It’s 20% undervalued, and there is a potential triple in 5 years. This makes it a buy at $255.

This post completes our undervalued Mag5 series with META, MSFT, and GOOGL.

BEFORE YOU READ ON:

I’m planning to be more active on X. Would appreciate it a lot if you follow me there first before continuing.

Here is the link to my profile.

If I get 100 followers, the META post will become free. Thank you!

Here is a mental model for you to understand Amazon:

They report revenue as 7 segments and operating income as 3 segments, but they can be grouped into 3 businesses: retail, cloud, and ads.

Revenue for each segment: 70% retail, 20% cloud, 10% ads. But the profits come from two sources: 60% cloud, 40% ads, and 0% retail. Ads revenue is broken out while its profits are reported inside the retail business, so this is an estimate.
Growth is around 10% for retail and 20% for cloud and ads, while all of them accelerated in the last quarter.

Now, let’s zoom out and answer a few important questions.

Try for 7 days with zero risk:

• Posts: Every idea with its quality, growth, and valuation.

• The full 20-name portfolio and 100-name watchlist.

• Trade Alerts.

Stay longer for $10/month or $120/year.

Subscribe now

Quality: Is this a compounder?

Absolutely.

Price and Performance

AMZN grew OCF at 28% and returned 25% annually since 2006. I’ve seen so many charts like these, and I’m 100% convinced that price follows business performance in the long term.

Now, let’s see what a committee of about 20 experienced Morningstar equity analysts have to say about its moat.

Moat and Management

Amazon is one of the few businesses where each business—retail, AWS, and advertising—has a wide moat (this is not AI, as I finally found a use case for the long dash). As a result, it earns 4 out of 5 moats that Morningstar considers when evaluating a company’s moat. The fifth moat type, efficient scale, simply doesn’t apply because it is for companies that service a market of limited size.

Here is what they are saying about it:

The retail biz, a virtuous circle:

This segment enjoys 30% to 60% operating margin, depending on who you ask:

About 60% of the operating profits come from the cloud biz, while it brings in only about 20% of the revenues:

Untouchable business as a whole:

Also, they assign an exemplary rating for their management. We are definitely in good hands.

A wide moat is a strong defense against rivals that is expected to last 20 years or more, while a narrow moat is for at least 10 years. I trust Morningstar’s moat ratings more than mine because a committee of about 20 experienced analysts beats my analysis for sure.

I do trust Morningstar’s moat ratings because a committee of about 20 experienced analysts beats my analysis for sure, but I see it as a directional instead of an absolute rating. Once you stack a lot of them in your portfolio, you’re likely to end up with a higher quality portfolio.

Moat ratings are forward-looking, but they can serve as a historical quality rating because companies don’t get them in one day.

Read about the Morningstar rating system here.

Growth: Is it likely to continue?

Yes.

What does the analyst consensus say?

Analysts, about 50+ of them, think that AMZN will grow revenue 14% and CFO 30% in the next 5 years.

Can they model this business?

Yes, especially over the last ten years. Analysts can model revenue fairly accurately, with +/-2% surprises. I checked, and the 22% quarterly miss in October 2004 is bad data. The actual miss was less than 1%.

There is no surprise chart for OCF, so I’m using EBITDA as a proxy for it. I see about 10% positive surprise on average, potentially less if you make it annual. Another thing is negative surprises almost disappeared after 2014.

Let’s take a look at the fair value and potential return in the next section.

Valuation

What the Market Has Paid

AMZN has been trading around 23.5 P/OCFps while growing at 28% in the last 20 years. Currently it trades around 16.5x.

Fastgraphs uses a weird version of multiples; they call it the blended. It is calculated by averaging the last 6 months with the next 6 months. So, it differs from the widely used LTM or NTM multiples.

It seems like the market re-rated it to around 20x since 2022, as it’s been trading in between 16.5 and 23.5x. It could be due to the revenue slowing down from about 20% growth to 10% since 2023. It might rerate, as it’s been picking up in the last 2 quarters with above 15% growth.

If you are comfortable using the LTM P/OCFps, here it is from Koyfin with 26x in the last 20 years and 17x currently. There isn’t a huge difference between FASTGraphs and Koyfin.

What You Actually Make

It is projected to grow around OCF 30% in the next 5 years. Fastgraphs shows FactSet data, while Koyfin shows S&P Capital IQ. So, they differ slightly.

The bear case is just using the current blended multiple of 16.5x because AMZN hasn’t been trading below that line consistently as it dips and recovers. The base case is 20x, which is the middle of the current 16.5x and the historical 23.5x. And the bull case is returning to the historical 23.5x, which might be in the cards due to the revenue acceleration towards 20%.

I used the 19.8 NTM OCF per share to calculate the fair value from today’s price of $255.

I’m showing the returns until 2028 and 2031, as one might not want to hold for 5 years.

Showing returns until 2028:

Showing returns until 2031:

Even using the current multiple shows that it’s undervalued by 22%, or in other words, your margin of safety. It’s a potential double in 3 years, even at the current multiple. If the growth and the current multiple hold up, it’s a potential triple from $255. I think there is nothing complicated here for the long-term investors. Buy and hold.

Please restack and like if you found this useful so other investors discover it too. Thank you!

It’s 13% undervalued conservatively, and the potential return is 20% annual in the next 3-5 years, even in the bear case with a 65% upside in the short term if it rerates to its historical multiple.

I have META in the compounders portfolio along with GOOGL, AMZN, and MSFT. It’s the fastest growing among the Mag5 and the cheapest.

Meta keeps you and 4 billion other monthly active users looking at your screen through their Facebook, Instagram, WhatsApp, and Messenger apps. It makes money by showing you ads. It benefits from more dollars going into digital advertising and improving average revenue per user as the user growth is slow.

Meta agreed to pay $17 billion in fines over 10 years and make safety changes for under-18 users, including a two-hour daily cap across Facebook and Instagram. No access from midnight to 6am. Notifications blocked during school hours. Forced pauses after 15, 60 and 90 minutes of scrolling. These run five years or drop to a one-hour daily cap for ten years if TikTok, YouTube, and Snapchat agree to the same terms. Meta also accepted an independent auditor who reports to the attorneys general.

As per Morningstar, these changes are a nothingburger because teens don’t matter for the top line:

Now, let’s zoom out and answer a few important questions.

Upgrade for $10/month or $120/year:

• Posts: Every idea with its quality, growth, and valuation.

• The full 20-name portfolio and 100-name watchlist.

• Trade Alerts.

Quality: Is this a compounder?

Yes, not a clean one.

Price and Performance

Meta grew its EPS at a 32% CAGR while it returned closer to 21%. Clearly a case of multiple contraction that we will see in the valuation section.

Meta is currently in another drawdown of another -30%, which isn’t as bad as the 75% drawdown in 2023. So, it isn’t a super clean compounder like Mastercard.

Now, let’s see what a committee of about 20 experienced Morningstar equity analysts have to say about its moat.

Moat and Management

I do trust Morningstar’s moat ratings more than mine because a committee of about 20 experienced analysts beats my analysis for sure, but I see it as a directional instead of an absolute rating.

Moat ratings are forward-looking, but they can serve as a historical quality rating because companies don't get them in one day.

Morningstar assigns Meta a wide moat, meaning they expect their competitive advantage to last 20 years or longer:

Also, they assign a standard rating for their management (poor, standard, or exemplary), saying their cash position is solid and internal R&D is good, while Reality Labs is a sinkhole, their M&A track record is good, and finally, Zuck’s passion projects “zuck.”

Read about the Morningstar rating system here.

Growth: Is it likely to continue?

Yes.

What does the analyst consensus say?

Analysts, about 50+ of them, think that META will grow revenue 18% and OCF 21% in the next 5 years. Estimates are from S&P Capital IQ through Koyfin.

The 5-year EPS growth estimate is 12% due to depreciation, and FCF is a mess due to the ongoing AI CAPEX.

Can they model this business?

Yes. Analysts are able to model this business pretty accurately because the average miss for revenue is less than 1%, while it’s about 5% for EPS. And for some reason, META keeps surprising the analysts with EPS beats. Underpromise and overdeliver?

In short, meaningful beats on EPS, with 10-20% being the norm.

Let’s take a look at the fair value and potential returns in the next section.

Valuation

Next, I’ll explain why there is a case for a potential 65% return in the short term.

What the Market Has Paid

The historical chart for META’s EPS is pretty clean, but the future projections are affected by depreciation. So, I’m using OCF to value the company’s growth.

META has been trading at a 0.6 PEG with 33% OCF growth while trading at 21 P/OCF in the last 14 years. Currently it trades around 11.5x (black straight line).

I’ll use 21x as the bull case, as it used to trade at this valuation before 2016 and then touched this valuation during 2019-2020 (price chart touching the blue line).

Fastgraphs uses a weird version of multiples; they call it the blended multiple. It is calculated by averaging the last 6 months with the next 6 months.

If we use the more familiar NTM OCF, then it trades around 9.8 P/OCF (570/58).

The average is higher due to the higher multiple in the first 4 years after the IPO as it grew 51% annually while trading at 34 P/OCF. Since it trades at 11.5x today, that’s a huge multiple contraction. Lesson? Initial valuation doesn’t matter if a company grows super fast for a decade, but it definitely increases your odds of coming on top if you start out with a lower multiple.

Then it rerated to 16.5x as growth slowed down to 25% with 0.7 PEG. When the market loves Meta, it tends to trade at that multiple (blue line below). So, I’ll use it as the base case.

And for the bear case, I’ll just use the current 11.5x multiple, meaning no multiple expansion from here.

Now, let’s see what are the potential returns and the fair value.

What You Actually Make

Meta is projected to grow around 20% in the next 5 years. Fastgraphs shows FactSet data, while Koyfin shows S&P Capital IQ. So, they differ slightly.

As mentioned above, the bear is 11.5x, the base is 16.5x, and the bull is 21x.

I’m showing the returns until 2028 and 2031, as one might not want to hold for 5 years.

61% total or 22% annual returns minimum by 2028:

166% total or 20% annual returns by 2031:

I used the 58 NTM OCF per share to calculate the fair value. Even using the current multiple shows that it’s undervalued by 13%.

So, if Meta rerates to 16.5x, which wouldn’t be unexpected, we might see a 65% return just from multiple expansion in the short term. And in the long term, say 5 years, the bear case is showing 166% total return. This is an easy buy here.

If you found this useful, make sure to restack it so other investors see it too. Thank you.

Since I want to keep the portfolio capped at 20 names, I’ve made room for GOOGL by selling RELX. I didn’t have GOOGL when I built this portfolio due to its valuation. Now, it has come back to earth because they haven’t released a new model. Whatever.

RELX has a wide moat and exemplary management ratings just like GOOGL, but it’s the slowest-growing company in the portfolio in terms of top line with 6%.

Their consensus EPS growth for the next 3 years is 11%; combine that with a 2.65% dividend yield, and you get about a 14% return. Also, they have bought about 18% of their shares in the last 12 years.

If you already own GOOGL, you might want to take a look at it.

It’s an interesting business. It builds large collections of data that are hard to assemble (court rulings, insurance claims histories, scientific papers, and identity records), wraps software around them, and rents access by annual subscription. Roughly 54% of revenue is subscription, and 58% comes from North America despite the London listing. It is a British-listed company that is mostly an American data business.

Morningstar has this to say about their management:

Here is my GOOGL post if you want to take a look at its valuation:

Portfolio and watchlist for more companies like RELX:

Please like and restack if you find it useful. Comment to start a discussion.

Upgrade for $10/month or $120/year:

• Posts: Every idea with its quality, growth, and valuation.

• The full 20-name portfolio and 100-name watchlist.

• Trade Alerts.

And it’s not an AI name. I haven’t seen it covered on Substack either. Potential return in the next 5 years is double digits.

Upgrade for $10/month or $120/year:

• Posts: Every idea with its quality, growth, and valuation.

• The full 20-name portfolio and 250-name watchlist.

• Idea Generator: The 10,000-stock dataset to find your own Compounders.

Read more

It might eventually go through a drawdown again in the future depending on how AI CAPEX turns out, as a lot is riding on OpenAI and the ROI of their investment in the short term. But in the long run, say, more than 5 years, none of this should matter. They might’ve as well burned that cash literally and still be ok because these businesses are super dominant. Ask META with their metacrap.

Here is a mental model for Microsoft so you have a better understanding of the business: 50-30-15.

Think of Microsoft as having three engines. About half of the business is Azure and cloud services, which power businesses behind the scenes. About one third comes from Microsoft Office, including Microsoft 365, Teams, Outlook, and LinkedIn. The remaining about one fifth comes from Windows and Xbox, including Windows licenses, gaming, Surface devices, Bing, and Edge.

Now, let’s go through each step in detail.

Like and restack before you read on. It means a lot to me! Thank you!

Upgrade for $10/month or $120/year:

• Posts: Every idea with its quality, growth, and valuation.

• The full 20-name portfolio and 250-name watchlist.

• Idea Generator: The 10,000-stock dataset to find your own Compounders.

Quality: Is this a compounder?

Absolutely.

Price and Performance

MSFT really took off after 2015 with Nadella as the CEO as it started growing closer to 20%. Look at the growth rates at the bottom of the chart, basically flat coming into 2015. Basically a dog of the market.

And the return reflected that with a 7.4% annual return for 8.4% EPS growth from 2006 to 2015. This is clear evidence that price follows the business performance.

Now look at the return since 2015. The company’s EPS increased by 18% annually while returning 27% due to the multiple expansion.

Now, let’s see what a committee of about 20 experienced Morningstar equity analysts have to say about its moat.

Moat and Management

Everyone knows that Microsoft is one of the world’s most dominant businesses, thanks to Azure, Office, and Windows.

I trust Morningstar’s moat ratings more than mine because a committee of about 20 experienced analysts beats my analysis for sure.

