The story somewhat echoes that of the Japanese equity market, which has seen two decades of stagnant returns, only to rise up and accelerate massively. The Nikkei 225-index is up 77% over the past twelve months, and international eyes are turning to the Tokyo Stock Exchange once again. Sanrio is one of the more interesting names on that list, and one that is severely undervalued.

Price chart 1: Nikkei 225 index, 1-year quote history

Company profile

Theme: Compounding, Direction: Buy

Symbol: 8136, Exchange: TYO
Sector: Consumer Discretionary, Industry: Leisure, Apparel & Entertainment Retail
Fair intrinsic value: ¥1 696 (89%), as of June 2, 2026
Market capitalization: ¥1 109 696 million
Pricing data: P/S 6.05x, P/E 14.95x

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

The explosive turnaround

Seldom does a chart paint a visual so vividly as the revenue history of Sanrio. A beloved brand with storied characters that frustratingly stood stagnant as a merchandise company until management recently engineered a revival. By bringing in new leadership and by centralizing brand management from a historically fragmented approach, Sanrio was able to have its second founding in 2020.

Figure 1: Historical revenues

Sanrio, famous for characters such as Hello Kitty, can be described as a character IP platform. Sanrio designs, licenses, and commercializes character-based IP across merchandise, retail, theme park experiences, and other emerging entertainment formats such as VR.

There are three main revenue segments, licensing being the most lucrative line of business. Royalties through licensing feature much more attractive margin profiles compared to traditional retail, are easily scalable, and do not carry inventory risk. Sanrio grants third parties the rights to feature Sanrio characters on merchandise such as toys, plushies, clothing, beauty products, stationery, and even food. It also grants the right to feature the characters in collaborations and various media content. Licensing and royalties currently account for about 42% of overall segment revenues.

Figure 2: Royalty and licensing revenue

Accounting for about half of overall revenues is product sales, which comprise direct sales of its merchandise through Sanrio retail stores, e-commerce, and outlets. A simple, easy-to-understand line of business that has found rapid growth, even though a mix shift to licensing is preferred from a margin and risk perspective.

Figure 2: Product revenue

The third main segment is the theme parks Harmonyland and Puroland, as well as other types of experiences. The theme parks feature attractions, parades, character interactions, and in-park merchandise. The revenue the theme parks represent as a percentage of total sales is less than 10%, but their existence is important in order to drive overall enthusiasm for the brand. The average number of visitors to the theme parks keeps increasing, and so does the spending per person who visits the theme parks. The theme parks, despite limited capacity, drive an increasing amount of traffic through seasonal events and limited-edition merchandise.

Figure 3: Theme park visitors and spending per visitor, TTM

The reason the business was able to turn around and accelerate has several key reasons. The company had grown too reliant on Hello Kitty, and there was no clear path forward to grow interest in the character roster that now spans over 450 different characters. The first step was to start monetizing its global IP strategy, which included aggressive pushes into markets outside of Japan and diversification in its licensing. Hello Kitty was joined by other characters such as Kuromi, My Melody, Cinnamoroll, and others, many of whom are now more beloved than Hello Kitty in key markets.

The push for a more diversified portfolio of characters was bolstered by digital marketing in order to appeal to younger fans and new demographics. That included leaning into YouTube, social media channels, games, and other channels frequented by younger demographics in order to gather feedback and reduce reliance on physical retail stores. In the most recent character ranking globally, Hello Kitty ranks fifth, with Pompompurin, Cinnamoroll, Pochacco, and Kuromi ahead, in order from first. In some key markets like the U.S., Hello Kitty is not even in the top 10 at the time of writing. Another key reason for the turnaround was an improved licensing structure, where Sanrio focused on acquiring higher-quality partners and pushed more character mix and category selection. Going for quality over quantity allowed for improved brand control, as well as monetization per license.

Today, Sanrio looks less like a toy and gift seller and resembles a global, multi-channel character operator. The value now stems from a growing, globally monetized IP flywheel with better reach, better licensing, broader character use, and tighter brand control. The flywheel is simple:

Character creation and brand building → Spread awareness through merchandise, stores, digital channels, collaborations, and theme parks → Scale the characters into many product categories and geographies through licensing → Higher visibility feeds fan engagement and character demand → Reinvest in digital content and further expansion, IP protection, and further character creation → Repeat the cycle.

Understanding the geographies and segments

While the revenue chart tells of an impressive turnaround narrative on its own, it is necessary to look deeper in order to grasp the dynamic fueling the growth. Most important of all is understanding why a mix shift towards licensing is substantially more lucrative for Sanrio. I already mentioned that licensing carries less risk and that it has a more attractive margin profile. However, much like the turnaround story, a chart vividly paints the picture of how much more attractive the licensing margins are.

