margins on factory-floor production costs, transforming an upfront logistical expenditure into a highly profitable, long-term wholesale arbitrage.

Validation from the J&T Global’s Earnings Call

You don’t just have to take MELI’s word for it. Their partners are seeing the results. During the Q4 2025 earnings call for J&T Express, a massive global logistics company, management specifically shouted out MercadoLibre. J&T’s executives made it clear that MELI is the biggest player in Latin America and a key partner for their regional deliveries and expansion. They noted that MELI’s aggressive push into Asia is bringing a ton of Chinese merchants over, especially those selling high-volume items like car parts and phone cases. J&T Express has repeatedly highlighted their deepening cooperation with MercadoLibre, eventually announcing they had won the “Mercado Libre 2025 Best Carrier Award.”

In their Q1 2026 earnings call, J&T Express reported that parcel volume in their “Other Markets” segment, which is anchored heavily by Brazil and Mexico, accelerated to 150 million pieces. That is marking over 100% year-over-year growth. J&T management explicitly pointed to their deepening ties with MELI, confirming they “increased our partnership since last year” and expect further collaboration as MercadoLibre expands across Latin America. J&T also highlighted that the per capita parcel volume in Latin America sits at just 11. Compared to the highly saturated markets of China and Southeast Asia, this proves the China-LatAm corridor is not just a high-volume operation today, but a pipeline with a massive, untapped ceiling for future logistics capital.

Battlefront 3: Turning Friction into an Advantage

The Regulatory Catalyst

For a long time, cross-border apps used a tax loophole to flood Latin America with cheap products. To stop this, the Brazilian government rolled out the Remessa Conforme program, which locals call the Taxa das Blusinhas. This new rule officially ended the old $50 import tax exemption. Now, imported goods get hit with a mandatory 20% federal tax plus a 17% state value-added tax. Because of how these taxes compound on top of each other in Brazil, the total tax burden actually jumps to as high as 44% of the final retail price.

MercadoLibre’s “Full China” integration shifts the unit economics away from what previous competition focused on. Instead, MELI is establishing a highly efficient B2B wholesale importation model, funneling bulk inventory directly into their local Latin American fulfillment centers to legally optimize landed costs and completely shield its top Chinese merchants from administrative nightmares. They basically took a massive regulatory roadblock and turned it into a competitive edge, becoming the safest and most legally compliant way for these merchants to sell into Brazil.

Agentic AI

But having great physical warehouses isn’t enough to dominate international trade. Trying to get tens of thousands of Chinese merchants to easily sell to Spanish and Portuguese speakers creates massive friction. To fix this, MercadoLibre went beyond basic translation tools and rolled out advanced AI. Their AI Seller Assistant basically acts as an autonomous digital manager for Chinese businesses. Using large language models, the AI automatically translates text, cleans up product details, and tailors millions of listings to local cultures. MELI even checks out local competitors to dynamically suggest the best pricing. This completely removes the language barrier that usually stops small Chinese businesses in their tracks. Even more impressively, MercadoLibre isn’t forcing these foreign merchants to learn a proprietary, clunky seller dashboard. To ruthlessly eliminate onboarding friction, MELI is integrating its tools into the platforms Chinese sellers already use day-to-day, specifically WeChat. By launching their seller assistant directly on WeChat, they are meeting the supply side exactly where they already live, seamlessly embedding Latin American e-commerce into the daily digital routines of Chinese factories.

During their Q4 2025 earnings call, management confirmed the efficiency of this invisible moat, noting that transactions advised or directly influenced by the AI Seller Assistant account for approximately 20% of the platform’s total GMV. This means an AI tool is actively optimizing nearly $4 billion in GMV per quarter.

Battlefront 4: The Financial Ecosystem Lock-In

The Temu Squeeze

The market assumes all Asian competitors run the exact same playbook, but they don’t. Right now, Temu is actively squeezing its Chinese suppliers by demanding massive deposits, sometimes upwards of $10,000, and slapping them with strict financial penalties for any issues. Chinese merchants are looking for a way out of this, and MercadoLibre gives them a structurally better home. Since MELI’s average order price is over $20, they aren’t trying to fight for the absolute cheapest, lowest-quality goods. Instead, they are able to pitch their “Full China” route as a premium channel, attracting higher-quality factories that actually want to keep their pricing power.

The Ecosystem Alpha & Credit Risk

MercadoLibre’s biggest physical advantage is its leadership in local delivery network, which would cost Asian competitors a lot of time and money to copy. But the final layer of their dominance is how they tie in their fintech. Pure cross-border apps only compete on price, but MELI has logistics and finance seamlessly blended together. In 2025, Mercado Pago processed a massive $277.8 billion in total payments. Their highly profitable credit portfolio grew by 90% year-over-year to hit $12.5 billion, while bad credit card loans dropped to a historic low of 4.4%. Plus, they now have nearly 78 million people actively using the fintech app every month. Basically, MELI controls the Chinese supply chain, takes a short-term hit to offer cheap R$19 shipping in Brazil, and then finances those exact purchases through its own digital bank.

Expanding a $12.5 billion loan book this fast in Latin America, especially by extending credit to the lower-income shoppers attracted by that new R$19 shipping tier, naturally brings up fears of default risk. Even though bad loans are at stable lows, big institutions want a wider safety net in case the economies in Brazil or Mexico take a sudden hit. But looking at this solely through a risk lens misses the strategy. For Mercado Pago, this credit portfolio is basically just a customer acquisition cost. By hooking underbanked shoppers with cheap items and instant credit, MELI is fighting a war to become their primary bank. The e-commerce profits they are sacrificing on shipping today are funneling millions of new users into a highly lucrative fintech ecosystem. And since these loans are very short-term (around 4 months), MercadoLibre could quickly recalibrate if the economy crashed. This would allow them to bounce back much faster than a traditional credit business ever could. The risk is still very real as a macro shock would leave MELI exposed.

The Argentine Alpha

While Brazil is the main battlefield, you can see the real strategic power of MELI’s China network over in Argentina. For years, Argentina suffered a terrible macroeconomic environment and incredibly strict import bans. But heading into 2026, everything shifted rapidly when President Javier Milei started aggressively deregulating the economy. Because MELI had its Chinese supply chain ready to go, they immediately capitalized on years of pent-up consumer demand the second those trade barriers fell. This direct integration helped drive a massive 42% surge in FX-neutral GMV in Q4 2025. This proves just how adaptable their China strategy is. It functions as a cheap defensive weapon in cutthroat markets like Brazil, and as an instant supply-shock mechanism in newly opened markets like Argentina.

US Goods

While the broader market is hyper-focused on the war against Asian competitors, MercadoLibre is quietly expanding its cross-border moat to the North. This strategy is not strictly a China play. In 2024, they opened a remote fulfillment center in San Antonio, Texas, specifically designed to allow established US brands to bypass export complexities. US sellers ship their inventory into this Texas hub, and MELI handles the cross-border logistics directly into high-value markets like Mexico. This actively diversifies their product assortment away from ultra-low-ticket Asian goods and funnels premium American inventory into their pipeline, establishing a structural defense against the pure-play, cheap-goods models. They are actively expanding US cross-border trade, which is doing very well in Mexico, growing in Argentina and now starting in Brazil.

The End Game

Capital Intensity and Reverse Logistics

Naturally, investors are waiting to see exactly when MercadoLibre’s profit margins will finally bounce back. Building out this massive infrastructure is incredibly expensive. Just keeping the ecosystem running costs millions. Because these massive costs are front-loaded, the market might keep discounting MELI’s stock until the financial payoff becomes obvious.

There are also some very real risks to this “Full China” strategy, especially when customers return items. If you are a seller sitting in China, it is way too expensive to ship a returned product all the way back across the ocean just to inspect or fix it. Because of this, returned goods often just sit in Latin American warehouses, piling up storage fees and getting written off as unsellable. To avoid this logistical nightmare, sellers are increasingly just giving customers a refund and letting them keep the item, eating the total cost of the product. Unless MELI builds a strong local marketplace to liquidate these returns, this reverse logistics problem will remain a headwind on profits and could scare off the exact Chinese merchants they are trying to attract.