Here is what Morningstar has to say about the moat:

Also, they assign an exemplary rating for their management (poor, standard, or exemplary). We are definitely in good hands.

Read about the Morningstar rating system here.

Growth: Is it likely to continue?

Yes.

What does the analyst consensus say?

Analysts, about 40+ of them, think that MSFT will grow revenue 17% and EPS 19% in the next 5 years. It’s kind of ridiculous to grow close to 20% at this scale.

Can they model this business?

Yes. Analysts are able to model this business pretty accurately because the average surprise for revenue is 1-2%, while it’s about 10% for EPS. And for some reason, MSFT keeps surprising the analysts with EPS beats. Underpromise and overdeliver?

Let’s take a look at the fair value and potential return in the next section.

Valuation: Is it a buy today?

Yes.

What the Market Has Paid

EPS went from $1.49 in FY07 to $17.28 in FY26. Two down years, as FY09 was -12% and FY13 was -7%. Every year was up and the market handed the stock a 22x average multiple.

That 1.58 PEG is the number that the buyers have been willing to pay.

Today the blended multiple is 26.6x.

Now, let’s see what we can make based on what we assume.

What You Actually Make

Consensus has Microsoft growing earnings 17% a year for the next 5-6 years. This is from FactSet research, while S&P Capital IQ has it at 19%. I’ll be using the FS as that’s the default on FASTGraphs and is more conservative of the two.

Using the 1.58 PEG the market has paid, then flexing 20% each way:

Here are the returns. Two hold periods, the shorter one first.

Hold to June 30, 2028:

Why two hold periods? Because the shorter you hold, the more your outcome depends on the exit multiple and the less it depends on the business. Two years out, a multiple that goes from 26x to 17x costs you money even though earnings grow 17% a year. Stretch the hold, and earnings do the heavy lifting while the multiple noise washes out.

Hold to June 30, 2032:

Look at what six years does to the bear case. Losing 6% a year over two years becomes making 10% a year over six. Same multiple collapses but completely different outcome.

Now the fair value math on NTM EPS of $19.18:

The MSFT fiscal calendar ends in June. So, the NTM EPS is basically for the year 2027.

At $464 the stock is trading around a conservative fair value estimate. Nothing crazy. If you buy here and the multiple holds at 22x, the six-year math gives you about 14% a year. If it compresses to 18x, you still get 10%.

Do you own it? Why not? Comment below.

Subscribe now

Think of Alphabet as three engines: 55-25-20

About 55% is Search. Ads on Google’s own results pages, plus the smaller “other” properties. This is the core that pays for everything else.

About 25% is the rest of the consumer business. YouTube ads, subscriptions (YouTube Premium, YouTube TV, Google One), Android and Play, Pixel devices, and ads placed on third-party sites through the Network.

About 20% is Google Cloud. Infrastructure, Workspace, and now TPU systems sold to enterprises.

Other Bets, Waymo mainly, is 0.3% of revenue.

These words from the Morningstar team describe it very well:

Now, let’s go through each step in detail.

Like and restack before you read on. It means a lot to me! Thank you!

Upgrade for $10/month or $120/year:

• Posts: Every idea with its quality, growth, and valuation.

• The full 20-name portfolio and 250-name watchlist.

• Idea Generator: The 10,000-stock dataset to find your own Compounders.

Quality: Is this a compounder?

Absolutely.

Price and Performance

Look at this chart. It is beautiful. 27% annual EPS growth since 2006 with a single down year in 2022 with -19% EPS drop.

The company’s EPS increased by 27% annually while returning 18% due to the multiple contraction as Google traded at 50-80 P/E after it’s IPO and before the GFC.

If a company continues to grow, valuation is irrelevant; however, since the future is uncertain, you should demand a margin of safety with a lower multiple.

Now, let’s see what Morningstar has to say about its moat.

Moat and Management

I trust Morningstar’s moat ratings more than mine because a committee of about 20 experienced analysts beats my analysis for sure.

A wide moat means their competitive advantage will last 20 years or more, while narrow maot is at least 10 years.

Here is what Morningstar has to say about their moat:

And they assign no moat to their hardware portfolio, which includes the Pixel, Nest, Chromecast, and Fitbit, due to the lack of any material disclosure. The same for other bets that include Waymo, which burns capital.

Also, they assign an exemplary rating for their management (poor, standard, or exemplary). We are definitely in good hands.

Read about the Morningstar rating system here.

Growth: Is it likely to continue?

Yes.

What does the analyst consensus say?

Analysts, about 50+ of them, think that GOOGL will grow revenue 18% and EPS 17% in the next 5 years. It’s kind of ridiculous at this scale.

Can they model this business?

Yes. Analysts are able to model this business pretty accurately because the average surprise for revenue is 2-4%, while it’s about 4-5% for EPS, not counting the 212% surprise due to investment gains, as GOOGL only reports GAAP.

Let’s take a look at the fair value and potential return in the next section.

Valuation: Is it a buy today?

Yes

What the Market Has Paid

I’ll be using OCF instead of EPS or FCF due to equity gains, depreciation, and ongoing AI CAPEX.

Two clean windows, and I’m reading the multiple and the growth rate off the same timeframe each time so the ratio actually means something.

Google grew cash flow at 18%, and the market paid 18xish multiple for it from 2010 to 2019.

Over the full history cash flow compounded at 23% and the market paid 20x.

Notice what that means. Faster growth, roughly the same PEG. The market has never fallen in love with this stock the way it fell in love with Costco. Google has spent twenty years being priced at slightly less than one PEG.

Today the blended multiple is 21x. On next year’s cash flow though it’s about 15x, and on 2028 it’s 12x.

Now let’s see what we can make depending on what multiple we assume.

What You Actually Make

Forward OCF growth comes out at around 21%, and Google has actually done better than that historically (23% over the full run).

The PEG set from those two windows:

Multiply each by the 21.32% forward growth rate:

0.87 x 21.32 = 18.5x. Call it 18x. That’s the bear. It’s also almost exactly what the market paid through the entire 2010s, so this isn’t a panic scenario. This is Google being priced normally.

0.95 x 21.32 = 20.3x. Call it 20x. Base case, and it happens to be dead on the twenty-year average.

22x is the bull, which works out to a PEG of 1.03. This is me pulling a 20% premium out of my rear for GOOGL relative to its 20-year multiple.

Two hold periods because some of us might not want to hold for 5 years.

To end of 2028:

To end of 2031:

Fair value today, on NTM OCF of $19.29:

The stock is below fair value on all three cases. Not by a mile on the bear case. By enough on the base.

My Take

Undervalued.

At $319 the base case says roughly 20% a year for five years. The bear case says 18%. Even if the multiple compresses to what the market paid through the whole 2010s you still make money, which is a rare setup for a $3.9 trillion company.

For this to go badly the cash flow estimates have to break. To lose money holding to 2028 you’d need the exit multiple down at 12x, and Google has not traded there in twenty years.

What are you doing with Google here? Buying the capex fear or waiting it out?

Netflix‘s quality score of 44 surprised me. This is one of the best growth stories of the last twenty years, and it screens as a C Tier.

Quality Score is built for analyzing compounders, more mature companies. It did its job by telling us that Netflix was a growth story that is now maturing. If you’re a strict quality-first, consistency-first buyer, Netflix has always looked “not quite a compounder,” and this 44 explains why in numbers.

It’s fairly valued. See the valuation section below for potential returns.

Read on if you want to know why it scored 44 and the assumptions behind the fair value calculations and returns.

Like and restack before you read on. It means a lot to me! Thank you!

Upgrade for $120/year:

• Every paid post: Quality Scores, Valuations, and Trades.

• The full 20-name portfolio and 250-name watchlist.

• The 10,000-stock dataset to find your own Compounders.

  Subscribe now

Step 1. Historical Returns: how has the stock done, and how steadily?

Great returns. Ugly line.

The bar: I score six timeframes (3, 5, 7, 10, 15, 20 years). A timeframe passes only if the stock compounded at 10% or better AND the price line was smooth, measured by R-squared of 0.85 or higher. R-squared is just how close the price hugs a straight trend. High R-squared means a calm, steady climb. Low means a rollercoaster. Ten points if all six pass, and it steps down from there.

Look at that footer. The CAGR row is almost all green. 17% over 3 years, 23% over 10, 31% over 20. The compounding is there. It’s the R-squared row that guts the score. Only the 15Y (0.86) and 20Y (0.93) windows are smooth enough to clear the bar. The 3, 5, 7 and 10 year windows all fail on linearity even though the returns are fine.

Why? Because the last five years have been a demolition derby. The stock ran to a blow-off top in late 2021, then cratered in Q1 2022 when Netflix reported its first subscriber decline in a decade and guided to more losses. It then tripled off the lows, ran up through 2025, and has been sliding again into 2026. Down ~44% over the past year, ~24% YTD, and another ~9% leg down after the Q2 ‘26 print on July 16. Fantastic if you timed it. A mess if you’re measuring smoothness. So 2 of 6 pass. That’s a 2 on Step 1. And this is a price problem, not a business problem. Hold that thought; it comes up again.

Step 2. Historical Growth: is the business actually growing?

Yes. And it’s not close.

The bar: I check revenue, diluted EPS and free cash flow across all six timeframes. Revenue needs a 7% CAGR to pass. EPS and FCF need 10%. Any window where EPS or FCF started negative gets marked N/A and dropped. Score is how many of the available checks clear.

This is a clean sweep. Revenue compounded 13% over three years and 23% over twenty.

EPS compounded between 32% and 56% depending on the window.

FCF passes every window where it can even be measured. Sixteen of sixteen available checks pass. That’s a 10.

Two windows show N/A on the FCF row. Those are the 7Y and 10Y, where free cash flow started deeply negative (2018 FCF was minus $2.9B, and 2015 was minus $919M), so a growth rate is meaningless. We’ll get to why the cash was negative in the next step. For now, the machine is unambiguously growing. Revenue went from $6.8B ten years ago to $45B last year. Nobody argues with Step 2.

Step 3. FCF / Net Income: is that growth real cash or just accounting?

For a full decade, it wasn’t cash at all. This is the ugliest step in the whole report.

The bar: for each timeframe I take the median FCF-to-net-income ratio and find the highest tier where every window clears it. All windows at 100% or better is a 10. All at 90% is an 8. All at 80% is a 6. Below that, it falls off a cliff fast.

Read the yearly ratios and you’ll see the story. 2016 free cash flow was negative 887% of net income. 2017 was negative 360%. 2018 negative 239%. Netflix paid cash for content years before that expense flowed through the income statement, so it reported accounting profits while free cash flow stayed deeply negative.

So the medians are all over the place. The 3Y median is a healthy 86%. But the 5Y is 79%, the 7Y is 70%, the 10Y is 17%, and the 15Y is actually negative 20%. Only one window out of six clears the 80% bar. Per the rubric that’s a 0.

Here’s the context read though, and it matters. The cash turned. Netflix went FCF positive in 2020 and has stayed there, printing $9.5B of free cash flow in 2025. The 3Y median at 86% is the real Netflix now. The business model flipped from “borrow to build the library” to “the library funds itself.” One more thing worth saying: stock comp is only 3.4% of net income here, so the recent FCF/NI is genuinely clean. The raw score stays 0 because the QS is measuring consistency and Netflix simply wasn’t consistent. But if you’re underwriting the future, the recent trend is the tell, not the 15Y median.

Step 4. Share Count: is management buying back stock or diluting me?

Slowly buying back. No games here.

The bar: I score four windows (3, 5, 7, 10 years) on annualized share change. Shrinking 3%+ a year is a 10. Shrinking 1%+ across all four is an 8. Stable or slightly shrinking, within plus 1% a year, is a 6.

Shares outstanding peaked around 4,453M in 2022 and have drifted down to 4,222M. The 3Y annualized change is minus 1.76%, the 5Y is minus 0.95%, and the longer windows are basically flat. All four windows are stable or shrinking, so it clears the 6 tier. It misses the 8 because the 5, 7 and 10 year windows don’t shrink a full 1% a year.

My read: Netflix started returning cash through buybacks in earnest once it went FCF positive, and it’s been steady about it. Worth knowing: after briefly pausing buybacks for the Warner Bros. bid it later abandoned, Netflix just ran its biggest buyback quarter ever. The 6 scores the history; the present is more aggressive.

Step 5. Margins: is the business getting better or worse?

Better. Meaningfully better.

The bar: I average gross, net and FCF margin over each window and compare to the 20-year average. A margin passes if it’s within 100bps of, or above, that long-run average. Fifteen checks total. Ten points if all fifteen expand.

Gross margin runs from a 36% long-run average up to 45% over the last three years. Net margin went from an 8.8% average to 21% recently. Both of those pass every window. The only misses are on the FCF margin, where the 15Y and 10Y windows sit below the long-run average, again because of the burn years. Thirteen of fifteen checks pass. That’s an 8.

The margin story is the bull case in one chart. Operating margin hit 29.5% in 2025, up from 26.7% the year before, and management is guiding to 31.5% for 2026. The business is getting better, not worse, and the QS agrees.

Step 6. Moat: can anyone else do this?

The QS says “meh.” I disagree, and here’s the honest wrinkle.

The bar: this one’s tiered on historical average gross margin AND ROIC together. A 10 needs 60% gross margin and 25% ROIC. A 6 needs 40% gross margin and 15% ROIC. If gross margin can’t clear 40%, the best you can do is a 2.

Netflix’s historical average gross margin is 36.5%. That’s below the 40% gate. So even though average ROIC clears 15% with room to spare (10Y ROIC is 20.7%, current is 35.6%), the low gross margin caps this at a 2. The rubric is built for asset-light software and it punishes a business that runs content amortization through cost of goods. That’s what’s happening here.