Figure 4: Segmented domestic operating margins

Segmented operating margins in the 80-90% range are on par with the most oppressive, high-moat, and unfair businesses in the world.

The rest of the research report is for paid subscribers only.

Please consider becoming a paid subscriber to help me do this full-time. That would allow me to deliver more frequent as well as higher-quality research.

Read more

If investors are purchasing the SpaceX IPO, it is for the narrative, not for the numbers that the business has put up historically, or even reasonably will put up over the next 10 years. SpaceX may have unparalleled potential among any publicly traded company, but it also comes with severe risks, most of which are tied to the valuation.

Company profile

Theme: Frontier, Direction: Sell

Symbol: SPCX, Exchange: NASDAQ
Fair intrinsic value: $458.14 billion, as of May 29, 2026
Targeted IPO valuation: $1,750- $2,000 billion
Major underwriters: Goldman Sachs, J.P. Morgan, BofA Securities, Morgan Stanley, Citigroup
Implied pricing data at $2 trillion: P/S 107x, P/E N/A

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

All about the numbers

SpaceX operates three main business segments: space, connectivity, and AI, all of which are synergistic. Adopting X and its AI business at SpaceX enabled a powerful self-reinforcing loop, and it is all driven by the overarching goal of making human life multi-planetary.

Space and connectivity:

• Reusable rockets drastically reduce launch costs, enabling more frequent deployment of low-Earth-orbit (LEO) satellites at a fraction of competitors’ costs.

• Starlink (LEO) generates recurring, high-margin revenue that feeds directly back into capital expenditures for next-gen rockets and scale manufacturing.

AI:

• Space’s payload capacity and orbital refueling capacity will enable orbital data centers in the future, where solar power and natural cooling provide AI with a structural cost and energy advantage.

• AI is already present in rocket design optimization, autonomous landing and flight control, constellation management, and collision avoidance. Furthering SpaceX’s AI capabilities and Colossus compute accelerates SpaceX’s development cycles.

• Starlink’s network traffic routing, predictive maintenance, and spectrum allocation can be optimized using AI. X integrations can drive new consumer and enterprise use cases on the network.

In short, connectivity’s cash flow subsidizes both Space and AI. Space’s cost advantages make connectivity difficult to compete with. AI’s compute scale and data from X and Starlink create a new high-margin revenue stream that further funds the other segments. The more satellites that are launched, the more cash SpaceX generates, which means more AI build-out, which turns into smarter rockets, which makes satellites cheaper, and the cycle repeats.

Figure 1: Segmented revenue

The business flywheel is solid, and it is hard to root against a company with such an inspiring long-term goal, but is it worth paying for? Total revenue has grown 33-35% over the past two years, and at a targeted $2 trillion valuation, the implied price-to-sales ratio is 107 times, and that’s without a price-to-earnings ratio since the business is not profitable.

As most readers know, I don’t put a lot of weight on pricing metrics as I prefer to focus on intrinsic value, especially in companies that are still relatively early-stage in their growth story. However, the satellite launch vehicle market (SLV) is expected to compound at 13% over the next 10 years (Global Market Insights), and the LEO market is expected to compound at 17% over the same period (Goldman Sachs). That’s not a market that would justify excitement to pay ~100 times revenue for. AI, on the other hand, is hard to gauge, but third-party estimates place expectations at 44% CAGR over the next 10 years (Precedence Research), something to definitely be excited about. But it is also the segment with the most competition, and Grok is not in a pole position to capture a majority of it.

Space

None of the markets is clearly defined, and the total addressable market size can change rapidly per segment. What we can do as potential investors, however, is to gauge how the businesses are developing and how SpaceX is setting up to capture potential TAM.

Space revenues have only grown in the single digits over the past two years, but the business development has been significant. Two core operational metrics that SpaceX provides in its S-1 filing are the number of launches and the mass to orbit. Higher mass and more frequent launches mean more billable payload, as government contracts and commercial customers pay based on capability and reliability.

SpaceX had 170 launches in 2025, which accounts for about half of all global orbital launches. Number two is China as a whole, with about 90 launches, most of which were expendable rockets, and with lower mass per launch. Third place is Rocket Lab with 21 launches, with much smaller payloads. The launch cadence alone speaks of SpaceX’s clear market dominance, but it is also important to note that almost all competitors operate on expendable vehicles and have higher costs per delivery. The resuability advantage is massive, and Starship will further widen the gap.