Accelerating Consolidation

Right now, competitors are locked in a brutal, localized war of attrition. MercadoLibre is betting that its massive digital and physical footprint will eventually create an insurmountable barrier to entry. If management can keep a lid on bad credit loans while finishing this build-out, the recent drop in the stock price might actually be a historic opportunity.

If you want to know what the financial payoff looks like, look at Mexico. This cross-border strategy is already becoming a proven, scaled operation, representing more than 15% of MercadoLibre’s total GMV in Mexico. Across the board, the cross-border platform grew more than 60% year-over-year in the last half of the year. The real upside, however, lies further south. While Brazil accounts for more than half of the company’s total business in Latin America, management admits that it remains pretty untapped from a cross-border perspective. With the infrastructure now in place, including newly opened physical offices in Shenzhen and Shanghai to lock in supply, the expected cross-border growth runway in their most dominant market is looking massive.

Ultimately, this drop in profit margins isn’t a sign that MELI is struggling to compete. It is a highly calculated investment. By using their strong local delivery network and their massive fintech bankroll, MELI is systematically pricing out rivals who rely on tax loopholes and asset-light logistics. By choosing to take a short-term hit to their profits today, MercadoLibre is actively speeding up its domination of Latin American e-commerce.

Subscribe now

The Consensus is Broken

The prevailing bear case for ServiceNow and the broader software as a service sector is built on the notion that generative AI increases efficiency, corporations reduce headcount, and seat-based software revenue collapses. This consensus has driven a severe compression in ServiceNow’s valuation, effectively pricing it as a legacy system of record facing secular decline.

This consensus is structurally obsessed with seat compression while missing the pricing expansion currently underway.

ServiceNow is evolving from just selling software licenses to effectively capitalizing digital labor. Through the introduction of Pro Plus pricing and the Now Assist Token consumption model, the company is executing a pivot from a tool that taxes by the number of humans to a utility that taxes by the work done.

The market views $NOW as a deflationary victim. The data suggests it is the infrastructure layer of the autonomous economy.

The Unit Economics of the Assist Token

The consensus bear case relies on 2015 math where a customer that reduces headcount by 15% causes ServiceNow to lose 15% of its seats, and growth collapses. This view is now factually incorrect because it ignores the dual-layer pricing architecture ServiceNow has engineered to hedge against labor deflation.

The new model functions as a SaaS Floor with an AI consumption Token Ceiling. CFO Mastantuono describes this as a “reload” mechanism where “tokens are used, then they get reloaded,” creating a built-in revenue hockey stick.

Pro Plus and Discount Normalization

ServiceNow has created a structural pricing uplift that acts as a hedge against seat compression. To access any GenAI features, customers must upgrade to the Pro Plus tier. This typically triggers a ~60% uplift in average selling price per seat.

But the floor is stickier than just the list price. Institutional pricing data shows that ServiceNow utilizes discount normalization to compound revenue. Even if a customer’s seat count remains flat, the effective spend index doubles by year 7 as initial discounts burn off. This ensures that even in a headcount neutral environment, the contract value naturally inflates.

The Operating Leverage of Digital Labor

This is the pivot point the market is missing. The Assist Token model taxes the outcome.

When a customer replaces human labor with AI, they potentially increase velocity on top of saving money. In Q4 2025, management cited a leading US consumer services company that purchased a $1 million assist pack. After realizing a 400% ROI, the customer scaled usage, requiring “eight times more assists” in year two.

The New Math

Now let’s consider a scenario where a client lays off 15% of their support staff. ServiceNow loses 15% of low-margin legacy seats. The remaining 85% of seats are upgraded to Pro Plus (+60% ASP). The work previously done by the 15% is now executed by AI Agents burning Assist Tokens.

Because digital agents are faster and cheaper than humans, the volume of work explodes. Management notes that “consumption creep” is now the fastest-growing cost driver, often sitting outside traditional procurement controls.

ServiceNow has effectively swapped a linear headcount tax for an exponential productivity tax.

The Government Efficiency Arbitrage

While the market obsesses over enterprise capital expenditure, a far larger shift is occurring in the public sector. The systematic conversion of government labor operating expenses into ServiceNow revenue.

In Q3 2025, ServiceNow and the US General Services Administration announced the “OneGov” agreement. This is not a standard procurement vehicle. It is streamlining the years-long Request for Proposal cycles that typically plague federal tech adoption.

Institutional Efficiency Mandates

During the Q1 2025 earnings call, CEO McDermott explicitly confirmed “productive discussions with... DOGE officials.” This is the highest signal comment in the fiscal year. The current administration’s mandate to “modernize operations and return savings to the American people” creates a forced migration from labor to software.

ServiceNow is positioning itself not just as a vendor, but as the execution layer for federal cost-cutting. In early pilots, one federal agency using ServiceNow AI agents to automate contract reviews projected 30% efficiency gains across the board. This is already showing in the numbers. U.S. Public Sector revenue grew over 30% YoY in Q1, with 11 federal deals exceeding $1M (up from 8 the prior year).

Bypassing the Swivel Chair

The Federal government is the largest employer in the U.S., and its workforce is famously fragmented across legacy systems (the “swivel chair” problem McDermott frequently cites). In the old way the government hired more people to bridge the gap between disjointed systems (such as manually moving data from a mainframe to a spreadsheet). In the OneGov way ServiceNow wraps the legacy estate in a single data model. The government doesn’t need to rip and replace 40-year old mainframes (this is a high risk endeavor) because they just need to overlay ServiceNow to automate the workflow.

The Fiscal Arbitrage

The U.S. government sees immense value in software that stabilizes long-term budgets. OneGov allows agencies to transition from the complex, variable costs of a rapidly expanding workforce to the predictable, transparent pricing of Assist Tokens.

ServiceNow has effectively secured an exclusive license to tax the modernization of the U.S. Government.

The Unbreachable Moat

One bear case for ServiceNow assumes that generative AI is a democratizing force. If everyone has access to the same LLMs, the software layer becomes a commodity.

This view is fundamentally flawed because it fails to distinguish between raw intelligence provided by an AI model and situational awareness provided by the business context. In an autonomous enterprise, the competitive advantage does not come from the model itself. The real value belongs to the platform holding the cleanest, most real-time record of the business’s operations.

ServiceNow’s moat is built on three structural pillars that competitors cannot mathematically replicate without rewriting their entire existence.

Single Data Model

AI agents are only as good as the data they can access. If an AI has to hop between different databases to resolve a query, it introduces latency and hallucination risk. Competitor platforms are stitched together through decades of mergers and acquisitions. Salesforce is effectively a federation of acquired databases such as Tableau, MuleSoft, Slack, and Demandware loosely connected by application programming interfaces. An AI agent trying to navigate this landscape faces high friction and fragmented context. ServiceNow was built on a single data model from day one. There are no acquired codebases tacked on. Whether it’s IT, HR, or Customer Service, every workflow shares the exact same table structure. In the age of AI, integration debt is effectively technical bankruptcy. ServiceNow has little integration debt within its own platform. This allows their agents to move laterally across the enterprise with little friction, while competitors struggle.

RaptorDB Pro and AI Agent Speed

Human employees work at human speeds, which takes at least minutes of time. AI agents work at machine speeds, which takes seconds of time. Legacy databases (MariaDB, Oracle SQL) are the bottleneck that cannot effectively handle the load of high-frequency agents. ServiceNow migrated its underlying plumbing to RaptorDB Pro (a heavily modified, high-performance PostgreSQL fork). RaptorDB Pro delivers 27x faster analytics and 53% faster transaction times. ServiceNow has effectively built the AI agent layer for enterprise work. While competitors are running on legacy rails designed for humans, ServiceNow has laid the high-speed infrastructure for digital labor.