So take the 2 with a grain of salt. Morningstar rates Netflix a narrow moat, built on intangibles and a network effect, and treats it as the clear standout among the streamers they cover. The current gross margin is 48% and rising as the library scales. I’d argue the true moat is wider than a 36.5% gross margin suggests, and the expanding margin trend is the evidence. But the raw score reads what the historical average reads. It’s a 2 on the page and a “narrow moat and widening” in my head.

Step 7. Management: is capital being allocated well?

Pretty well, and getting better.

The bar: tiered on average ROE and ROIC. A 10 is 30% ROE and 25% ROIC. An 8 is 25% and 20%. A 6 is 20% and 15%.

On the longest-available average, ROE is 20.8% and ROIC clears 15%, which lands it at a 6. But look at the trend. On a 10-year basis, ROE is 25.2% and ROIC is 20.7%, which would score an 8. The current-year ROE is 41% and ROIC is 36%. So the honest read is “6 on the full history, 8 on the modern business, and the arrow is pointing up.” I scored the 6 because that’s what the rubric’s longest-average rule produces. The capital allocation quality is clearly improving as the burn years roll off the average.

Step 8. Balance Sheet: can it survive a storm?

Easily now. It was hairier a few years ago.

The bar: I score interest coverage and net-debt-to-EBITDA together, taking the lower of the historical average and the current year. A 10 needs 25x coverage and under 1x leverage. A 6 needs 15x and under 3x.

The net-debt-to-EBITDA picture is pristine. It’s 0.18x today and has averaged around 0.2x. Netflix is basically net-cash-neutral. The thing dragging this to a 4 is the interest coverage average. Back through the 2012–2020 stretch, as Netflix borrowed its way to a debt load that peaked above $16B, coverage sat as low as 2 to 4x. That pulls the historical average down to 12x, which caps the score at the 4 tier, even though current coverage is a comfortable 17x.

My take: this is a backward-looking penalty. Netflix deleveraged hard once the cash turned, and today the balance sheet is a non-issue. The 4 is the ghost of the debt-funded content era. Real, but fading.

Step 9. Future Growth: does the market expect the compounding to continue?

Yes, but the EPS path is lumpy.

We just established what Netflix looks like from the inside. Steady revenue, expanding margins, cash that finally shows up. Now the question is whether the street expects that to keep going. Here’s consensus.

The bar: I check revenue and EPS estimates for the next three fiscal years plus their average. Each check passes the base tier if revenue is 7%+ AND EPS is 10%+. All four checks at 15% revenue and 20% EPS is a 10. Three of four at the base tier is a 4.

Revenue estimates are solid all the way out: 13%, 11%, 10% for the next three years, decelerating gently. EPS is where it gets bumpy. FY1 is a huge 40%, then FY2 drops to just 7.5%, then FY3 back up to 20%. That FY2 EPS number is the problem. At 7.5% it’s below the 10% bar, so one of the four checks fails. Three of four pass. That’s a 4.

Is the FY2 dip real or noise? It reads like a margin-investment year sitting between two operating-leverage years, the kind of lumpiness you get when a company steps up content or ad-tech spend. Compare it to Step 2, where Netflix actually delivered 30%+ EPS growth for years. The street is modeling slower and lumpier EPS than the business has historically produced. Whether that’s conservatism or realism is the whole valuation question. Netflix’s own FY2026 guide is 13-14% revenue and $12.5B of FCF, which lines up with the revenue estimates here.

Step 10. Predictability: can analysts even model this thing?

Yes, but it isn’t clear from the score.

Quick framing first. Consensus estimates cluster because analysts herd. There’s a whole matrix of being right or wrong alone versus right or wrong with the crowd, and most analysts sit in the “with the crowd” boxes. For a compounder, clustering is a feature. If the business is genuinely modelable, analysts should cluster tight, and they should be right a lot. So the question is simple: have they been able to model Netflix?

The bar: I count every beat and miss on the surprise chart and score on the lower of the revenue and EPS beat rates. 95%+ is a 10. 90% is an 8. 80% is a 6. 70% is a 4. 60% is a 2. Below 60% is a zero.

Across roughly 95 quarters, Netflix missed on revenue 33 times and on EPS 17 times. That’s a 65% beat rate on revenue and 82% on EPS. Since the score takes the lower of the two, revenue’s 65% sets the grade, and 65% only clears the 60% band. That’s a 2.

Revenue misses are very small, with the beat and miss range being 0.5-1% on average. So, this score is a little bit harsh.

The Final Score

44 out of 100. C Tier.

Here’s the honest one-line version: Netflix is a genuinely great growth business that the QS grades harshly because it spent a decade not looking like a clean compounder. Two categories carry the score. Growth is a perfect 10 and margins are an 8, and both are still improving. Share count and management sit at 6, solid and trending up.

The damage is spread across four places, and they’re worth separating.

FCF/NI at 0 and the balance sheet at 4 are scars from the 2015-2019 content-borrowing era. That cash burn was real, and the QS refuses to forget it, but the business flipped to $9.5B of free cash flow and a net-cash balance sheet. Those are backward-looking. Eventually this one should trend closer to 10.

Moat at 2 is a rubric-fit problem: the framework wants a 40% gross margin and Netflix runs content through COGS, so a narrow-moat streaming leader screens like a no-moat company. Maybe a 6, using the Morningstar rating as a proxy.

Returns at 2 is a price problem, not a business problem. The compounding is there, the path was just volatile.

Predictability at 2. Revenue misses were tiny, closer to 0.5%, and beats were 1% on average. Maybe 6-8 going forward.

Not very low like the four above, but the management and balance scores are 6 and 4. Both are trending upwards towards a score of 8 and 10 based on the rubric shared in their respective steps.

So, adjusting for the trends and adding the net number to the current score:

10 (FCF/NI) + 4 (Moat) + 6 (Predictability) + 2 (Management) + 6 (Balance sheet) = 28. Then 44 + 28 = 72. B Tier. This looks more like it.

Now, let’s take a look at valuation.

What the Market Has Paid

That 90x number is not a valuation anchor. One thing I want to be honest about. I do not have a clean 20-year or 10-year normal multiples as usual to build a proper PEG range from. The window I can read is polluted by the early years. So the three cases below are built off multiples I picked by hand rather than PEGs the market handed me. Treat them as my assumptions, not as history.

Now, let’s see what we can make based on what multiple we assume.

What You Actually Make

Three cases. I am using 15x, 18x and 22x.

15x is the mechanical fair value the formula spits out for 14.53% growth. Call it the bear. It assumes Netflix is a mature media company and gets a mature media multiple.

18x is my base. It gives Netflix a 20% premium for pricing power and the ad business. My base case is “the multiple shrinks further.”

22x is the bull and 20% premium from 18x, and it is simply where the stock trades today. That should tell you something about how far this thing has already de-rated. My bull case is “the multiple stops falling.”

Or you could think of it like this: the current 22x is the starting point. Then discount that by 20% to assume a multiple contraction to account for the lower growth of 14%. Then set the floor multiple at 15x, basically a PEG of 1 (15 P/E divided by the growth rate of 14%).

The shorter the hold, the more everything rides on the exit multiple. Stretch it out, and the earnings do the work instead.

Shorter hold, to end of 2028:

Longer hold, to end of 2031:

The short hold at the bear multiple gets you exactly nothing. Zero. Two and a half years of your capital for a rounding error.

Stretch to 2031 and the bear case turns into 6.5% a year. Still not good. But that is the whole point. The earnings compound underneath you and the multiple stops mattering as much.

Now the fair value math. On NTM EPS of $3.46:

Do you own Netflix? Why and why not? Comment below.

After scoring Mastercard in the previous post, it made sense to score Visa next.

It scored 80 vs 86 for MA. Investors know that Visa is a slower version of Mastercard, and my Quality Score picked up on that, hence the points difference. Still, that makes it an A Tier business.

I think Visa is overvalued at $362 with about a 28x blended P/E (second half of 2025 and the first half of 2026). Over five years that math spits out roughly 12% a year if the multiple just holds. If it doesn’t, you are in for 5-12%.

Let’s find out why it scored 80 and is overvalued compared to MA.

Upgrade for $120/year:

• Every paid post: Quality Scores, Valuations, and Trades.

• The full 20-name portfolio and 250-name watchlist.

• The 10,000-stock dataset to find your own Compounders.

  Subscribe now

Before you read further, please like and restack. Means a lot to me! Thank you!

Step 1. Historical Returns: how has the stock done, and how steadily?

Short answer: great over the long haul, choppy lately.

The bar: I check six windows (3, 5, 7, 10, 15, 20 years) and each one passes only if the stock compounded at 10% or better AND did it steadily, meaning an R-squared of 0.85 or higher. R-squared is just how close the price line hugs a straight trend. High R-squared means smooth compounding. Low means a rollercoaster. Ten points if every window passes. Eight if all but one. Six needs two-thirds. Below that you start bleeding points fast.

Visa only has price history back to the March 2008 IPO, so there’s no 20-year window. Of the five I can measure, three pass: 7Y (11.6% CAGR, R-squared 0.86), 10Y (17.7%, 0.93) and 15Y (21.3%, 0.96). Beautiful, smooth compounding over the long windows.

The two that fail are recent. The 5-year CAGR is 9.6%, which misses the 10% bar by a hair, and its R-squared is 0.82. The 3-year R-squared is 0.72, which is the rollercoaster showing up. Three of five passing gets 4 out of 10.

Now, a 4 on a business this good looks wrong, so let me confront it. This isn’t the business breaking. Look at Steps 2 through 8 and you’ll see the machine underneath is humming. This is a textbook example of stock price getting ahead of the business, then multiples compressing. We will see it clearly in the valuation section.

Vs Mastercard: dead heat at 4/10. Both stalled the last three years. MA has the longer record (20-year CAGR of 26.7% since its 2006 IPO) and Visa only goes back to 2008, but the recent chop hit both the same way. Visa’s 5-year CAGR of 9.6% actually held up a shade better than MA’s 5.6%.

The market has rewarded this business enormously over 15 years. But price alone doesn’t tell us why. Let’s look under the hood.

Step 2. Historical Growth: is the business actually growing?

Short answer: yes, and it barely takes a breath.

The bar: three metrics (revenue, diluted EPS, free cash flow) across all six windows. Revenue passes at 7% CAGR or better. EPS and FCF need 10%. That’s up to 18 checks. Ten points if all 18 pass. Eight if you miss only one or two.

Revenue is 6 for 6. Every window from 3 to 20 years compounds between 10% and 15%.

EPS is even better, 6 for 6, running 14% to 26% depending on the window. That leaves FCF, which passes 4 of 6.

Here’s the context about Visa’s lumpy-looking free cash flow. Three things actually move the cash. Client-incentive payments, which spike when a batch of big deals renews at once. Income-tax timing, exactly as thrilling as it sounds. And the litigation escrow, the pile Visa keeps feeding to settle the interchange fight with merchants.

That escrow explains the FY2024 dip. On September 26, 2024, four days before the fiscal year closed, Visa dropped $1.5 billion into it. That one deposit is most of why FY2024 FCF landed at $18.7B while 2023 printed $19.7B. The business was fine. Just a calendar quirk.

16 of 18 checks pass. That’s an 8 out of 10.

Vs Mastercard: MA takes this one, 10 to 8. That’s the first of three places MA pulls ahead. MA’s free cash flow cleared 10% in all six windows. Revenue and EPS growth are basically neck and neck between the two. MA just had cleaner FCF across the board.

Growth is nice. But growth without cash is just accounting. So let’s check the cash.

Step 3. FCF / Net Income: is that growth real cash or just accounting?

Short answer: real cash. Every dollar of profit becomes a dollar of cash, and then some.

The bar: I take the median FCF-to-net-income ratio in each window and find the highest tier they all clear. All windows at 100% or more gets a 10. All at 90% gets an 8. All at 80% gets a 6. The score caps at 100%, so there’s no bonus for going higher.

Every single window sits above 100%. The medians run from 101.8% (20-year) up to 110.8% (recent). Full marks, 10 out of 10.

And before you ask the obvious question. No, this isn’t stock comp inflating the cash flow. SBC is only 4.5% of net income and has averaged 4.3% over five years. That’s tiny. The reason FCF runs above net income is boring and good: depreciation runs a bit ahead of capex on an asset-light payment network, plus favorable working capital. This is about as clean as cash conversion gets.

Vs Mastercard: tie at 10/10. Both turn every dollar of profit into more than a dollar of cash. Two toll bridges doing the same clean thing.

Real cash. But cash conversion with a leaky share count is a treadmill. Is management diluting it away?

Step 4. Share Count: is management buying back stock or diluting me?

Short answer: buying back. Steadily.

The bar: annualized share change across four windows (3, 5, 7, 10 years). I don’t score 15 or 20 years because the early post-IPO share math is noisy. All four shrinking 3% a year or more is a 10. All four shrinking 1% or more is an 8. All four flat-to-down is a 6.

Shares outstanding are down every window: 2.5% a year over 3 years, 1.7% over 5, 1.8% over 7 and 2.2% over 10. Class A shares outstanding, the clean count, from 3,377 million in 2008 to 1,933 million today. All four windows shrink more than 1% a year, so this clears the 8 tier. None reaches the 3% a year needed for a 10. Call it a steady 8 out of 10.

Vs Mastercard: tie at 8/10. Both shrink the count every year at roughly the same clip, around 2% a year, and neither hits the 3% bar for a 10. On a clean shares-outstanding basis they’re basically twins here.

Step 5. Margins: is the business getting better or worse?

Short answer: better. And it wasn’t bad to begin with.

The bar: gross, net and FCF margin, each measured as a window average against the longest-available average. Stable (within 100bps) or expanding passes. That’s up to 15 checks. All 15 stable or expanding is a 10.