Armis and the Physics of the Enterprise

Software companies usually stop at the software layer. With the ~$7.75B acquisition of Armis, ServiceNow is crossing over to the physical world. A traditional Configuration Management Database tracks servers and laptops. It doesn’t track the MRI machine in a hospital, the robotic arm on an assembly line, or the HVAC system in a data center. Armis gives ServiceNow real-time visibility into Operational Technology and Internet of Things assets. This makes ServiceNow the only platform that can orchestrate a workflow that starts in the digital realm (a cyber threat detected) and ends in the physical realm (shutting down a specific factory valve). Competitors have little visibility here. They are blind to the physical assets that actually run the economy.

The AI Control Tower

In 2026, the biggest risk to a CEO is that their AI is too fast and does something costly. Enterprises will soon have thousands of agents from different vendors (Microsoft Copilot, Salesforce Agentforce, custom Python agents). ServiceNow is positioning itself as the authority of enterprise AI. The AI Control Tower allows the business to govern all agents, regardless of who built them. By owning the governance layer, ServiceNow ensures that even if a customer builds an agent in Microsoft Copilot, they still pay a tax to ServiceNow to validate that agent’s actions against company policy. They monetize their competitors’ innovation.

The One of One Reality

ServiceNow has evolved from just an Information Technology Service Management tool to the Operating System of the Autonomous Enterprise. There is no “next ServiceNow.” The conditions that allowed it to be built with 20 years of organic development on a single code base without dilutive M&A cannot be replicated by any incumbent or startup today.

The Risks

While the bull case for ServiceNow is structural, the risks are temporal and executional.

The J-Curve Revenue Gap

The most potent risk is a timing mismatch between seat deflation and consumption inflation. Let’s say a Fortune 500 customer deploys ServiceNow Agents and immediately cuts 20% of their support staff (reducing seat revenue). They may initially hoard efficiency gains rather than immediately spend more on tokens. Management already mentioned that businesses likely need more time to trust in the return on investment in assist packs. This creates a temporary dip in revenue growth where seat revenue falls faster than token revenue ramps. If this happens, ServiceNow could miss its 20% growth target for a couple of quarters, triggering further multiple compression.

The Armis Integration Drag

The ~$7.75B Armis acquisition is a rare deviation from ServiceNow’s history of organic scaling, and it presents a unique operational test. Unlike competitors who simply bolt on acquisitions, ServiceNow’s single platform standard requires a deep, native integration of Armis’s OT and IoT data. This is a massive engineering lift. The risk is that maintaining this architectural purity could distract from the critical Pro Plus AI rollout, or signal to investors that the company must now buy growth to supplement its core engine.

The Agentforce Counter-Attack

ServiceNow’s growth depends on expanding beyond IT into Customer Service Management. However, Salesforce has launched a competent counter-offensive with Agentforce. While ServiceNow owns the back office (IT/Ops), Salesforce owns the front office (Sales/Service). If Salesforce successfully convinces CIOs that Agentforce can handle IT tickets (encroaching on ServiceNow’s core), it forces a pricing war. ServiceNow needs to prevent Salesforce from commoditizing the agent layer. If agents become a commodity, pricing power collapses.

The Discount Normalization Backlash

As mentioned previously, ServiceNow relies on discount normalization (reducing discounts at renewal) to drive growth. In a macro environment where CIOs are under pressure to cut costs, aggressive renewal tactics can backfire, leading to churn or shelfware audits. If customers feel held hostage by rising costs without seeing immediate AI value, net retention rates (currently elite at 98%) could slip.

Current Valuation

Despite the latest drawdown, ServiceNow is not optically cheap. It still commands a premium multiple. However, that valuation is a little misleading. The market is currently pricing $NOW as if its growth will permanently taper off, assigning little value to the upside of successfully transitioning from seat-based to outcome-based revenue. Their stock-based compensation is also becoming less impactful to revenues and free cash flows over time.

The Plumbing of the Future

The market views NOW as a legacy seat-based software company facing a deflationary AI wave, and it has priced it accordingly. ServiceNow stock has suffered a ~50% drawdown from its highs, compressing its valuation multiple to levels not seen since it was a pure-play IT helpdesk tool.

This view fails to account for the Token Economy. If companies hire people, ServiceNow wins on seats (linear). If companies use AI, ServiceNow wins on Tokens (exponential). By building the OneGov infrastructure and the AI Control Tower, ServiceNow has effectively built the toll road for the Agentic Enterprise.

The core business is at a historical discount, with a free call option on the largest productivity shift in history. This is an arbitrage on the market’s skepticism and inability to model the math of digital labor.

Disclaimer: The information provided in this newsletter is for educational and informational purposes only and should not be construed as financial or investment advice. I am not a professional financial advisor. All content reflects my personal thoughts and research. I make no representation as to the accuracy or completeness of any information found here.

Forward-Looking Statements: This post contains forward-looking statements (such as predictions about ServiceNow’s future economics and the adoption of the “Token Economy”) which are speculative and subject to change.

Disclosure: As of writing this, I hold a position in ServiceNow ($NOW). I may buy or sell shares at any time without notice. Please conduct your own due diligence before making any investment decisions.

The market still treats Netflix like a media company selling subscriptions. But Netflix has evolved into something far more dangerous: a yield management engine for leisure time. The alpha isn’t in the subscriber count anymore. The real weapon is buried deep in their technology edge.

The Exclusivity Tax

For the last two years, Netflix operated with a passive monetization strategy. By outsourcing its ad tech to Microsoft (via Xandr), they were able to launch ad tier quickly and have an insurance policy against slowing growth. The Microsoft deal included a massive minimum revenue guarantee. This was critical especially when their ad tier had zero initial subscribers. However, Netflix paid a fee estimated to be 15% to 30% of every ad dollar. Now the training wheels for Netflix are off because they are on track to chase the upside they have been missing out on.

The real killer wasn’t just the fee. The structure of the Microsoft partnership created operational rigidity that limited inventory liquidity. By keeping inventory within a closed ecosystem, the partnership inadvertently fragmented demand. This wasn’t just a margin play, it also gave them more control. Netflix can now pipe bid requests directly into multiple demand side platforms to force price discovery upwards. This strategy is coupled with the utilization of dynamic insertion for high-value executions, such as live events.

Owning the Tech Stack

Netflix’s model differs from the closed-loop ecosystems typical of the last decade of ad-tech. Netflix is effectively acting as a massive supply side platform. They control the decisioning engine, allowing them to pipe demand from multiple major DSP’s into a unified auction

By forcing the buy-side to plug into their sovereignty, Netflix changes the power dynamic. They evolved from just selling ad slots to commoditizing the DSP’s role. Netflix is shifting leverage back to the publisher side of the equation.

The Scarcity of Quality

Advertisers aren’t bidding up the price simply because there are more slots. They are bidding it up because they have an addressable ad-tier that is growing faster than any linear network. Netflix offers clean inventory that has massive frequency control. Because they own the server, they can ensure a user sees the ad once at a high impact. This efficiency is what creates a floor for CPMs that defies standard supply/demand gravity.

Subscribe now

The Churn Risk Algorithm

Unlike legacy ad servers designed to max out CPM for the current slot, Netflix solves for maximum lifetime value. By capping frequency at the subscriber level rather than the show level, Netflix prevents the repetitive ad fatigue that plagues linear TV, ensuring the ad load never crosses the threshold that triggers churn.

This is Frequency Capping 2.0. The legacy model caps ads by time but Netflix caps ads based on psychology.

Monetizing the Danger Zone

If the algorithm senses ad fatigue pushing a user toward the cancel threshold, it suppresses the inventory. The genius is not just suppressing the ad for tired users but monetizing the mechanism itself. Take the Keep Watching format. When a user binges three episodes in a row, the algorithm may potentially flag them as high churn risk (interrupting a flow state is dangerous). Instead of showing a standard commercial, Netflix serves a binge sponsorship. This is an ad-free episode presented by a brand. When Netflix offers an ad-free episode for binging three in a row, it isn’t just a gamified perk. Netflix successfully bundled the revenue. They charge the advertiser a premium for the hero moment of not showing an ad. This strategy turns a retention liability into a high-CPM product.

The algorithm is not designed to just toggle ads on and off. It adapts the format intensity. Low risk users get standard high interrupt ads, medium risk users get low interrupt pause ads and high risks users get zero interrupt sponsored ads. By dynamically matching the ad format to the user’s patience, Netflix achieves the holy grail of maintaining high ARPU without triggering the cancel button.

The WBD Catalyst is Bid Density

Now, look at the Warner Bros. Discovery deal through this lens. Wall Street hates the debt load. They see a balance sheet problem. They analyze this like a studio merger, worrying about culture clashes and interest payments. They are missing the arbitrage. It isn’t about cost savings. It’s about the unified buy.

The Supply Glut vs. The Frequency Solution

One bear case is that flooding the market with WBD inventory creates a supply glut that crashes the prices. Although this concept is true in a vacuum, Netflix is not just buying impressions because they have the ability to deduplicate them. If a user watches a show on Max and then watches a show on Netflix, they may get the same ad today, which is wasteful. By owning the ad stack, Netflix can pipe WBD’s inventory through Netflix’s ad server and institute global frequency capping. They can sell a universal reach that guarantees a user sees an ad once across the entire ecosystem. Advertisers pay a massive premium for this efficiency.

The Sports Bridge

Another bear case is that WBD lost some critical sports assets such as the NBA. While the domestic live rights are gone, WBD retained a massive international sports footprint and the digital highlight rights. When Netflix buys the Streaming and Studios unit, they inherit the streaming tier of these rights. They are acquiring a turnkey sports infrastructure with production, licensing, and ad sales that would take a decade to build from scratch.

Becoming the Market Maker

This scale solves one of Netflix’s primary challenges: competing with the scale of massive social video platforms. Until now, massive digital platforms dominated ad inventory through sheer volume. By combining Netflix’s 300M+ subs with WBD’s massive footprint, they create a bundle that rivals the scale of social aggregators but with superior, brand-safe content. They become the market maker for premium video. They are not just setting price floors, but compelling agencies to consolidate their spending on high-quality inventory.

The Key Risks

The thesis is not complete without stress testing the bull-case.

Execution Risk

Merging tech stacks is notoriously hard. Disney’s integration of Hulu was a multi-year engineering challenge. If Netflix’s pipelines can’t ingest WBD’s library metadata and Max subscriber viewing history cleanly, that churn algorithm loses its edge.

The bigger risk is that a talent exodus at HBO begins the moment the deal closes. If Netflix tries to force its ideals onto HBO’s development process, the show-runners who created The White Lotus and Succession could defect to Apple or Amazon. Netflix is buying a library, but they need to retain the architects or else they destroy their assets.

The Entanglement Trap

The spin-off of the linear networks sounds clean on paper, but legal reality is messy. Most of WBD’s remaining sports rights (March Madness, NHL, MLB) are structurally tethered to their linear cable networks. The streaming rights exist primarily as simulcasts of the linear feed. The risk is that Netflix cannot cleanly sever the streaming rights for Max without triggering potential legal clauses. If the leagues block the transfer to a spun-off entity, Netflix could be forced to retain and operate declining linear assets it preferred to shed.

The Antitrust Pivot

As of 2026, the DOJ’s definition of monopoly is the swing factor. If the market is defined as Premium Streaming Video On-Demand (Netflix, Disney+, Max), this deal could be considered a monopoly (combined ~40% share) and will be blocked. Netflix’s legal team will argue the market is total video attention (including Big Tech platforms). In that massive ocean, Netflix and WBD are not considerably larger than most of the top competitors. The recent FTC cases suggest the courts are moving toward this broader definition.

The Break-Up Fee

Wall Street is pricing in the chance of a regulatory failure. The WBD deal includes a $5.8 billion break-up fee that Netflix has to pay in the event the deal gets blocked. While painful, it is not fatal. The real cost of a blocked deal would be the opportunity cost. Netflix would lose out on a great opportunity to expand its ecosystem.

Macro Risk

Netflix is moving from a pristine balance sheet to a leveraged one with some equity dilution. Generally, Netflix is trading content risk for execution risk. With the deal potentially closing in late 2026, Netflix is betting on a soft landing. If the economy turns before the deal closes, they are acquiring a levered asset into a contracting ad market. This wouldn’t kill the strategy, but it would suppress the payoff timeline significantly.

The Trade

The market sees a streamer trying to get bigger, weighed down by a terrifying leverage. I see a data company financializing leisure time.

Netflix isn’t trying to buy cash flow. They are manufacturing it by running better software on better content. The legacy TV assets will be gone. They will keep the growth engines (HBO, DC, Warner Bros. Studios) and will hook it up to their sovereign stack. That’s the trade.

How do they service the leverage without the cable cash flow? Yield. WBD was monetizing its superior IP with a legacy sales team and fragmented data. Netflix will monetize that same IP with its own sovereign ad stack, applying the churn risk algorithm and bid density to multiply the yield per viewing hour.

This deal is about operating system dominance. Wall Street focuses on debt. They should be focusing on the moat. If this deal closes, Netflix becomes the default operating system for global entertainment. They will evolve from just competing with the likes of Disney to competing with the entire global attention economy.

Disclaimer: The information provided in this newsletter is for educational and informational purposes only and should not be construed as financial or investment advice. I am not a professional financial advisor. All content reflects my personal thoughts and research. I make no representation as to the accuracy or completeness of any information found here.

Forward-Looking Statements: This post contains forward-looking statements (such as predictions about Netflix’s future economics) which are speculative and subject to change.

Disclosure: As of writing this, I hold a position in Netflix ($NFLX). I may buy or sell shares at any time without notice. Please conduct your own due diligence before making any investment decisions.

The catalyst is nuclear energy.

Subscribe now

The quarterly filings, the transcripts and the strategic partnership documents reveal a thesis that the market is overlooking. When you isolate the unit economics of the nuclear portfolio, the implied valuation suggests the market is effectively pricing Brookfield’s nuclear business at a steep discount.

Here is the deep dive on the Brookfield nuclear strategy.

The $80 Billion Signal

In the Q3 2025 Letter to Shareholders, buried amidst the standard commentary on “resilient cash flows,” was a detail that signals a paradigm shift:

“...a transformational partnership with the U.S. government to deliver $80 billion of new nuclear plants in the U.S.”

This is capital deployment at a nation-state scale. Crucially, the partnership introduces a mechanism rarely seen in public markets: asymmetric sovereign alignment.

The U.S. Government receives a participation interest entitling it to 20% of cash distributions, but only after Brookfield and Cameco have recouped a $17.5 billion hurdle. By converting the U.S. Government into a partner with a vested interest in the equity upside, Brookfield has structurally mitigated a primary risk in nuclear development: political volatility.

While some will see this as a cap on upside, in nuclear, the primary risk is completion. Giving the U.S.government a 20% participation interest after locking in $17.5 billion is a cheap insurance policy for Brookfield.

The government has effectively outsourced the managerial execution risk to Brookfield (via Westinghouse) while retaining financial tail risk via the hurdle structure. Why? Because the government has regulatory authority, but lacks operational efficiency. Brookfield has the balance sheet ($178 billion in deployable capital) and the technical capability.

The Hardware Catalyst: Santee Cooper

While the $80 billion partnership is the macro signal, the execution is happening on the ground in South Carolina.

On December 8, 2025, Brookfield signed a Memorandum of Understanding to pursue the restart of the abandoned V.C. Summer nuclear project. For context: V.C. Summer was a failure. It was abandoned in 2017 after billions in cost overruns, leaving two partially built AP1000 reactors dormant.

The landscape has changed since 2017. The Vogtle 3 & 4 units in Georgia are now operational, proving the AP1000 viability. The supply chain is warm.

The details of the Santee Cooper MOU are significant:

• $2.7 billion potential cash payment to Santee Cooper (contingent on final investment decision).