All 15 pass. Start with gross margin, and here’s a correction worth making. Visa’s real cost of revenue is one small line: network and processing, the direct cost of running VisaNet, about 2 to 3% of revenue. Strip that out and gross margin runs ~96 to 98%. More important than the level is the direction. It has climbed from 95.7% in FY2011 to 97.8% in FY2025, a steady grind higher as the network scales and each incremental transaction costs almost nothing to carry. That’s real operating leverage at the very top of the P&L.

Net margin is the other eye-opener. Recent windows average 52% against a 32% long-run number. FCF margin, same story, 56% recent versus 39%.

Vs Mastercard: tie at 10/10, and it’s the same shape. Both carry only a tiny direct network cost, so both run gross margins in the high-90s with net and FCF margins expanding underneath. Two nearly identical P&L structures. Dead tie.

Expanding margins are great. But margins without a moat are temporary. So can anyone actually replicate this?

Step 6. Moat: can anyone else do this?

Short answer: no. This is one of the widest moats in the market.

The bar works in tiers. A 10 needs a long-run gross margin of 60% or better AND ROIC of 25% or better, with the current trend stable or rising. Step down from there.

Gross margin is ~97% and rising (net of Visa’s small network and processing cost, as covered in Step 5) and current ROIC is 47.9%, way above the long-run average and climbing. The long-run ROIC average is around 30% (or 25% if I strip out one artifact I’ll mention in a sec). Either way it clears the 25% bar with the current trend pointing up. That’s a 10 out of 10.

Vs Mastercard: tie at 10/10. Two halves of the same duopoly. MA’s ROIC is even higher (~48% average, 60% now, against Visa’s ~30% average and 48% now), but both are miles past the bar. Nobody is building a third global network.

A quick note on ROIC calculation without going into detail: I checked 4 different data sources, and depending on how it’s calculated, the data range can be 25-50%. Yeah, that’s a big range, but for our purposes the score goes from 8 to 10.

A moat is great. But a moat with bad capital allocation is wasted. So how’s management deploying?

Step 7. Management: is capital being allocated well?

Short answer: very well, and getting better.

The bar: long-run ROE and ROIC plus the current trend. A 10 needs ROE of 30% or more AND ROIC of 25% or more. An 8 needs ROE of 25% and ROIC of 20%, current stable or rising.

Current ROE is 52.9% and marching straight up the chart. The catch is the long-run average ROE, which lands around 26 to 28%. That’s below the 30% bar for a 10. So this is an 8 out of 10.

Why is the average dragged down? Two reasons, both explainable. First, the 2008 IPO dramatically expanded Visa's equity base. ROE started at just 3.6% because earnings were being measured against a much larger equity base, then climbed steadily as aggressive share repurchases shrank equity over time. Second, FY2012 net income was depressed by a $4.1 billion pretax litigation provision (roughly $2.9 billion after tax) related to the interchange antitrust settlement. That’s the visible dip in the middle of the chart. The trajectory here is a 10. The 20-year average is an 8. The rubric scores the history, so 8 it is.

Vs Mastercard: MA wins, 10 to 8. And here’s the honest asterisk. MA’s average ROE runs around 71% partly because it has bought back so much stock its equity base is tiny (tangible book is actually negative). Visa still carries the big equity pile from its 2008 IPO, so its ROE looks lower on paper. On ROIC, the cleaner read of return on capital, both are elite. MA scores higher on the letter of the rubric.

Capital allocation is strong. But can the business survive a storm?

Step 8. Balance Sheet: can it survive a storm?

Short answer: easily. This is a fortress, basically a needless exercise.

The bar: interest coverage and net-debt-to-EBITDA, both on a historical median and current basis, scored on the lower of the two. A 10 needs coverage of 25x or better AND net debt under 1x EBITDA.

Current interest coverage is 45x. Median is 31x. Net debt to EBITDA is 0.29x today, 0.09x median. Both metrics sit in the top tier. 10 out of 10.

One standout on the chart. Interest expense jumps from $3M in FY2015 to $427M in FY2016, and net debt swings from negative $3.5B to positive $10.3B. That’s the Visa Europe acquisition, which closed June 2016 and was funded with $16B of senior notes issued the prior December. Even after taking on that debt, interest coverage never fell below roughly 23×, and net leverage remained comfortably below 1× EBITDA.

Vs Mastercard: tie at 10/10. Both are fortresses.

We’ve established what Visa looks like from the inside. The fundamentals are elite and they fully support the price performance from Step 1. Now the question becomes: does the market expect this to continue? Here’s what the analyst consensus says.

Step 9. Future Growth: does the market expect the compounding to continue?

Short answer: yes, steady double-digits, just not screaming.

The bar: consensus revenue and EPS growth for the forward years plus the average. Each check passes a tier only if BOTH revenue and EPS clear it. A 10 needs 15% revenue and 20% EPS. An 8 needs 10% and 15%. A 6 needs 7% and 10%.

Revenue estimates: 13.9%, 10.5% and 10.2% for the next three years. EPS estimates: 14.6%, 13.1% and 13.2%. All four scored checks clear the 6 tier comfortably. What stops them at 8 is the EPS side. Those forward EPS numbers cluster around 13 to 15%, and the 8 tier wants 15% or better across the board. So close. 6 out of 10.

Stack it against Step 2, where EPS actually grew 13 to 26% historically. The estimates are in line to a touch below what Visa has delivered. The out years are interesting too: FY4 estimates look strong (Rev 12%, EPS 16%), FY5 softens (Rev 5.5%). The near-term picture is healthy, dependable, mid-teens growth.

Vs Mastercard: MA wins, 8 to 6. The third and final spot where it edges ahead. Revenue outlook is basically the same, low-teens for both. The difference is forward EPS. Analysts have MA around 16% and Visa around 13 to 15%. MA clears the 15% EPS bar for an 8. Visa lands just under it.

Analyst estimates are useful. But are they reliable for this specific name? That’s the last question.

Step 10. Predictability: can analysts even model this thing?

Short answer: mostly, yes. On earnings especially.

The bar: I count every quarterly surprise bar, beats versus misses, for both revenue and EPS, then score the lower of the two overall beat rates. 95% or better is a 10. 90% is an 8. 80% is a 6.

On EPS, Visa beat or met in about 67 of 71 quarters. That’s a 94% beat rate. The four misses were tiny and clustered in 2013 to 2016 plus the 2020 COVID quarter. On revenue, the beat rate is lower at 88.7%, eight misses out of 71 quarters, because revenue surprises are noisier than EPS.

Score the lower of the two, which is revenue at 88.7%. That clears 80% but sits just under 90%. So 6 out of 10. If you scored EPS alone, this would be an 8. Revenue is the binding constraint here.

Vs Mastercard: tie at 6/10. Both get gated by a sales beat rate under 90% (Visa 88.7%, MA 85%). Both nail EPS (Visa 94%, MA 96%). Visa’s revenue prints are actually a touch more reliable than MA’s. Same score anyway.

The Final Score

80 out of 100. A Tier.

The business categories are basically flawless: 10s on cash conversion, margins, moat and balance sheet, plus 8s on growth, buybacks and management. That’s a genuine compounder with one of the widest moats I score.

So what’s holding it out of S Tier? Two things, and they’re different in kind. Steps 9 and 10 (future growth at 6, predictability at 6) are the honest, boring truth that Visa is a mid-teens grower whose revenue prints are a little noisier than its earnings. That’s fine. It’s still elite.

The one that stings is Step 1 at 4. And I want to be clear about what that 4 is. It’s a price problem. The assumption for Step 1 is that compounders should just keep compounding and the price should reflect that but the price sometimes gets ahead of the business and then “waits” for the business to catch up. How do I know, steps 2-10 confirm that it’s business as usual.

Now the head-to-head, since I ran the same scorecard on Mastercard. MA came out at 86, S Tier. Visa at 80, A Tier. Here’s where the six points went:

Seven of ten categories are a dead tie. The whole gap is three boxes: Growth, Management and Future Growth, two points each. And I’d put an asterisk on one of them. MA’s Management 10 leans on a 71% average ROE that’s juiced by a bought-down, negative-tangible-book equity base. On ROIC, the two are neck and neck. So on pure business quality these are closer than 80 versus 86 makes them look. MA genuinely grows FCF a touch more evenly and analysts expect slightly faster EPS from it. The rest is a coin flip between the two best toll bridges in the market.

Which is the one question the QS deliberately does not answer: is it cheap today? The score tells me Visa is a A Tier compounder worth owning. It does not tell me the price to pay. That’s what we will check next.

What the Market Has Paid

Here is what Visa’s earnings have done and what the market handed over for them, across three windows compared to Mastercard.

In the early years Visa grew earnings at 22% and the market paid a 24x. Over the last decade the growth cooled to 16.6%, yet the multiple the market was willing to pay actually went UP to 30x. Slower growth, higher price tag. That is the market falling in love with the quality of the rails.

Today the blended multiple is 28x, so the stock sits right in the middle of its own history. Cheaper than the 30x froth of the last decade. Pricier than the 24x it paid in the early days when it was growing faster.

The dividend is almost a rounding error here at 0.74%, so make no mistake. This is an earnings-growth story.

Now, let’s see what we can make based on what multiple we assume.

What You Actually Make

The consensus forward growth rate is running around 14% for the next 5 years, but Visa has never grown at that rate before.

Using the historical periods below, we get 1 to 1.9 PEG ratios:

If the consensus growth had matched one of these periods, we could’ve assumed one of the historical multiples, but it is lower.

So, should we adjust it for growth? Using 1 PEG would give us a 14x multiple. The problem is, Visa has essentially never traded there. So using 14x as my base case would be assuming the market is about to re-rate a wide-moat monopoly. So, no durability premium. Still, we could use it as the bear case.

How about 1.3 PEG? 1.3x14=18 multiple. This is the base case. MA traded at 1.3 PEG up until 2016. It respects what the market has always been willing to pay for Visa’s quality, and it pulls the multiple down honestly to account for the slower growth.

And using the 1.84 PEG with 14% yields a 26x multiple.

Now the returns. And here is why I’m showing two different holding periods instead of one.

The further out you forecast, the shakier the estimate gets. There are usually 10-30 analysts providing estimates for the next 3 years, while five years out there are 1-5.

On top of that, the shorter your hold, the more your return depends on where the multiple happens to land on the day you sell. Stretch the hold longer, and the earnings growth starts to dominate, and the multiple noise washes out. So I show a near horizon and a far one. You get to see how much of the outcome rides on the multiple versus the compounding.

Shorter hold, to the end of FY2028. Not looking good in the short term:

Longer hold, to the end of FY2031. Now this looks better at our base case of 18x.

I feel like the multiples assumed were a little bit harsh, but it helps us see the returns with potential multiple contraction.

So the fair value band. On NTM EPS of roughly $13.90:

Again, the longer you hold a compounder, the more the business does the heavy lifting and the less the exit multiple matters.

Would you buy now or wait for the market to hand you a better price?

It’s an S-tier business, and my quality scoring system confirmed it with an 86-point score. It’s a buy at $484. See the valuation section at the bottom for potential returns.

Let me go step by step.

Upgrade for $120/year:

• Every paid post: Quality Scores, Valuations, and Trades.

• The full 20-name portfolio and 250-name watchlist.

• The 10,000-stock dataset to find your own Compounders.

Before you read further, please like and restack. Means a lot to me! Thank you!

Step 1. Historical Returns: how has the stock done, and how steadily?

Done well, like a straight line.

The bar: for each timeframe I check (3, 5, 7, 10, 15, 20 years) the stock has to compound at 10% or better with an R-squared above 0.85. R-squared is just how straight the line is. Clear all six and it’s a 10. Clear half and you get a 4.

Since the 2006 IPO, Mastercard has compounded at 26.7% a year. On an R-squared of 0.97. For the non-nerds, R-squared near 1.0 means that line went up in an almost straight diagonal for twenty years. That’s about as clean a compounding chart as exists.

So why did this only score 4/10?

MA clears three. The 10, 15 and 20-year windows are gorgeous, 19% to 27% a year on a 0.94-plus R-squared. The recent ones are ugly. The 5-year CAGR is 5.6%. The 3-year is 9%. The stock has basically gone sideways since 2021 while the business kept printing low double-digit growth.

The business is fine while the multiple compressed. MA was a premium-priced growth name; the air came out of that whole bucket after 2021, and the price spent three years digesting it while earnings caught up.

Step 2. Historical Growth: is the business actually growing?

Yes. And it’s not close.

The bar: revenue compounding at 7% or better, EPS and FCF at 10% or better, across all six timeframes. Three metrics times six windows is eighteen checks. You need to pass basically all of them for a 10.

Revenue grew 13% a year over the last decade. EPS grew 17%. Free cash flow grew 16%. Pick any window. 3, 5, 7, 10, 15, 20 years. Every single one clears the bar. That’s a 10/10, and it’s rare.

One honest note. The 5-year growth numbers look extra juicy because they start in 2020, when revenue actually fell 9% and net income dropped 21%. Cross-border volume cratered 29% when nobody was flying. Low base, fat-looking CAGR. The 7 and 10-year numbers are the clean read, and they’re still excellent.

Step 3. FCF / Net Income: is that growth real cash or just accounting?

It’s real cash.

The bar works in tiers. Convert at 100%+ across every window and it’s a 10. 80% gets an 8. 60% gets a 6. MA sits around 106% everywhere, so it’s a clean 10. Every dollar of reported earnings shows up as a dollar of cash, and then some.

Free cash flow beats net income basically every year. Median FCF to net income sits around 106% across every timeframe.

This is what asset-light looks like. They spend about half a billion on capex to produce thirty-three billion in revenue. The rest falls to cash.

Step 4. Share Count: is management buying back stock or diluting me?

Buying back.

The bar: shrink the count and you score. Buy back 3%+ a year and it’s a 10, 1%+ is an 8, dead flat is a 6, dilution is a fail.