• 25% ownership stake retained by the utility.

• 2.2 Gigawatts of clean, baseload power brought online.

While the original licenses were terminated in 2019, the data remains valid. In nuclear regulation, the timeline killer is site characterization (seismic analysis, flood modeling, soil sampling). For V.C. Summer, this work is already codified in the NRC’s database. Brookfield isn’t starting from scratch. They are starting from a paused state, potentially shaving years off the standard timeline.

The Liability Shield

Why is South Carolina letting a private asset manager take over their grid assets? To create a financial firewall. They are using a legal structure known as ‘bankruptcy remoteness’ to ensure that if the project fails, the state isn't left holding the bag.

The state cannot afford a second failure on its balance sheet. By housing the restart within a private vehicle, Brookfield creates a liability shield. If the project spirals, the financial fallout is contained within the SPV. The state captures the power upside, but maintains plausible deniability.

The Silicon Valley Endorsement

The government de-risks the construction via the $80 billion partnership, while Big Tech underwrites the revenue.

The answer may already be signed. Let’s look at the trend. First came the 10.5 GW big tech agreement in 2024. Now, in Q3 2025, we have a 3 GW partnership with another hyperscaler specifically targeting 24/7 baseload hydro. The language is shifting from 'renewables' to 'carbon-free capacity', which is a direct invitation for nuclear.

Brookfield is not building into a void. They are essentially slotting this nuclear capacity directly into a pre-sold order book.

The Valuation Disconnect

Buying a share of BN today at ~$45 effectively provides exposure to a basket of publicly traded stocks (BAM, BEP, BIP, BBU) worth ~$28 per share (net of corporate-level debt and preferreds). This implies the market assigns a value of roughly ~$17 per share to the unlisted ecosystem, which includes $26B of real estate equity, the Wealth Solutions platform, and the proprietary carried interest rights.

This pricing offers a significant margin of safety. Even if one were to mark the unlisted commercial real estate equity entirely to zero, the current share price implies the Wealth Solutions growth engine and the carried interest rights at a distressed valuation. The market is pricing in a worst-case scenario that ignores the cash-generating reality of the non-real estate assets.

The most overlooked catalyst, however, is the $80 billion nuclear partnership. The market views this primarily as an equity story for BEP, missing the credit arbitrage that flows directly to BN.

The partnership structure creates a proprietary double dip for BN:

• BN captures ~46% of the equity upside via its ownership stake in BEP.

• BN’s Wealth Solutions business can capture financing economics on projects it originates or arranges.

The Risks

Why is the stock trading at ~$45 and not ~$68?

The Interest Rate Drag

Brookfield is built on leverage. While they have insulated themselves with fixed-rate, long-duration debt, the market fears that a sustained 4.5%+ 10-year-yield environment will permanently compress their IRRs. If the cost of capital stays high, the spread they earn on their massive infrastructure projects shrinks.

Office Risk

You cannot ignore the office portfolio. It acts as a heavy sentiment anchor that keeps generalist investors on the sidelines.

Duration Risk

Nuclear IRRs are exceptionally sensitive to time. In a high-cost-of-capital environment, time is the enemy.

If the V.C. Summer restart drags from a planned 4-year sprint to a 7-year slog, the math begins to break. Every month of delay compounds the carrying costs without generating revenue, effectively incinerating the project’s IRR.

The Complexity Discount is Sticky

Wall Street applies a complexity discount because the compounding happens across entities rather than within a single P&L. Brookfield’s structure is moving assets between private funds, public vehicles, and insurance floats. This makes it more opaque. Historically, conglomerates trade at a discount because investors assume where there is complexity, there is risk. There is a non-zero chance this gap never fully closes.

The Ultra-Heavy Forging Bottleneck

The most critical bottleneck in the global nuclear renaissance is ultra-heavy forging capacity. The main facilities capable of forging the 600-ton single-piece ingots required for the AP1000 RPV are Japan Steel Works and Doosan Enerbility.

This creates a binary risk profile. If the existing on-site reactors at V.C. Summer have suffered micro-cracking or corrosion during their 8-year dormancy, Brookfield loses its time arbitrage and joins the back of a multi-year global queue.

However, they currently possess the only available AP1000 hardware in the Western hemisphere. A September 2024 inspection by the Governor’s Nuclear Advisory Council confirmed the site is in “good condition” with significant amounts of well-preserved parts.

Execution Risk

Building nuclear power plants in the West is hard. Labor shortages, regulatory creep, and welding inspections can turn a ~5-year project into a ~10-year nightmare. Even with the lessons from Vogtle, restarting V.C. Summer is a heavy lift.

The risk here is managerial. Managing thousands of skilled tradespeople and adhering to NRC documentation standards requires perfect choreography. If costs blow out again, Brookfield’s reputation for operational excellence takes a massive hit.

The Verdict

This is a time arbitrage. In a world where data centers need power now, time is the most expensive commodity. A greenfield nuclear project is a 2032 story. V.C. Summer, being nearly half-built with major components on-site, is a 2030 story. That 2-3 year delta allows Brookfield to command premium pricing from hyperscalers who cannot afford to wait.

The market is pricing BN based on backward-looking office sentiment. It is ignoring the structural shift that has placed Brookfield at the center of the U.S. energy re-industrialization.

Disclaimer: The information provided in this newsletter is for educational and informational purposes only and should not be construed as financial or investment advice. I am not a professional financial advisor. All content reflects my personal thoughts and research. I make no representation as to the accuracy or completeness of any information found here.

Forward-Looking Statements: This post contains forward-looking statements (such as predictions about the Nuclear market and Brookfield’s future economics) which are speculative and subject to change.

Disclosure: As of writing this, I hold a position in Brookfield Corporation ($BN). I may buy or sell shares at any time without notice. Please conduct your own due diligence before making any investment decisions.

The market continues to price Uber as if it is fighting a war it is destined to lose. The prevailing fear is simple: autonomous vehicles will achieve scale, commoditize the network, and displace the human driver, leaving Uber as a relic of the past.

This view fundamentally misunderstands the mechanics of the mobility market. The future of transport is likely not going to be a monopoly of vertical integration. It is more likely to be a utility grid. And in this grid, Uber has quietly pivoted from a labor aggregator to the system operator.

Subscribe now

The Economics of “Base Load” vs. “Peaking Power”

To understand why the winner-take-all AV thesis fails, you just have to look at the math.

An autonomous vehicle, whether from Avride or Tesla, is a high-CapEx asset. Like a nuclear power plant or a base load generator, it carries heavy fixed costs. To generate an acceptable Return on Invested Capital, that asset must run at near-100% utilization. It cannot sit idle. If a ~$100k AV sits parked during a Tuesday morning lull, it destroys shareholder value.

Human drivers are the opposite. They are the “peaking power” plants. They are low-CapEx, high-marginal-cost assets. Crucially, their supply is elastic. They absorb the volatility of Friday nights peaks, rainstorms, and major events.

Uber isn’t competing with the AVs. It is the only platform capable of balancing them. A standalone AV fleet cannot survive the economics of peak demand. If they buy enough cars to cover Friday night, half their fleet sits idle on Tuesday. They need Uber’s hybrid network to ensure their expensive robots stay utilized while human drivers handle the overflow.

The Austin Proof

We are already seeing this play out in the data. CEO Dara Khosrowshahi noted that in Austin, the average AV on the Uber platform is busier than 99% of human drivers.

This isn’t an accident. It is structural engineering. Uber’s algorithm is feeding steady, predictable volume to the AVs (protecting their ROIC) while using human drivers for the complex, messy, or surged trips. Uber has solved the utilization problem for the hardware manufacturers.

Looking Through the Financial Noise

If you look at the Q3 headline numbers there is a lot of noise. The $6.6 billion net income figure is distorted by tax valuation releases and equity revaluations.

But… if you look at their cash generation, Uber is a machine. Free Cash Flow hit $2.2 billion, converting nearly ~100% of Adjusted EBITDA. The business is generating cash as fast as it reports profit. This is a hallmark of a high-quality compounder. With trips growing 22% and margins expanding, the network effects are actually accelerating alongside AV integration, rather than deteriorating.

The Liability Moat

The final piece of the puzzle, which is often ignored by the tech-focused crowd, is insurance. I admittedly struggled to understand this at first as well.

Uber has constructed a form of liability arbitrage. When a human driver is waiting for a trip, Uber pays for a low-cost contingent liability policy. It only steps up to the expensive, full commercial coverage when a ride is active. Commercial insurance on average is about ~5x more expensive than personal car insurance.

An AV when used as a robotaxi (even for a private owner) does not have this luxury. Regardless of how often it is used as a robotaxi, for insurance purposes it is considered a commercial asset 24/7 and therefore burns cash on commercial premiums every second it is idle. The regulatory hammer that recently came down on Tesla Insurance in California highlights just how difficult it is to bypass commercial fleet requirements. The barrier to entry here isn’t just software. It is the financial reality of liability.

My Conviction

I do not believe that Uber is just a taxi company. I believe it is the operating system for the autonomous world. While the market worries about a robotaxi war, Uber is securing its place as the inevitable grid operator.

Disclaimer: The information provided in this newsletter is for educational and informational purposes only and should not be construed as financial or investment advice. I am not a professional financial advisor. All content reflects my personal thoughts and research. I make no representation as to the accuracy or completeness of any information found here.

Forward-Looking Statements: This post contains forward-looking statements (such as predictions about the AV market and Uber’s future economics) which are speculative and subject to change.

Disclosure: As of writing this, I hold a position in Uber Technologies Inc. ($UBER). I may buy or sell shares at any time without notice. Please conduct your own due diligence before making any investment decisions.

• The Current Wall Street Consensus: Nu and MercadoLibre are a “Cozy Duopoly” with a seemingly impenetrable runway in Latin America.

• My Theory: A three-way war has begun. Sea Limited is attempting to pre-bank Gen-Z gamers. This could be potentially a big headwind for Nu’s future acquisition funnel.

• The Metric to Watch: Yield Dilution. Nu is aggressively pivoting to low-yield Secured Lending (payroll) to build a defensive moat. They are trading Gross Yield for Credit Stability. They are knowingly accepting lower asset returns to protect ROE via lower provisioning costs.

• The Verdict: The nearly infinite runway thesis may be getting a little bumpier. Nu is very attractive, but it may be shaping more into a defensive utility, over a high-growth compounder.

Subscribe now

The Death of the Cozy Duopoly

The institutional consensus on Latin American fintech has shaped into a potentially dangerous duopoly: MercadoLibre owns commerce, and Nu owns banking. The market has priced Nu at ~25x forward earnings based on the idea of a cordial duopolistic landscape. This assumes that Nu will frictionlessly inherit the financial lives of the region’s youth for the next decades to come.

This narrative may be getting tested. This doesn’t necessarily mean doom and gloom for Nu. It just requires potentially a slight change in expectations.

By analyzing the latest Q3 2025 filings it appears to reveal that Nu may have already entered a three-way war. While Nu posts record headline metrics ($783M GAAP Net Income, 127M customers), a silent, asymmetric threat has emerged from the East: *Enter Sea Limited*.

Through its fintech arm (Monee), Sea is not trying to build a better bank. It is building a pre-bank. They are strategically capturing Gen-Z users via gaming (Garena) and commerce (Shopee) years before they enter the formal banking system. Sea may be on the verge of eroding Nu’s future acquisition funnel or at least hurting it. I will attempt to strip away the accounting noise and quantify why Nu is forced to trade yield for safety to survive.

The Business Truth: The Pivot to Utility

The Illusion: The market sees Nu growing its credit portfolio at 42% FX-neutral and assumes the business model is unchanged.

The Structural Difference: Nu is aggressively pivoting from a high-yield Consumer Lender to a lower-yield Digital Utility.

The Yield Dilution Mechanism

Nu’s Net Interest Margin compressed by 40 basis points to 17.3% this quarter. This is by design. I already touched upon this a few times on Twitter/X. It is a feature of their new defensive strategy.

• The Shift: Nu’s Secured Lending portfolio exploded by 133% YoY, vastly outpacing the 29% growth in high-yield Credit Cards.

• The Trade-off: Secured loans are safe (deducted directly from paychecks) but offer lower ~30% yields compared to the 200%+ APR of revolving credit cards.

• Why? In a three-way war, capacity (access to the paycheck) becomes more valuable than intent (financing the purchase). Nu is diluting its yield today to secure the regulatory moat that Monee cannot easily cross.

The Financial Reality: Stripping the Adjusted Noise

Management highlights an Adjusted Net Income of $829M. I will adjust this metric largely because it adds back Stock-Based Compensation (SBC).

• Adjusted Net Income: $829M

• (-) SBC Expense: ($88.3M)

• (-) Maintenance CapEx: (~$60M)

• (=) Owner Earnings: ~$680M

The Reality: Nu is trading at ~25x GAAP Annualized Earnings, but is actually trading at ~30x Annualized Owner Earnings. The valuation is a little tighter than it looks. At current levels it is definitely not a complete steal unless you assume continued success and more tailwinds (which I think are likely).

My Evidence of the Three-way War

The Pre-Bank Arbitrage

Monee’s global loan book has reached $7.9 billion, growing at ~70% YoY. While not that big in Brazil yet, it proves they have the balance sheet velocity to continue. Based on Shopee’s current success in Brazil it is imminent that Monee will likely succeed as well. In the latest quarter Monee’s loan book in Brazil specifically grew 3x YoY!

The structural threat is friction.

• Nu has higher friction: While Nu does offer accounts for under-18s, it requires a parent to invite the teen into the banking system. It involves KYC, parental consent, and app downloads. It is a utility.

• Sea has very little friction: Monee is able to create a Loyalty Moat before the Nu Regulatory Moat. A 14-year-old cannot legally hold their own credit card, but they can be addicted to Shopee’s gamified coins and Free Fire’s ecosystem. Sea is underwriting the user’s identity years before Nu can underwrite their risk.

Sea is able to engrain itself into somebody’s life as early as the age of 10 with its gaming segment. This can then naturally act as a funnel to consumption on Shopee potentially by the age of 14. By the time that person turns 18, their digital wallet habits could already be hard-wired differently than the previous generation.

The NPL Mirage & The Mix-Shift Mask

Bears point to Monee’s 1.1% NPL ratio vs. Nu’s 6.8% as a sign of superior underwriting, but I would counter that Nu’s NPLs have improved YoY (down from 7.2% in Q3’24). There is more nuance for the big spread and also the improvement for Nu.

Nu’s headline NPL improvement is a actually a bit illusionary because of the Mix Shift.

• The Strategy: Nu has flooded the denominator with Secured Payroll Loans (+133% YoY). These loans have near-zero default rates because they are deducted from salaries.

• The Issue: If you strip out the Secured portfolio and isolate the unsecured Credit Card vintage, I estimate the Real NPL is trending closer to ~8-9%.

• My Thoughts: Nu is not fixing its credit quality completely. It is diluting it with safe, low-yield assets to hide the deterioration of the core card business. Generally this is ok but slightly misleading and should be monitored closely.

Monee’s NPL Mirage

• The 1.1% NPL looks much better, but their duration is likely much shorter (ecommerce BNPL turnover is weeks, not months). Short duration loan books always look cleaner during high growth because the denominator grows faster than the numerator.

The Risks that are likely not Priced-in yet for Nu

1. The Off-Platform Data Bleed & The NuCel Data Bridge

The integration of ShopeePay into Brazil’s Pix ecosystem creates a blind spot for Nu. When a user pays via Pix, Nu loses the merchant category code. They see “Transfer” over “McDonalds.”

The Counter-Move: *Enter NuCel*. Nu launched its MVNO in late 2024. I already talked about this a bit on Twitter but I don’t think Nu’s intent is about building an MVNO for the $10/month subscription revenue. There is some cross-selling and ecosystem building but the real goal in my eyes is that they are building it for the Android permission layer.