Shares shrink about 2% a year, every year, across the 3, 5, 7 and 10-year windows, so an 8. The buybacks comfortably outrun the stock comp, so owners get a bigger slice each year without lifting a finger.

Step 5. Margins: is the business getting better or worse?

Better.

The bar here isn’t about the level, it’s about the trend. Each margin, at each timeframe, has to be flat or expanding versus its own long-run average. Gross, net and FCF across five windows, fifteen checks.

Gross is pinned at 100%, and net and FCF are both climbing, so all fifteen pass. Clean 10. Net margins went from about 40% to 45-46%. Free cash flow margins went from the mid-30s to over 52%. A 100% gross margin business that keeps half of every revenue dollar as free cash.

Step 6. Moat: can anyone else do this?

Not many.

The bar: average gross margin above 60% and average ROIC above 25%, with the current readings holding or rising, earns a 10. Step down to 50% and 20% and it’s an 8. 40% and 15% gets you a 6.

MA has a 100% gross margin and ROIC averaging 48% over fifteen years, currently at an all-time high near 60%. It clears the top bar with room to spare. A two-sided network with switching costs and a brand on every card and terminal on earth. 10/10.

Step 7. Management: is capital being allocated well?

Very well.

The bar mirrors the moat step. Average ROE above 30% and average ROIC above 25%, current readings holding or rising, is a 10.

MA’s average ROE is 71% and ROIC is around 48%, both climbing.

Here’s why ROE is so high. Mastercard buys back so much stock that there’s barely any equity left on the books. Tangible book equity is actually negative. When the denominator shrinks toward zero, the ratio explodes. So a 200% ROE isn’t some accounting trick. It’s what happens when a cash machine returns almost everything to shareholders.

Step 8. Balance Sheet: can it survive a storm?

Easily.

The bar: interest coverage above 25x and net debt under 1x EBITDA, and it has to hold on both the long-run average and today. Coverage above 20x or debt under 2x drops you to an 8.

MA earns 27x its interest bill and carries net debt of 0.4x EBITDA. They went from net cash to lightly levered to fund buybacks, but this is still a fortress. 10/10.

Step 9. Future Growth: does the market expect the compounding to continue?

It does.

The bar runs on the forward estimates for the next three years plus their average. Revenue above 15% and EPS above 20% across all of them is a 10. 10% revenue and 15% EPS is an 8. 7% and 10% is a 6.

Consensus has revenue growing 12-13% and EPS around 16%. That clears the 8 tier comfortably but falls short of the 15/20 needed for a 10. So an 8.

Step 10. Predictability: can analysts even model this thing?

Mostly.

The bar is the beat rate, and I take the lower of sales and EPS so one good line can’t carry it. 95%+ is a 10, 90% an 8, 80% a 6.

There are 80 quarters of miss and beat data since the IPO. EPS: 77 beats, 3 misses, 96%. Sales: 68 beats and 12 misses, so 85%. The misses bunch up in the 2010 to 2020 stretch, where both five-year windows ran beat rates of 75%.

The lower of the two is sales at 85%, so a 6. Still, for a compounder, analysts clustering tight is a good thing, and on EPS they nail it. The business is modelable.

The Final Score

An 86, which lands it squarely in S Tier.

I can defend every point. The only category still dragging is returns, and that’s the price stall. The growth, the cash conversion, the margins, the moat, and the balance sheet are all 8s and 10s. There is no quality question here.

The only real question is price. The QS confirmed the business. It says nothing about whether the stock is cheap today. I’ll take a look at that next.

What the Market Has Paid

Here is what the market actually pays for this thing over time:

MA compounded earnings at about 23% a year since IPO, with its multiple expanding in the last 10 years.

Today the stock trades at 26.5x blended earnings or a 24x forward P/E. Above where the cold formula (growth=P/E) says it belongs but very close to the 2010-2016 era with 19% growth.

What You Actually Make

The consensus has forward earnings growing about 16% a year in the next 5 years.

The fair value formula for a 16% grower is 16x. Only briefly after the GFC did MA trade near that level. So I'm not going to use 16x as my base case. I chose 21x as my base case due to the slower growth rate of 16% versus the historical 19% with 25x.

That’s actually a similar PEG to the historical: 25/19=1.32 and 16×1.32=21.

First, the shorter timer period, holding to the end of 2028:

Then the longer hold, out to the end of 2031:

Hold for three years and you could earn 6-14%, while five years could yield 11-15%. This is why holding longer can help with multiple contractions, even if you end up overpaying for a quality company.

And for the returns, I’m not betting on a return to 30x, and I won’t underwrite 16x on a business this good. Call fair value 21x to 25x. Or using the NTM EPS of 20.23, that’s 424 and 506, about 15% overvalued and 4% undervalued. In other words, it’s trading around fair value based on the multiples that I assumed with 11-15% potential annual returns in the next 5 years.

Would you pay 26.5x for the rails or wait for a pullback?

All company research boils down to quality, growth, and valuation. This system crowdsources the whole process and beats the market.

Quality: I use Morningstar moat ratings that are assigned by a committee of 20 analysts. They cover about 1000 companies: 200 are wide and 400 are narrow moat rated. My watchlist is about 100 names plus.

Growth: Consensus of 5-50 analysts; large caps are usually covered by around 30 analysts. I ignore anything with earnings growth less than 10% and usually go for companies growing revenue more than 10%.

Valuation: market-determined historical multiple in the last 10-plus years. The key word here is market-determined. It means the market weighed the substance of a company and decided that this is the multiple it deserves.

Seriously, that’s it. I’ve worked 6 years to finalize this system. And I’ve come to realize that simple is better.

Here is the current portfolio:

And my watchlist:

Here is a recent post showing the system in action.

So, Compounders is quality at a fair price and systematic. “Long-term, hands-off, for busy people; zoom out and check your portfolio once a year” are the words to describe this strategy. Potentially the next best thing to buying indexes in terms of time spent and returns to try to beat them.

Upgrade for $10/month or $120/year:

• Posts: Every idea with its quality, growth, and valuation.

• The full 20-name portfolio and 100-name watchlist.

• Trade Alerts.

  Subscribe now

My Readers Find It Useful

Every single one of these keeps me going! Thank you!

This is What the Other Writers Think

Very kind words! Thank you!

Does It Work?

I’ve been managing this Compounders portfolio since 2021, beating the SPY and QQQ.

Returns are money-weighted.

Total returns: 74% for 3 years and 76% since inception.

Annual returns: 15% since inception.

I believe a long-term portfolio should be built around AMZN, GOOGL, MSFT, and META. So, this portfolio has them.

This portfolio was built using a simple system:

1. Use Morningstar moat and management ratings to find quality companies

2. Use consensus estimates to cut off anything growing EPS below 10%

3. Use historical multiples as a reference for valuation

You get the portfolio in your email when you upgrade.

Upgrade for $10/month or $120/year:

• Posts: Every idea with its quality, growth, and valuation.

• The full 20-name portfolio and 100-name watchlist.

• Trade alerts.

What made it into the portfolio and why

Here is the final portfolio (THIS PART IS VISIBLE TO PAID SUBS ONLY):

What I threw out and why

Here’s why the big ones didn’t make it.

Wonderful, just too expensive. Finally, a long list of great companies that failed on one thing: price. These aren’t rejects so much as a watchlist for the next correction. Costco at a 49% premium to fair value. Walmart at 96% over. Apple at 68%. Deere at 91%. Caterpillar at 111%. Goldman Sachs at 47%. Cisco at 97%. Alphabet at 32x vs the historical 24-27x multiples. The semiconductor equipment names, ASML, Lam, KLA, Applied Materials, all trading at huge premiums after their run. Every one a business I’d happily own. None at today’s price. When these come back to earth, we’ll talk.

Unrated, so I trust the system instead of my gut. A whole group of genuinely great businesses got cut for one reason: Morningstar doesn’t rate them, so including them would mean overriding my own system on a hunch. KKR and the other alternative-asset managers. The Mexican airport monopolies are OMAB and PAC. Corpay in payments. The gold royalty names. The advisory boutiques like Houlihan Lokey.

Skipped because I already own the better version. A few names got cut simply because their slot was filled by something cheaper or stronger.

The financial-data cluster, where I kept the cheap ones. This corner is full of wonderful tollbooths, and they all do roughly the same thing, so I only need a couple. They are both great businesses, with 7% revenue and 12% earnings growth, both top-rated. But one trades at 29x versus 22x for the other. Easy pass. There is another one that is even richer at 32x for similar growth. Two more have narrow moat ratings while trading at 25x and 30x for similar growth rates.

A tale of two multiples. Both are wonderful software businesses with top-rated management. The first one grows faster, 20% revenue and 21% EPS against the other’s 12% and 15%. So why did the slow grower make the book? The multiples. The first one trades around 33x, a full multiple already. For it to really pay off, it needs to grow as projected without any multiple compression. The other trades at 13x, a low multiple for its quality and growth. Its upside comes from its multiple rerating to around 20x, which the fundamentals justify on their own, not back toward its own history of above 30x. Me being a value investor at heart, decided to go for a lower multiple. And I’m usually surprised when a high multiple goes even higher. I’m guessing this will be one of those cases.

Or maybe I’ll just find an excuse to include both later on.

So tell me. Which name would you have left out, and which one am I going to regret cutting? Drop a comment.

I ended up with 648 companies. That’s about 7.8% of all companies!

Next, I pulled consensus estimates for the next three years. Revenue and earnings.

Any company projected to grow revenues by less than 5% and EPS by less than 10% over the next three years was removed.

About 250 survived (3%) as a watchlist.

Then I prepared the watchlist for portfolio building by adding Morningstar moat and capital allocation ratings as a second opinion and historical multiples for valuation.

This post will follow a reverse order to keep things interesting:

1. The watchlist

2. Using the Idea Generation Tool to find compounders

3. How to use the Idea Generation Tool

4. How I came up with the Tool

5. Thinking behind it

Bonus

Here is free access to the spreadsheet with about 4,000 tickers across NYSE, OTC, and international listings. Just click “request access” and then make your own copy to use the filters.

Upgrade for the latest list that includes the NASDAQ.

Upgrade for $10/month or $120/year:

• Every paid post: Quality Scores, Valuations, and Trades.

• The full 20-name portfolio and 250-name watchlist.

• The 10,000-stock dataset to find your own Compounders.

Subscribe now

Here’s what the final list looks like under the hood.

250 companies made it through. 85 have wide moats. 88 are trading below my estimated fair value right now (May 25, 2026). 32 have a PEG under 1.0.

Yes, the usual suspects are here. All of the Mag 7 made the cut. So did the semiconductor heavyweights: AVGO, ASML, TSM, and AMD. Visa and Mastercard. Costco. Eli Lilly. If the screen didn’t catch those, I’d question the methodology.

Industrials dominate with 74 names. Tech and financials are second and third. There are three Mexican airport operators on here. A Canadian gold miner. 26 international names total across 10 countries.

The list skews large cap (169 names), but 48 mid-caps made the cut. Those tend to be the most interesting.

The cheapest name trades 47% below my estimated fair value. It’s a large-cap biotech with a PEG of 0.8. The highest estimated 3-year return on the list is 51%, a mid-cap pharma with a PEG of 0.9. And there’s a mid-cap broadcaster trading at a PEG of 0.3 with a 47% estimated 3-year return. I almost didn’t believe that one.

Every company on the list comes with a fair value estimate, PEG ratio, and estimated 3-year total return.

Here is the final watchlist for my paid readers with estimated fair values and 3-year returns and moat/management ratings.

One more thing. This methodology isn’t perfect. No methodology is. Consensus estimates are wrong all the time. Historical multiples shift. Moat ratings are opinions, not facts.

This is just one way to fish. It narrows 8,000+ names down to 250 and gives you a starting point. The real work starts after the screen.

Using the Idea Generator to find compounders

Welcome to the fun part. Here is our first batch of winners.

Filters used: Combined score higher than 90, first year (IPO) before 2010.

It looks like TDG took a breather this year from being a compounder with -44% CAGR this year (-10% YTD).

Another one is RBC; we click through the chart column. What a nice chart! A 35-bagger.

This is why I created this tool; it is a great way to quickly discover companies.

We loosen the First Year a little by filtering for companies that IPOed between 2010 and 2015.

More interesting ideas: KEN, AXSM, and FTAI.

I see some Singaporean and Canadian companies with market capitalizations as low as $5 billion in industries ranging from biotech to defense.

So far we have 26 companies in our watchlist.

You can also filter by each column instead of the Combined Score to have better control of the time periods.

I went ahead and cast the widest possible net by loosening the filters to a CAGR above 10% and filtering for 10, 15, and 20-year periods separately for the linearity columns above 0.8.

I ended up with 648 companies out of 8264! That’s about 7.8% of all companies!

If you’re saying, “Hold on, what kind of watchlist has 648 companies?” I’m aware of that.

So I reduced further.

Remember when I mentioned combining this data with consensus estimates? Any company projected to grow revenues by less than 5% and EPS by less than 10% over the next three years was removed.

After that, there were 250 companies left. That’s about 3% of all companies!

That’s an interesting number. Bessembinder’s research on 26,000 U.S. stocks since 1926 found that roughly 4% of all stocks created all the net wealth in the market. The other 96% collectively added nothing above T-bills. My screen landed on 3% using completely different inputs. Different method, same neighborhood.

These companies are historical winners with at least 10 years of performance and are projected to grow their revenue at least 5% and EPS more than 10% in the next 3 years.

I didn’t stop there, of course, as I compared this watchlist to Morningstar’s database of about 970 companies. About 23% out of those are rated wide moat.

It turns out the watchlist has about 85 or 34% wide moat companies vs. the 23% in the database.