• The Blind Spot: As a bank, Nu cannot know exactly what a user is doing on competitor apps.

• The Data Bridge: By bundling NuCel with the Nu App (“Manage your plan here”), Nu creates a valid incentive to request Usage Access permissions.

• The Alpha: This permission bypasses the need to inspect encrypted network traffic. Instead, it potentially opens the door to requesting Usage Access permissions to see exactly how many minutes a user spends on Shopee vs. MercadoLibre. It captures intent and attention, if not the specific transaction. This gives Nu a needed edge against its commerce native competitors.

2. Mexico Regulatory Issues

Nu has 13 million customers in Mexico, but the unit economics are facing a potential headwind.

• The Threat: COFECE (Mexico’s antitrust body) has issued a preliminary finding that the card payment market lacks effective competition. They have recommended that Banxico (The Central Bank) intervene to cap interchange fees.

• The Impact: Mexico’s interchange fees average ~1.36%, nearly 6x higher than Europe’s 0.20%. Nu’s High Yield Deposit strategy (paying nearly 15% on savings) is partially subsidized by these large interchange fees. If Banxico slashes credit card fees to the proposed 0.60%, Nu’s Loss Leader acquisition model in Mexico may suffer more headwinds.

3. Capital Allocation: Defensive Reinvestment

The final proof of the three-way war is in the cash flow statement. Despite generating record net income, Nu management has notably avoided share buybacks.

• The Signal: This is Defensive Reinvestment. Management is hoarding cash to fund capital-intensive projects like the NuCel infrastructure build-out and the high cost of Mexican deposits ($0.15 of every $1.00 acquired).

• The Implication: They do not view this current period as victory lap moment but as an existential land-grab. They are fully focused on securing their utility status in the LatAm digital economy. And to be clear I think this is definitely the right move!

My Final Thoughts

The age of uncontested growth is likely coming to an end but that is not all doom and gloom for Nu. We may be seeing a period of a defensive war against an asymmetric rival (Sea) and yield dilution in an exchange for safety.

• The Bull Case: Nu builds a Regulatory Fortress via Secured Lending and maintaining a $0.90 Cost to Serve that incumbents cannot touch. This gives them ownership of the customer’s salary!

• The Bear Case: Monee successfully executes the Gen-Z funnel, creating a big credit headwind where future cohorts have structurally lower lifetime value.

My Conviction: Nu remains attractive for its operational excellence and Brazilian dominance, but there is definitely competition (and competition is always costly). If they did not have such great management I would probably be more worried about the risks. It just so happens to be that the same company causing margin compression for MercadoLibre may also start creating similar problems for Nu.

The PEG Trap: Investors paying ~25x earnings for 30% growth may feel extra safe. But if Sea erodes the terminal value of the 2028+ cohorts, the growth may slow down a little more than expected.

Keep in-mind its easy to look at competition and say that the stock is overvalued. I think this is not the case. There are several tailwinds for Nu that are not being accounted for either. Colombia will likely be another growth engine. Any kind of United States adoption would be an incredible unlock. These tailwinds could be huge but my job as an investor is to understand the risks. I refuse to own something without fully understanding the downside.

I dug into the Q3 2025 filings, and honestly, the market is reading this wrong. What looks like cash burn on the P&L is actually a massive capital allocation play. They aren’t just buying growth but front-loading their liability moat.

Subscribe now

Here is the one chart that might change your mind.

The Liability-First Playbook

Most fintechs enter a new country and lend aggressively to show growth (Assets). Then, they scramble to find the money to fund those loans (Liabilities), usually relying on expensive wholesale debt. It’s risky and kills margins.

Nu is doing the exact opposite in Mexico.

They are using their massive Brazilian profits, $783 million in Net Income just this quarter, to subsidize high-yield savings accounts in Mexico first.

Look at the numbers:

• The Fact: In Q2 2025, Nu Mexico explicitly reported $6.7 billion in deposits.

• The Trend: By Q3, the customer base swelled to 13.1 million. Based on that trajectory, the deposit fortress is now tracking toward ~$7.5 billion.

They now have a massive surplus of cash (Liabilities) relative to their loan book (Assets).

Addressing the Bear Case: Is This Just Hot Money?

This is the smartest critique I hear: “They are paying ~14% interest to have those deposits. That’s not a bank. That’s a yield farm. The moment rates drop, that money flees.”

It is a valid risk. High-yield deposits are expensive CAC disguised as interest expense.

But here is why I think the Hot Money thesis is wrong:

1. The Primary Shift: Nu isn’t just capturing savings but capturing the operating account. In Brazil, Nu became the Primary Banking Relationship for over 60% of active users. Management has confirmed Mexico is following the exact same S-Curve adoption, just faster.

2. The Usage Data: If these were just yield tourists, you would see high deposits but low activity. Instead, Nu reported a global Activity Rate of 84%. People aren’t just parking cash. They are logging in and transacting.

3. The Trap: Once a customer sets up bill pay, direct deposit, and gets a credit card limit (which Nu is now ramping up), the switching cost becomes incredibly high. Nu is paying 14% now to lock in a customer who will generate 30%+ ROE for the next decade.

Why This is Alpha (and not just accounting)

You might look at Nu’s consolidated margins and see them compressed by the high interest they pay to Mexican savers. Short-term traders hate this.

But here is why long-term investors should love it:

1. De-Risked Growth: Nu doesn’t need to ask the capital markets for permission to lend in Mexico. They already have the cash. That ~$7B fortress means they control their own destiny.

2. Structural Advantage: Deposit funding is structurally superior to wholesale debt. By securing the liability side now, they are guaranteeing lower funding costs and higher margins when they eventually turn on the lending tap in 2026.

The Bottom Line

Nu just posted a record $4.2 billion in revenue this quarter. The burn in Mexico is barely a scratch on their Brazil cash printer.

Management is engaging in Scale Economies Shared: intentionally deferring monetization to widen the moat. They are paying for a low-cost funding base today so they can be the low-cost lender tomorrow.

Notes:

• Mexico Deposits: $6.7B figure verified in Q2 2025 Earnings Release. Q3 figure (~$7.3B) is my conservative estimate based on verified Q3 customer growth (13.1 million).

• Activity Rate: 84% global activity rate verified in Q3 2025 Release.

The market is getting MercadoLibre’s Q3 wrong. Everyone is fixating on the operating margin dropping to 9.8% and freaking out. It’s a classic knee-jerk reaction from Wall Street observers who are conditioned to favor short-term profits over long-term dominance.

If you actually dig into the filings, the shareholder letter, and CFO Martin de los Santos’s comments, you’ll see that this margin compression isn’t because MELI is getting sloppy. It’s not operational inefficiency. It is the deliberate price tag of building the biggest competitive moat in the history of emerging markets.

The real story this quarter isn’t just the 39% revenue growth to $7.4 billion (though that is huge). The real story is the engineering behind it. MELI is playing a game of internal arbitrage. They are using their commerce dominance to get credit card users for free. They are using empty space in trucks to ship cheap items for free. They are using fintech to fix a broken cash economy.

I believe MELI has moved past the growth phase and entered a capacity arbitrage phase. They are using a surplus in one area (user data, empty trucks on Fridays) to subsidize another (credit underwriting, shipping costs). Banks can’t do this. Retailers can’t do this. Only MELI can.

Here is what is actually happening.

Subscribe now

The Latin American Reality (Why This Works)

To understand MELI, you have to understand that Latin America in 2025 still has frictions that simply don’t exist in the US or Europe.

• The Banking Gap: Traditional banks in LatAm are expensive dinosaurs with high fees and physical branches. Neobanks like Nubank are great, but they lack the data that MELI has. They don’t know if a customer is a good credit risk because they don’t see their exact spending habits. MELI does.