It means this method has 1.5x as many wide-moat companies compared to the Morningstar database. That’s good evidence that the “reverse-engineer from price” approach is systematically surfacing quality businesses.

You can cut further. 51 companies remain out of 250 if you only pick the ones with Exemplary management ratings.

This is another way to cut down the list by adding a qualitative filter to it.

So, 8,265 companies. Filtered for 10%+ CAGR and linearity across 10, 15, and 20-year periods, higher than 0.8. Down to 648. Removed anything below 5% projected revenue growth and 10% EPS growth. Down to 250. Overlaid Morningstar wide moat and Exemplary management. Down to about 50.

That’s the top 0.6%.

How to use the Idea Generation Tool

Every stock has a Combined Score which is the average of Trend Quality and Growth Strength.

For instance, ALSN below has a Combined Score of 84 while having a Trend and Growth score of 88 and 80, respectively.

Blue means a score above 90, while green is a score between 80 and 90. They were color-coded so you could use them to filter quickly, as shown below:

Trend Quality is derived from the linearity data below and is how straight or exponential the stock price is. It is a number from 0 to 1, with higher being better.

Time periods for CAGR and linearity data go from 1 to 20 years.

Each stock is analyzed for a best-fit model and categorized into Exponential (think Nvidia) and Linear (think Visa).

Growth Strength is derived from the CAGR rate (annualized returns). And of course higher is better.

And there is a quick information section, such as an industry and business description, to get us up to speed from the get-go.

Also, the links in the chart column take you directly to Google Finance for a visual check of the CAGR and linearity.

The first-year column is the IPO date, while the cutoff is 2001. I think 20 years of stock performance is good enough to gauge the performance.

How I came up with it

First of all, a wide net needs to be cast to catch them all. There are a few ways to go about it.

I could plug in some filters and let a screener spit out a list. That works. But it makes you pick the reasons (business fundamentals) for the value creation in the market (stock performance).

How about we reverse-engineer the business performance with stock price?

Why? Because the market is a voting machine in the short term and a weighing machine in the long term. A stock can’t go up consistently for a decade by accident. The business has to be doing something right for it to go up in a straight or exponential line, such as below.

If you’ve been following my work, you know that I’ve built a tool that scans every US-listed stock (including OTC, over 8,000 tickers) and measures two things: how fast the price compounded (CAGR) and how straight (linearity) that compounding was.

It is called Generate Ideas and lives inside my paid Compounders spreadsheet.

That’s what we are going to use today and create this watchlist together.

Thinking behind it

Updating my watchlist has been in my backlog for a while now, but I’ve been putting it off because I wanted to look at everything that is listed on the US exchanges.

So, I’m going to walk you through the methodology that I followed to create this ultimate watchlist of compounders.

Before we kick off, there are a few questions that need to be answered.

What is a compounder anyway?

Compounder is a company that consistently, and over the long run (10-plus years at least for me), creates value (cash) for its shareholders.

Or to describe it quantitatively, it’s a business with fundamentals that go to the right and up over time with ROIC higher than the cost of capital.

How do we recognize one?

They usually have kind of a moat and good capital allocators that result in ever-increasing free cash flows per share for very long time periods.

Will it continue being a compounder?

The past is a good proxy for the future.

The longer a company has compounded, the more likely it is to keep compounding. Not because the past guarantees anything. But 15 years of consistent value creation is evidence of something durable underneath. A moat, a culture, a reinvestment engine. Something real. This is the Lindy Effect applied to stocks.

Mauboussin’s research on ROIC persistence backs this up. He tracked 1,000 companies across ROIC quintiles over a decade. Top-quintile companies stayed in the top quintile at rates well above what chance alone would predict.

I combine that with consensus analyst estimates to check if the compounding has legs going forward.

If you scrolled/read all the way here, please leave a like. Thank you!

If you are reading this for the first time, here is a brief intro:

I’m using stock price annual returns (CAGR) and how straight or exponential the stock price is to find compounders. The assumption is that the stock price and its smoothness reflect the quality of the underlying business, especially in the long run (10 plus years).

Here is free access to the dataset, which also has all of the US exchanges. Link.

Let’s take a look at the dataset.

Upgrade for $10/month or $120/year:

• Posts: Every idea with its quality, growth, and valuation.

• The full 20-name portfolio and 100-name watchlist.

• Trade Alerts.

Subscribe now

What’s in the dataset?

There are about 1500 companies in there. Sweden has the most listings, with about 880 of the total, followed by Norway. Iceland has the smallest exchange.

Nordic markets aren’t that different from the US in terms of market cap distribution. In both markets small caps lead in terms of numbers, followed by nano, mid, large, and mega in that order.

Small caps ($300mln to $2bn) make up most of the market with about 56% vs. 43% for the US.

In terms of sectors, Nordics are dominated by industrials, tech, and healthcare, while the US is dominated by financial services, healthcare, and tech in that order.

Energy is the smallest sector in both, with about 4% of the total market.

Now, here is the fun part of finding long-term compounders.

Finding Compounders

Filters applied: Trend Quality, Growth Strength, Combined score above 80, and First Year below 2015.

Only 44 out of 1532. That is 2.9% of all companies.

There are 13 large, 17 mid, 13 small, and 1 nano caps.

This is how the CAGR and linearity data look under the hood. Will they keep compounding and grow into large caps? They’re more likely to do so than a random company that is not a compounder. Winners keep winning.

Some familiar names top the list, such as ABB, Investor AB, and Lifco AB:

Now, I’m going to show you some companies from the database that had crazy returns in the last 10-20 years, but there will be no data except the total returns. Can you guess the company? Here we go.

A smooth compounder, large-cap, in an industry most people don't associate with 85x returns. I’m guessing the Nordic investors will recognize this one.

A big winner after the GFC, a mid-cap, 27-bagger.

A small cap, 128-bagger!

And finally, an 11-bagger nanocap that isn’t covered by Google Finance. That’s about 20% CAGR for about 15 years if you were wondering.

Any luck?

Please like and restack if you find it useful. Comment to start a discussion.

Subscribe now

Markets are shaped by group dynamics, not individual psychology.

Mauboussin:

A lot of the literature on behavioral finance deals with individual mistakes that you and I make... The nature of collective behavior is different from individual behavior... The aggregation of even suboptimal individuals can lead to optimal results. Part of the point is that we’re not really talking about psychology when we think about markets. We’re talking about sociology, which is how groups behave in a group setting.

This connects to the Wisdom of Crowds theory and how it applies to market efficiency. The core idea is that even if individuals have biases or make mistakes, markets can still be efficient if three conditions are met.

1. Diverse perspectives.

Firstly, ...diverse points of view. Some people are optimistic; some are pessimistic. There are technical traders, fundamental traders, short-term, long-term, etc.

2. A properly functioning aggregation mechanism.

The second condition is that there’s a properly functioning aggregation mechanism. The information that is out in the world is reflected, in this case in particular, in prices. Exchanges do that very well, but I would note that sometimes aggregation falls down because people simply do not participate.

3. Incentives for accuracy.

The third is incentives, which are rewards for being right and penalties for being wrong.

So when do markets get interesting? When one or more of those conditions break down.

Mauboussin says diversity is the one that breaks most often. And when it does, humans go to extremes and create bubbles:

By far, the most likely to be violated is diversity. Rather than each of us operating independently or with our own views, we correlate our behaviors. Humans are social, and investing is inherently a social exercise, so from time to time, it happens that people’s views collapse upon one another, and people all start to believe the same thing.

Here’s the thing. There’s a quote from Seth Klarman that Mauboussin loves, and I think it captures the playbook perfectly:

Value investing is, at its core, the marriage of a contrarian streak and a calculator.

Two parts to that. First, the contrarian streak:

The contrarian streak says when others are bullish or bearish, examine the other side of the case. That need not be a simple way to make money, because sometimes the consensus is correct, and by the way, positive feedback is something that we see a lot in nature in order to survive.

Being contrarian for its own sake isn’t enough. You also need the calculator:

So it’s not just being contrarian... Because everyone is uniformly bullish or uniformly bearish, that’s led us to a set of expectations that are unduly high or unduly low, and that company or that industry simply cannot satisfy that set of expectations that’s priced into that particular stock. As a consequence, there’s going to be a reversal... It seems like everybody has a uniform point of view, which seems to be what’s expressed in the security price, and through an expectations approach, I’m going to reverse engineer and say it’s a good bet to be on the other side of the argument.

So what does all of this mean? Herd behavior leads to bubbles and crashes because everyone starts believing the same story and reinforcing price movements in one direction. Maybe this applies to the current AI narrative. Only time will tell. Because a wise man said:

“The learned fool writes his nonsense in better language than the unlearned, but still ‘tis nonsense.” — Ben Franklin

I might end up being that fool. And I might not even be learned.

You can read and watch the full interview here.

Subscribe now

Yeah. I had to read that three times too.

Turns out there’s a scientific name for this. The Dunning-Kruger Effect.

Let’s take a closer look.

Stage 1: Peak of Mount Stupid

You bought Nvidia at $12. You’re up 10x and you’re telling everyone at Thanksgiving dinner about your “strategy.” Your uncle asks what a P/E ratio is and you say “doesn’t matter, just buy what you know.” You’re basically Warren Buffett. Except Warren spent 70 years reading annual reports and you spent 70 minutes on Reddit.

Of course I know everything!

Stage 2: The Oh No

Six months later your portfolio looks like a crime scene. You bought Peloton at $140 because “everyone’s working out at home now.” You loaded up on some crypto coin your coworker recommended at lunch. Your buddy who bought an S&P index fund is beating you by 40 points.

Oh snap. This is hard.

You start Googling “Is investing just gambling?” at 2am. Your confidence is in the basement. You’re ready to sell everything and put it in a savings account earning 0.5%. Your spouse is giving you that look.

Stage 3: I May Never Understand This

Rock bottom. You’ve lost real money. Not paper losses. Real money. The kind where you open the brokerage app and physically wince. You’re convinced the market is rigged and only insiders win.

But something weird happens here. You start reading. Not stock tips on Twitter. Actual books. Maybe it’s Peter Lynch. You stumble across the Substack FinStack community and realize there are people who actually think about this stuff systematically.

You can’t read and learn fast enough.

Stage 4: It’s Starting to Make Sense

You finally build a real portfolio. Not 47 random tickers you heard about on X. An actual diversified set of businesses you understand. You know what ROIC means. You know why free cash flow matters more than earnings. You start reading 10-Ks for fun (okay maybe not fun but you do it).

You stop checking your portfolio every 20 minutes. You start thinking in years instead of days. Your confidence is coming back but it’s a different kind of confidence. The kind built on competence instead of luck.

Stage 5: Sustainable Confidence

Was it easy? No. Complicated? Yes.

Your competence and confidence finally align. You understand your weaknesses. You know your circle of competence, and you actually stay inside it (most of the time). You have a process. You still make mistakes, but you know why you made them.

Not bad for a self-taught investor.

One important side note. This research gets completely butchered online. The original study didn’t just say “dumb people think they’re smart.” It showed that everyone misjudges their abilities because of blind spots in self-awareness. Skilled investors underestimate themselves too. They assume everyone else can also read a cash flow statement. (They can’t.)

Some researchers have even critiqued the study’s methodology and pointed out alternative explanations. So the most popular framework about people being wrong about their abilities might itself be wrong.

Mother of all ironies.

Subscribe now

I'd like to believe I'm somewhere between 4 and 5 right now. Where do you think you are in your investment journey?

First, the Numbers

132 companies. Seven-dimensional scoring framework. Here is what came out:

22 wide-moat downgrades (20 to narrow, 2 all the way to no moat). Adobe, Salesforce, ServiceNow, Oracle, Workday, Accenture, Infosys, ADP. Big names.

18 narrow-moat downgrades (to no moat). Sabre, FactSet, The Trade Desk.

2 upgrades to wide moat: Cloudflare and CrowdStrike. Both cybersecurity. That’s it. Two out of 132.

The full Morningstar report is free. Here is the link. 48 pages. Worth reading if you have the time. If you don’t, keep reading.

Where the Damage Hit

IT Services took the worst beating. Nine downgrades in one sector.

The logic is straightforward. These firms monetize billable hours, labor arbitrage, and simple workflow automation. AI takes a chunk out of all three. Accenture, Infosys, Tata, Wipro, Capgemini, EPAM, Endava, Globant all got cut.

Enterprise software was second. Salesforce, Workday, ServiceNow, Oracle, Adobe, OBIC. Common thread: application layer, per-seat pricing, AI can replicate the workflow.

Payroll processing got hit as a whole sector. ADP, Paychex, Paycom. Per-seat again. Plus the risk AI-native startups skip the incumbents entirely.

The resilient sectors tell you where the market is wrong:

Cybersecurity. More attack surface, more data, more demand. AI is a tailwind here, not a headwind.

Engineering and design software. 30+ years of proprietary data. Workflows with zero room for error. Certification barriers that take years to clear.

Financial exchanges and data providers. Regulated. Benchmark status. Transaction data that cannot be scraped.

Content and music rights. Cannot be replicated. Cannot be trained away.

Payment networks. Network effects, not code.

How about the moat types? Network effects beat switching costs in an AI world.

Network effects had the fewest downgrades by far. Switching costs had the most. That’s the single most important takeaway from the whole report.

Why? Network effects are not based on the product’s technology. They are based on the strength of the network. Technologies change. Networks persist. Payment networks, exchanges, travel platforms. AI does not solve the cold-start problem of building a network from scratch.

Switching costs? Different story. AI can automate the technical processes that created those switching costs in the first place. Data migrations, format conversions, system integrations. All getting easier. Close to half of the downgraded companies had switching-cost moats. Morningstar still cut them.

The 7 Dimensions

Morningstar built a seven-dimensional scoring system to evaluate AI impact on moats. Most AI frameworks I’ve seen are high-level. This one is practical. Each dimension has a “better positioned” and “worse positioned” side. You score your company on each.