• The Logistics Void: There is no USPS equivalent that is reliable. If you wanted to ship something fast and safe five years ago, good luck. MELI had to build their own network (Mercado Envios) out of necessity. Now, they ship ~95% of their own items.

While competitors like Amazon and Shopee are present, they are fighting with one hand tied behind their backs. Shopee still relies on third-party couriers (so they can’t optimize costs nearly as much as MELI), and Amazon lacks the deep fintech integration to lend money to the unbanked.

The Zero-CAC Flywheel: Killing the Banks

The most misunderstood number in Q3 is the credit card portfolio growing 83% to ~$11 billion. The growth was also very efficient.

Traditional banks burn money on ads, branches, and sales commissions to get you to sign up for a card. Their customer acquisition cost is high. MELI has driven that cost to near zero.

Let’s walk through a thought experiment. Imagine a user buying a refrigerator on MercadoLibre. They have already been pre-approved by MELI’s risk models. At the checkout, the exact moment they want to buy, MELI offers them a credit card to pay for it. One click. Done.

• Marketing spend? Zero. The user was already there.

• Friction? Zero. No visiting a branch.

• Intent? Validated. They aren’t asking for cash to blow it, they are asking for credit to buy a fridge (an essential).

This is how they issued a record number of new cards in Brazil this quarter, building on the 1.5 million added in Q2.

But what about the margins? The financials look ugly right now because of accounting rules (CECL). When you issue a ton of new cards, you have to set aside money now for potential bad debts later. This crushes your margins (NIMAL) in the short term. MELI’s NIMAL dropped to 21%, and the market panicked. But that panic ignores the J-curve. Management revealed a critical stat where approximately half of the Brazil credit card portfolio is already profitable. They aren’t burning cash on a gamble but are waiting for the newer half of the book to mature. They are spending money today to acquire customers who are proven to become profitable. They are spending money today to acquire customers who will pay them for years. It’s a land grab, and they are winning.

Logistics: The Friday Gap Magic

If the credit strategy is about arbitraging data, the logistics strategy is about arbitraging physics. MELI cut unit shipping costs in Brazil by 8% QoQ this quarter, even while aggressively subsidizing shipping. They slashed the free shipping threshold in Brazil from R$79 to R$19. This wasn’t a tweak but the opening of a floodgate. It drove a 29% YoY increase in unique buyers in Brazil. This is the fastest pace since 2021. How do you lower costs while shipping millions of cheaper items?

The Friday Gap. Logistics networks are usually slammed Monday through Wednesday fulfilling weekend orders. Trucks are packed and warehouses are buzzing. But by Thursday and Friday? The volume drops! Trucks run half-empty. But the truck costs the same whether it’s 50% full or 100% full.

MELI’s CFO revealed a brilliant hack: They offer free shipping but only if you accept a Slow Shipping window (3-5 days). This lets the algorithm hold the package back.

• If you buy that t-shirt on Monday, they don’t rush it.

• They wait until Friday when the truck has empty space.

• They throw your package on the truck for effectively zero marginal cost.

They are turning waste (empty truck space) into a value proposition (free shipping). Third-party logistics providers can’t do this because they charge per package.

The 1P Strategy: It’s a Plug, Not a Pivot

Investors get nervous when they see MELI selling its own inventory (1P revenue hit $933M accelerating to 69% YoY growth). They fear MELI is becoming a low-margin retailer like Amazon’s core business.

No need to panic here. Small merchants can’t make money shipping heavy fridges or low-margin laundry detergent. So MELI steps in to sell those items directly. Why? Assortment liquidity. MELI is willingly taking a hit on inventory turnover and working capital to ensure the customer never leaves the app. If you come to MELI for a fridge and they don’t have it, you go to a competitor. If you go to a competitor, you might buy other things. By selling the fridge directly (even at lower margins), MELI keeps you in the ecosystem so you buy your high-margin phone case from a 3P seller later.

Mexico: The Closed-Loop Money Machine

Mexico is growing like crazy (+36% GMV). Why? Because MELI built a wall around the money. Mexico is a cash-heavy economy. People actually print vouchers and pay cash at convenience stores (Oxxo) for online goods. It’s slow and friction-heavy.

MELI built a closed loop:

1. You deposit cash into the Mercado Pago wallet.

2. MELI gives you credit based on that data.

3. You spend that money on the Marketplace.

Now, 20-30% of sales in Mexico are paid via Mercado Pago’s own rails. The wall is quickly being built. Without fintech, a third of their sales in Mexico might just evaporate.

Ad-Tech: Following You to Disney+

MELI Ads grew 63%. But the shift is qualitative, not just quantitative. They are moving beyond Retail Media (ads on the search page) to Off-Platform ads. They are now partnering with Disney+ and Roku . MELI can now target you while you are watching a movie on Disney+ with an ad for running shoes, because they know you were looking at Nikes on MercadoLibre yesterday. This unlocks massive brand advertising budgets (think Coca-Cola ) that are much bigger than simple search ads. This moves MELI into direct competition with specialized Demand Side Platforms like The Trade Desk , significantly expanding its Total Addressable Market into the massive pool of brand advertising dollars, which dwarfs performance budgets.

Argentina: The Profit Anomaly

You might look at the headlines about Argentina like hyperinflation, currency crashes, economic chaos and assume it’s a drag on the business. But you’d be wrong. Argentina remains a surprisingly robust profit engine for MercadoLibre.

The launch of the credit card there in Q3 reveals a fascinating strategic twist. It’s completely different from the playbook in Brazil or Mexico.

In Brazil and Mexico, fintech usually starts at the bottom. You build products for the unbanked and people who traditional banks ignore. It’s a volume game with lower-income users.

In Argentina, the dynamic is inverted. Because inflation is so aggressive, everyone needs Mercado Pago. It isn’t just a banking alternative but a survival tool for your money. Even the upper-middle class and the wealthy use the app daily to pay bills, move cash, and earn yield to fight inflation.

Why this matters for the Credit Card? When MELI launched the credit card in Argentina this quarter, they weren’t starting from scratch with risky, unknown borrowers.

• The Target: They immediately tapped into the highest-frequency users in the country. Because of QR payments and wallet usage, MELI has years of data on payment reliability for users that traditional banks ignore.

• The Data: They already have years of data on these high-fidelity users. They know exactly who pays their bills on time and who has deep pockets.

• The Result: Management was clear and they expect the Argentina credit book to hit profitability faster than Brazil or Mexico. Why? Because the customers are high-frequency users, the data is cleaner, and the engagement is higher. In a chaotic economy, MercadoLibre has become the safe harbor, and that pays incredibly well.

But there is a catch that confuses the algorithms. Because Argentina is a hyper-inflationary economy, GAAP accounting forces MELI to take FX losses on local assets directly on the P&L. This quarter, that created a $102 million optical hit to Net Income. The operational machine is printing money while the currency accounting is hiding it.

The P&L and The Risks

The administrative costs (Headquarters, R&D) are actually dropping as a percentage of revenue. The company is getting more efficient. The costs that are rising are the variable ones: Bad debt provisions (because they are growing credit) and shipping subsidies (to lock in customers). They generated $206 million in free cash flow despite investing massively. That is a fortress balance sheet.

However, the risks are real:

1. Credit Spirals: Vintage Risk. If Brazil’s economy tanks, all those new credit cards could be useless. If defaults rise higher than their models predict, the profits may vanish.

2. Competition: Amazon is rich and patient. If they decide to burn cash to win market share in Brazil, it could force MELI to spend even more.

3. Argentina: It’s still 19% of revenue. One bad currency devaluation or political crisis there could hurt the reported numbers significantly.

The Verdict

You can look at this quarter two ways.

• The Bear Case: Margins are falling, bad debt is rising, and they are buying growth.

• The Bull Case: MELI has cracked the code on zero-cost customer acquisition. They are spending margin today to acquire users who will be profitable for a decade. The margin compression isn’t waste but it’s the capital expenditure of building a digital nation-state.

I’m personally in the camp with Bulls. The moat is widening, even if the construction costs look high on the quarterly print.