I’ve been using it on my own holdings. Here’s the framework with the examples that matter.

1. Layer Positioning: Infrastructure or application?

This is the single most important dimension. Infrastructure companies (cloud platforms, chip design tools, cybersecurity, databases) get more business when AI scales. Application-layer companies (basic CRM, scheduling tools, productivity apps) get replaced by it.

Think about it this way. AI agents need infrastructure to run on. They don’t need a separate scheduling app.

Microsoft scores well because Azure IS the infrastructure. Salesforce scores poorly because CRM sits at the application layer where an AI agent can do most of the job. The tricky ones are hybrids. ServiceNow orchestrates IT workflows (infrastructure-ish) but monetizes per-seat (application-ish). That’s exactly why it was a 3-2 vote to downgrade. Close calls live in the hybrid zone.

2. Data Moat: Unique or scrape-able?

Not all data is equal. Cadence and Synopsys have 30+ years of semiconductor design data certified to specific foundry processes. You cannot scrape that. You cannot assemble it from public sources. Each new chip generation adds more proprietary data that competitors will never touch.

Compare that to a company whose data is mostly customer-owned or can be rebuilt from public datasets. If an AI model can train on freely available information and replicate what you sell, your data moat is thin.

The test is simple. Could a well-funded AI lab recreate your dataset in 2 years? If yes, score low. AppLovin passes this test. Its advertising optimizer trains on proprietary auction win/loss data from a vertically integrated supply and demand network. The Trade Desk does not. It orchestrates client data and collects fees on third-party data. Neither is proprietary. That’s why AppLovin kept its moat and TTD lost it.

3. AI Demand: Does AI grow or shrink the TAM?

Cybersecurity gets MORE demand from AI. More agents running 24/7 means a bigger attack surface. More endpoints. More data to secure. CrowdStrike and Palo Alto benefit every time someone spins up a new AI workload.

Chip design tools also win. AI chips are among the most complex semiconductors ever built. More complexity means more EDA software. The AI boom is a direct tailwind for Cadence and Synopsys.

On the other side: basic CRM, contact management, scheduling, payroll. If AI reduces headcount or automates routine coordination, demand for those products shrinks. Fewer employees means fewer seats, fewer paychecks, fewer licenses. Score low.

4. Pricing Model: Per-seat is the kill zone.

This one is almost binary. Per-transaction and usage-based pricing survive because they are headcount-agnostic. If AI agents generate more transactions, you make MORE money. Adyen and the payment networks win here.

Per-seat pricing is exposed. If one person with AI tools does the work of two, you just lost half your seats. Workday, Salesforce, the payroll companies. All per-seat. All downgraded.

The smart incumbents will try to shift to value-based or usage-based pricing. Some will pull it off. Most will take a margin hit in the transition. And here’s the thing: even if they manage the transition, if they were previously earning economic rents above what AI-priced work costs, there could be compression anyway.

5. Workflow Complexity: Dashboards die. Regulated multi-step workflows survive.

Simple tasks get automated first. If your software mostly generates reports or organizes basic data, an AI agent can do that now. Score low.

Complex multi-step workflows embedded in regulatory frameworks are a different story. Semiconductor design (one wrong transistor layout and the chip doesn’t work). Insurance underwriting (Guidewire’s P&C workflows touch dozens of regulatory regimes). Multi-country payroll tax compliance. These workflows have too many domain-specific edge cases for a generic AI to handle reliably.

Dassault Systemes scores well here. CATIA is the “system of truth” for physical products. In aerospace, the model IS the airplane before it gets built. Regulatory certification references the Dassault model as the authoritative source. Switching CAD in aerospace would require recertification across 50+ year aircraft lifecycles. Good luck disrupting that with a chatbot.

6. Switching Costs: Still real. Just weaker than before.

Almost every software company has switching costs. That was not enough. Close to half of the downgraded companies had switching-cost moats at one point. Morningstar still cut them.

AI can automate the technical processes that CREATED those switching costs. Data migrations, system integrations, format conversions. All getting easier. The switching costs that survive are non-technological. Regulatory barriers (can’t switch your EDA tool mid-certification cycle). Network effects where the switching cost is leaving the network, not moving the data. Multi-decade product model libraries with no migration path.

If your company’s stickiness comes from “it’s annoying to move your data,” that moat just got thinner. If the stickiness comes from “switching could crash a $200M chip design program,” you’re fine.

7. Competitive Position: Distribution still matters. Maybe more than ever.

Getting people to adopt your product was always hard. Executing a sales motion, managing enterprise relationships, scaling across an existing client base. AI does not solve any of that.

Incumbents with big installed bases and competent go-to-market teams have more than a fighting chance, especially Microsoft with distribution into every enterprise on the planet. That advantage persists even if the product itself faces competition.

The risk is fragmented markets where AI-native startups can gain traction with lower prices and better tech. Small companies with no switching costs and no distribution moat are vulnerable.

Running This On My Portfolio

I scored my own holdings against the framework.

A few things to note here. My largest position, 40%, held its moat.

None of my core holdings got downgraded. My other 5 are not even on the 132-company list. Morningstar did not flag them as AI-susceptible. Good.

The AI framework’s verdict: nothing to act on. No downgrades, no fair value cuts, no exposure to contested sectors. The portfolio passed the test.

Weighted portfolio score: 27.9 / 35 sits one tick under “very resilient” (28+ threshold). Here is the result:

Run This On Your Portfolio

I built a Claude skill that automates the scoring. Drop in a CSV or paste your holdings. It runs each name through the 7 dimensions, cross-references Morningstar’s moat ratings with external sources, flags overlap with the top picks list, and gives you a weighted portfolio score.

The skill file is free:

Ai Moat Framework

148KB ∙ PDF file

Download

Download

Here’s how to use it:

1. Open any AI chat (I used Claude Opus 4.6 Extended)

2. Upload the AI_Moat_Framework.pdf file

3. Paste your portfolio in the same message

4. Type: “Run the AI moat framework on my portfolio”

Did your portfolio beat my score of 28? Share in the comments.

Subscribe now

Here’s how I TRY to avoid stupid mistakes.

1. Invest in companies you already know.

Massive time saver. I ripped through Adobe’s Digital Media segment (learned some Photoshop back in school) in one sitting. The Digital Experience segment (enterprise marketing tools I’ve never touched; I’m an engineer) took me a while to wrap my head around.

You still have to read the filings. latest 10-K, 10-Q, transcripts. Non-negotiable. Search SEC EDGAR or google “ticker IR” (example: ADBE IR). But starting with a product you actually use gives you a head start before you open the 10-K.

2. Stick to widely covered names.

I focus on companies with analysts on them. Here’s why: those analysts have channel checks, industry contacts, unit-level data, and access to management. My model isn’t going to beat theirs. My job is to understand the why behind the model.

So I use the consensus. Koyfin has it for free.

3. Pull all the earnings surprises.

Koyfin has them for free. The companies I own almost never miss consensus, especially on full-year numbers (visually ad up quarters into years).

Below are quarterly beats and misses for ADBE sales and EPS.

I believe this data says more about business predictability than analyst skill. If a name beats or misses by 40% every other quarter, it’s not that the analysts are bad. It’s that the business is unpredictable or too cyclical.

4. Keep valuation simple.

Historical multiples with context. Reverse DCF. That’s it. It just has to be reasonable.

Get the historical multiple and the growth. Find the PEG by dividing the average P/E by the growth rate. Then multiply that P/E by the next 5 years’ consensus future growth rate.

For instance, ADBE traded at median NTM P/E of 31x since switching to subscription model while growing 26% annually. So, 31/26=1.2 PEG.

ADBE is projected to grow about 14% in the next 5 years while trading at 8.6 NTM P/E.

So, 1.2x14=16.8 (PEG times the next 5 years’ growth rate equals P/E). ADBE should be trading around 17x P/E with that growth based on the historical multiples and growth rate.

Its fair value would be NTM EPS times the fair multiple, so 25.93 x 17 = 440. That’s a double from $222 (07 July, 2026).

Koyfin gives away this data for free.

Here is a link to my free reverse DCF calculator.

It looks like Adobe needs to grow about 8% to return 15% annually in the next 10 years.

5. Start small. Add on conviction.

I open new positions at 1-3% of the portfolio. Conviction comes with time as I keep digging through filings and read what other investors think.

Just wrote about this in my last post.

6. Hold it.

Help yourself to hold it. How? It especially helps if you have a diversified portfolio (aka multiple small positions).

I keep even at 20% drawdowns (even though it hurts). I’m changing this attitude a little because a 20% drop requires a 25% gain to break even.

If the drop is company specific, then it’s definitely time to dig deeper.

How do you avoid mistakes? Drop it in the comments. I want to hear it.

Upgrade for $120/year:

• Every paid post: Quality Scores, Valuations, and Trades.

• The full 20-name portfolio and 250-name watchlist.

• The 10,000-stock dataset to find your own Compounders.

Subscribe now

Maybe you didn’t read the filings. You trusted a Substack writer who turned out to be wrong. Hey, could be me.

The stock drops 50%. It was a 10% position. I’m pretty sure it has happened to all of us.

That’s a mismatch between how much you knew and how much you bet.

What Honest Sizing Actually Looks Like

You can do shallow research and still be ok if you size accordingly. You can build a 50-stock portfolio on checklists and quick write-ups and do just fine (I did just that as a busy person). Diversification is the last free lunch in investing.

The mistake isn’t doing light research. The mistake is doing light research and then betting 10% of your portfolio on the outcome.

Think about what you actually know before you buy. Not what you meant to read. Not the tab you have open. What you actually know, right now, today.

Most positions fall into one of three buckets.

You read a tweet or a headline. Maybe you caught a tip on Reddit. Someone on Substack wrote something compelling. You skimmed it, and it sounded good. That’s it. Position size: 0% (okay, maybe a tiny tracker position below 1%). You don’t own this yet. You have a hunch, not a thesis.

You ran a checklist and read a couple of write-ups. You cross-referenced a bull case and a bear case. You have a rough sense of valuation. You understand what needs to go right. This is actually a legitimate basis for owning something. Position size: 2-3%. At 2% across 50 names, a 50% drawdown costs you 1% of your portfolio. That’s a mosquito bite. Sleep is fine at this level.

You read the 10-Ks, transcripts, slides, and proxies. You understand the unit economics. You’ve recognized the patterns in management decisions. You’ve understood the KPIs that matter. You’ve thought about what would break the thesis. This is when you’ve earned the right to size up. Position size: up to 10%.

The Averaging Down Trap

Here’s where mismatched sizing gets really dangerous.

Stock drops 20%. You didn’t know enough to hold with conviction, but you sized it at 8% anyway. Now it feels wrong to sell. So you average down. It drops another 20%. You average down again. Now it’s your largest position, and you still haven’t read the filings.

The antidote is sizing correctly at entry. If it’s a 2% position and it drops 50%, it’s not going to be difficult to hold. The math is manageable (2%x50%=1% drop in your portfolio), and you have room to be wrong. If it’s a 10% position and it drops 50%, averaging down is how small mistakes become portfolio-defining ones (5% drop from a single position).

My Take

So, here’s the rule. Write this down.

Read a tweet or a headline: the position is 0%.

Ran a checklist and read a couple of write-ups: 2-3%. This is actually fun. Diversified, low-stress, and you’ll sleep fine.

Read the filings and can explain the business to a 10-year-old: up to 10%.

If you are busy with family and work commitments and unable to do enough work before buying, most of your portfolio should be in the first two buckets.

What’s the biggest position you’ve held where you knew, deep down, you hadn’t done the work? Drop it below. I’ll share mine.

Upgrade for $120/year:

• Every paid post: Quality Scores, Valuations, and Trades.

• The full 20-name portfolio and 250-name watchlist.

• The 10,000-stock dataset to find your own Compounders.

Subscribe now

Sold CRM and ADBE due to a 40% position in CSU, a software rollup, and losing interest in small positions.

Added to BABA at $125 and AMZN at $213.

The S&P 500 is in its longest correction since 2022. The VIX closed above 30 last week. Bearish sentiment in the AAII poll hit 52%. That is above 96% of all historical readings.

Scary numbers. But here is the thing. Every single piece of data I am looking at says that the current level is the bottom. And if you are a long-term investor, the current market is one of the better entry points you have had in years.

Let me walk you through it.

Volatility Spikes = Opportunity

Now here is where it gets interesting. Because while the macro picture is ugly, the data on what happens after volatility spikes is overwhelmingly bullish for long-term investors.

When the VIX is in its top 10% of readings (above 28, which is where we are now), the S&P 500 averages a 22% return over the next year. That is double the average.

And 85% total returns in five years. Not bad.

It gets better. When the VIX closes above 40 or 50, the average one-year forward return is above 30%. We are not there yet. But a year ago, the VIX closed above 50 during the tariff tantrum. Every single time that has happened historically, the S&P 500 was positive one year later. Every time.

And this is potentially the bottom. That’s what the data is telling me. Every time the put volume on SPY was above 8 million since 2022, that was the exact market bottom. It happened on April 5th.

If that was not enough, every time the Daily Sentiment index dropped below 15, that was the bottom.

And everyone’s favorite, Fear and Green Index, just bounced back from Extreme Fear.

What I'm Doing

I see two things at the same time. A genuinely uncertain macro picture and the cheapest stock prices in years for the highest-quality businesses, but the data is telling me that this is the likely bottom for now.

I am using this selloff to high-grade my portfolio. I made four moves today. Two exits, two adds: rotating into the compounders I want to own for the longer term.

What I Sold

I sold Salesforce (CRM) and Adobe (ADBE) in full. While they were closer to 5% positions initially, they were 1–3% positions when I sold them due to fresh cash.

If ADBE maintains the current 11-12x P/E multiple, we can expect total returns of 45% or a 15% annual return going forward.

If we see a multiple expansion towards 20x, then the expected return would be 140% total or 38% annual in the next 3 years. I believe it is possible if they maintain their current results in the coming years. Not accelerating growth, but simply maintaining the status quo. You know, put the AI worries to rest.

And for CRM, it is 45% total and 14% annual with the current P/E multiple of 15 in 3 years.

If there is a multiple expansion to 20x, then the returns would be 85% total or 25% annual.

I think they are solid choices if one wants to pick a side in this SaaSacre.

Why did I sell?

Two reasons. First, the valuations are attractive, but I feel uncomfortable adding to them regardless while I have a 40% high-conviction position from SaaSacre. So, I see no point in keeping additional positions. Second, I’ll be honest, I kind of started losing interest because they were small positions. There is a good chance someone is going to make a killing holding those two in the next 3 years: 140% and 85% total returns if they rerate to 20x multiples.

What I Bought

I added to two of my high-conviction names. Made each 10% position after they had dropped to around 5% positions due to new cash.

I added to Alibaba and Amazon.

The recent selloff created a window to add to two businesses I’ve held conviction on for a while. BABA pulled back from 190ish to 120ish.

The business itself has been improving, with cloud growing 36% in the last quarter.

Regardless, I’ll just add this screenshot from FastGraphs. It shows the OCF growth in the next 3 years. I’m using OCF because EPS is noisy due to the ongoing quick commerce wars, and FCF is due to the ongoing cloud investments.

With a slight multiple expansion from the current 13x P/OCF to 16x, total return would be more than 100% and 20% annually in the next 3 years.

AMZN has been flat YTD. AWS keeps compounding; some analysts are expecting it to reaccelerate towards 35%.

It traded around 24x P/OCF while growing at 28% annually in the last 20 years.

It wouldn’t surprise me to see it re-rate to a higher multiple, say 20x from here, as it is projected to grow at 25% annually in the next 3 years, which is close to the historical 28%:

Even if it doesn’t rerate, our returns would be around 90% total or 25% annual due to growth. Of course, if we don’t see a multiple compression from here.

Latest Portfolio

Concentrated portfolio. The SGOV position is essentially dry powder, earning something while I wait for the right opportunities.

I sometimes think about going 100% CSU and calling it a day. I mean, it is a super-diversified business with a well-aligned management that is trading at an attractive valuation. Isn’t that the definition of investing: quality businesses with great capital allocators that are trading at a reasonable valuation?

Also, I’ve been thinking about building KSPI into a bigger position, 20-30%, but couldn’t bring myself to do it for now. This stock could do a 2x in 1-3 years if Turkey investment shows further improvement and they keep the dividend going forward.

I’ll keep digging into KSPI. You’ll be the first to know if I increase my position.

What are you doing in this environment? Holding, adding to, or sitting on cash? Let me know in the paid chat or through email.

If I told you a 27-year-old from Oklahoma moved to Moscow in 1994 and bought 35% of Russia’s Google for $5 million, you’d think I was making it up. If I told you that $5 million turned into a 450x return, you’d walk away. Now add a 98% stock market crash he survived, an arrest by Russia’s secret service, and a stint in the same prison where a man died under mysterious circumstances. Still made billions for his investors.

All true. And it gets weirder.

The Name on the Door

Before we get to Mike Calvey we need to talk about the name. Baring Vostok. “Vostok” is Russian for “east.” The “Baring” part? That comes from Barings Bank. One of the most storied financial institutions in history. And one of the most spectacular failures.

Barings Bank was founded in 1762. For over two centuries it was royalty. Literally. The Queen of England banked there. At the Congress of Vienna in 1815, five nations reportedly called Barings “the sixth great European power.” This was a bank that financed the Louisiana Purchase in 1803. The largest land deal in history. Doubled the size of the United States. It funded the purchase of Alaska from Russia. It bailed out the Bank of England in 1839 and the Banque de France in 1847.

Barings also had deep ties to Russia going back centuries. It served as the Russian government’s financial agent in London. Managed their bank accounts. Arranged payments. Issued bonds. So when Baring Asset Management backed a team of twenty-somethings to raise a PE fund focused on Russia in 1994, there was a poetic logic to it.

Mike Calvey was part of that team. The Barings brand gave them instant credibility. A bunch of kids in their mid-twenties with no track record don’t raise $160 million from the World Bank, EBRD, TIAA-CREF, and Fidelity without a four-hundred-year-old name on the door. Right?

The Day the Name Almost Died

One year later the Barings name became a punchline.

On February 26, 1995, Barings Bank was declared insolvent. Destroyed. A 233-year-old institution gone overnight. Sold to the Dutch bank ING for one pound. Not a typo. One pound.

The man responsible was Nick Leeson. A 28-year-old derivatives trader in Barings’ Singapore office. Leeson grew up in Watford, a suburb of London. His father was a plasterer. He didn’t go to university. He failed his math exams. And somehow he ended up controlling both the trading desk and the back office at Barings’ Singapore operation. Nobody was watching the guy writing and grading his own tests. (This would be funny if it didn’t destroy a 233-year-old bank.)

Leeson started hiding losses in a secret error account numbered 88888 in 1992. A new employee under his watch had lost a small amount. Rather than report it and risk his reputation, Leeson buried it. Then he kept trading. When he lost more he doubled down. Classic gambler’s spiral. By the end of 1992 the hidden losses were £2 million. By late 1993, £23 million. By the end of 1994, £208 million.

On January 17, 1995, an earthquake hit Kobe, Japan. The Nikkei index cratered. Leeson had bet heavily that it wouldn’t fall below 19,000 points. It blew right through. He kept doubling down. Bought more than 20,000 futures contracts worth $180,000 each. Each bet made the hole deeper.

On February 23 Leeson left a note that read “I’m sorry” and fled Singapore with his wife. Total losses: £827 million. More than twice the bank’s entire capital. The Bank of England tried and failed to organize a rescue. Barings was dead.

Leeson made it to Malaysia, then Thailand, then Germany. Arrested in Frankfurt. Extradited to Singapore. Pled guilty to fraud and forgery. Six and a half years in Changi Prison. He was diagnosed with colon cancer while locked up. He survived. Released in 1999 for good behavior.

He wrote an autobiography called Rogue Trader. In 1999 it became a movie starring Ewan McGregor and Anna Friel. The New York Times called the book “dreary” and described Leeson as “a young man very taken with himself.” He later got a psychology degree and hit the after-dinner speaking circuit. In 2019 he said “two events a month is enough to keep me in the manner to which I’m accustomed.” He still trades stocks. Only with his own money now.

Sounds crazy right? Not so fast. The story’s just getting started.

Meanwhile, in Moscow

While Barings was imploding in London, Mike Calvey was in Moscow building the fund that would become legendary.

Born in 1967 in Wisconsin. Raised in Oklahoma. Studied computer science and business at the University of Oklahoma. Got an MS in finance from the London School of Economics. Started at Salomon Brothers in New York in 1989 doing energy M&A. His ambition at the time? To become the governor of Oklahoma. Wall Street was supposed to be a two-year pit stop.

His boss left to join the newly created EBRD in London. Calvey followed. He started traveling to Russia in 1991 at age 24. Moscow in those days was the Wild East. A country that needed everything from a business perspective. He fell in love with the process of working alongside entrepreneurs building something from nothing.

When Barings collapsed ING swept in and acquired the parent. But the private equity arms eventually completed management buyouts. Calvey and his mostly Russian colleagues kept the Barings brand and kept building.

Here’s the thing. Calvey grew up in Oklahoma. The oil industry there was built around the carried interest concept. A geologist who believed in a discovery would find three investors to put up a third of the capital for a quarter of the equity. They called these “third for a quarter” deals. Plenty of bullshitters. But the ones with real substance created fortunes. And the ones who got wiped out? Almost always the ones who borrowed too much.

Calvey carried that lesson into everything he built. No debt at the fund level. Never borrow in dollars and invest in local currency. It sounds obvious but it’s the cardinal sin of emerging markets investing. It works in the short term and you always lose it in the long term.

The 98% Crash That Made Everything Possible

In 1998 disaster hit. Russia defaulted on its sovereign debt. The ruble devalued by 4x. The stock market collapsed by 98%. Not 20%. Not 50%. Ninety-eight percent.

Most investors would have written it off. Calvey didn’t. Baring Vostok carried no debt and held cash reserves. So he doubled down. Increased stakes in his best companies from minority to controlling positions. The cash he injected gave those companies a massive edge over their struggling peers.

That first fund could have been a total wipeout. It ended up returning 4x to investors. $160 million in, $620 million out. Not bad for a fund that lived through a 98% drawdown.

More importantly it made Calvey the last man standing. The only foreign PE fund raising and investing in Russia between 1999 and 2003. Monopoly access to the best deals and the best talent.

One of those deals was Yandex.

$5 Million to $2+ Billion

In 2000 Baring Vostok led a $5 million seed round in Yandex, the Russian search engine. That bought roughly 35% of the company. It was the only equity funding Yandex received before its IPO.

Google came knocking. First offering $100 million. That would have been a 20x return. Calvey’s team listened to Arkady Volozh, the founder, who believed the company was worth billions. They passed. Google came back a year later with $250 million. Other shareholders wanted to sell. Calvey exercised a veto right to block the deal.

Six years later Yandex went public at an $8 billion market cap. Later grew past $20 billion.

So, Baring Vostok’s $5 million stake ended up worth well over $2 billion. Some sources cite a return in excess of 800x on the original investment. On a blended fund-level basis the number that circulates is 450x. Either way it’s one of the greatest venture returns ever recorded.

The lesson? Warren Buffett’s concept of watering the flowers and cutting the weeds. Calvey learned early (through some painful exits he made too soon) that the real money in PE is made by holding your best companies longest. Many firms sell their winners first under pressure to generate DPI (distribution to paid-in capital) for LPs. They end up holding the mediocre stuff for years. I’ve made this exact mistake in my own portfolio in the form of adding to my losers.

After that 450x return everyone who had wanted to sell at $250 million developed amnesia. They were appearing on Bloomberg and CNBC boasting about being early investors in Yandex. But they were the ones who wanted to sell when Google tried to buy it.

The FSB, the Prison, and the Way Out

By 2019 Baring Vostok had roughly $10 billion invested in Russia. Five portfolio companies had exceeded $10 billion market caps. Mike Calvey had become a legend in the Russian market.

Then in February 2019 Calvey was arrested in Moscow on fraud charges.

The case was linked to a shareholder dispute at Vostochny Bank where Baring Vostok held a majority stake. The charges alleged embezzlement of 2.5 billion rubles (about $38 million). Calvey said the accusations were fabricated by Artem Avetisyan, a minority shareholder with FSB connections. Avetisyan stood to gain tens of millions if Calvey was convicted. And Calvey had been suing Avetisyan in London. And winning. The arrest happened right in the middle of that lawsuit.

He was thrown into Matrosskaya Tishina (”Seaman’s Silence Prison”). Same facility where lawyer Sergei Magnitsky died under mysterious circumstances. Same judge assigned to Calvey’s case.

Russia’s former finance minister called the arrest “a shock” that triggered $40 billion in capital outflows. The U.S. Ambassador boycotted Russia’s premier economic forum in protest. Both the Trump and Biden administrations said it was a commercial dispute that shouldn’t have been handled through criminal courts.

After months of detention and house arrest a settlement was reached. In August 2021 Calvey received a suspended sentence. He left Russia in January 2022 just before the invasion of Ukraine. In April 2024 his conviction was formally vacated.

Remarkably the Baring Vostok funds had their most profitable years during this entire ordeal. 2019, 2020, and 2021.

After Russia

When the war started everything changed. Holding Russian assets would toxify the firm’s international portfolio. The exit model for Russian companies on international exchanges was gone for good.

The firm split in two. Vostok Investments took the Russian assets (owned by the former Russian managers). Baring Ventures took the international portfolio: Kaspi (now on NASDAQ, $20 billion+ market cap), Revolut, BlaBlaCar, and a Mexican digital bank called Plata. Baring Ventures has no Russian shareholders. No Russian employees. No Russian assets.

Calvey is now Chairman of Baring Ventures managing roughly $5.5 billion. He’s not raising institutional funds anymore. He’d rather do fewer things better. Investing his own money alongside co-investors on a deal-by-deal basis. (There’s a Lenin quote he likes: “Better, fewer but better.” He disagrees with almost everything else Lenin said.)

His biggest bet after Russia is Plata. A digital bank in Mexico built by roughly 200 former Tinkoff specialists who relocated to Mexico City after the war. It took 18 months and 40-50 iterations on the risk model before unit economics worked. Now it’s scaling fast. Reached 3 million customers in about half the time it took Tinkoff.

And he’s still bullish on Kaspi. The company is expanding into Turkey. Calvey draws a parallel to MercadoLibre’s logistics investment several years ago. The market punished MELI for investing in last-mile infrastructure. Turned out to be exactly the right strategic move. He believes Kaspi’s Turkey expansion will look the same way in three years. That is a post for another day.

My Take

The Barings connection is almost too perfect as a parable. One arm of the Barings empire was destroyed by a 28-year-old with no oversight who hid his losses until they swallowed the whole bank. Another arm of the same empire, under a 27-year-old with no track record in a country that barely had functioning markets, generated one of the greatest investment returns in PE history.

Same name on the door. Wildly different outcomes. Risk management and knowing when to double down versus when to cut your losses and some luck. That’s the whole game. (I’m still learning.)

Have you heard of Mike Calvey before? Comment below.

Upgrade for $120/year:

• Every paid post: Quality Scores, Valuations, and Trades.

• The full 20-name portfolio and 250-name watchlist.

• The 10,000-stock dataset to find your own Compounders.

Subscribe now