Figure 2: Mass per launch and revenue per ton to orbit

Looking at the operational metrics, it is clear that SpaceX is becoming more efficient with its payloads, as they are able to bring more tons into orbit per launch. However, the revenue per ton to orbit is rapidly declining, which may seem alarming. Revenue per ton to orbit translates to pricing power per unit of capacity delivered, where a falling metric means that scale is beating price, which is actually the bull case. Capacity is scaling far faster than the revenue dollars, which is expected when launch costs are collapsing, and Starlink is the dominant payload, as they fly their own cargo for free. SpaceX is already dominant, and making use of the capabilities to accelerate Starlink’s growth will feed right back into the company’s multiplanetary ambitions as Starlink generates funds to advance space capabilities even faster. When the payload mix shifts more heavily towards contracted payload, the unit economics of the business will change rapidly.

Connectivity

Global connectivity has an increasingly broad application and market, one that is just now starting to take shape. SpaceX themselves estimate the total market opportunity to be $1.6 trillion across consumer use cases, and that is without accounting for connectivity evolving into a critical infrastructure layer for the global economy. Historically, Starlink has primarily been used to serve the 40% of the global population that lives in rural areas, where terrestrial broadband infrastructure is not developed. And while rural and underserved geographies have been the primary target for Starlink historically, that is now quickly changing as Starlink begins to compete materially in suburban and urban environments.

Consumer broadband opportunities are just one side of the coin, though, as enterprise solutions are quickly gaining traction as well. SpaceX’s connectivity offerings are used across a wide variety of industries, including construction, agriculture, retail, telecom and hospitality. For example, we have seen an increasing number of airlines adopt Starlink for their on-board WiFi, most recently American Airlines.

Applications across government use cases are also rapidly increasing, as demand for resilient, low-latency, and secure communications in contested and remote environments is increasing.

In a mere two years, Starlink subscribers have increased by 448%, and Q1 2026 connectivity revenue alone just about equals the revenue for the full 2023 fiscal year. Not only is the adoption and revenue ramping up quickly, but so too is the margin expansion in the segment. As mentioned earlier, Starlink is the main cash generator of the business, which is necessary to allow the advancement of the space and AI segments.

Figure 3: Starlink subscribers

Similar to revenue per ton to orbit, Starlink's average revenue per user (ARPU) is dropping rapidly. However, much like with revenue per ton to orbit, it is not alarming but by design. The metric has to be looked at from a broader perspective. While ARPU is dropping, revenue for the segment has tripled over the same period, and the subscriber base has exploded. Simultaneously, operating margins have more than tripled.

Figure 4: Starlink average revenue per user

It is a case of a classical market penetration strategy, as is the case with many now successful platforms like Netflix, Amazon, and even Tesla with its Roadster. Early adopters are premium customers who pay high rates, and as the service expands globally and across more diverse use cases, ARPU drops. In order to reach the full breadth of the market, Starlink has to move down-market, which means pricing the service competitively. A dropping ARPU would be alarming, but only if it happened without a corresponding subscriber growth. SpaceX is prioritizing market share, global coverage, and long-term network effects over shorter-term per-user metrics.

AI

The rest of the research report is for paid subscribers only.

Please consider becoming a paid subscriber to help me do this full-time. That would allow me to deliver more frequent as well as higher-quality research.

Read more

I will do something different in this coverage and attack Robinhood from a bear’s perspective, and still show why Robinhood is severely undervalued as a business.

Company profile

May 18, 2026 Follow-up coverage
Direction: Buy
Previous fair intrinsic value: $158.88, as of February 27, 2026

Symbol: HOOD, Exchange: NASDAQ
Sector: Financial Services, Industry: Capital Markets
Theme: Growth
Fair intrinsic value: $155.29 (101%), as of May 18, 2026
Market capitalization: $70 586 million
Pricing data: P/S 16x, P/E 38x

Previous coverage:

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

Take-rates explain everything

Most investors forget that cyclical businesses have volatile results until they are reminded of it. And when they are reminded, the stock chart does not look pretty. My forecasting for Robinhood accounts for the business’s cyclical nature, which is why the periods are normalized. I was criticized for expecting 27% Y/Y growth in 2026 when 2025 just posted 52%, almost cutting the subsequent year’s growth in half. That is the result of normalization, and I have done it for all the forecasted fiscal years. Robinhood is the type of business that will post negative growth in transaction-based revenues in one year (2023), and then grow 110% the subsequent year. That is the nature of the business, and as long as investors account for it, seeing near-misses won’t sway emotions.

Instead of starting with the top-line results like usual, let’s begin by looking at the building blocks for getting that top-line. Here is a summary of earnings expectations from Wall Street, and what Robinhood reported:

• Total revenue estimates of $1.14 billion, reported $1.07 (-6%).

• Net interest revenue estimates of $383 million, reported $359 (-6%).

• Transaction-based revenue estimates of $658 million, reported $623 (-5%).

Drilling into transaction-based specifically:

• Equities revenue estimates of $83, reported $82 (-1%).

• Options revenue estimates of $280 million, reported $260 (-7%).

• Cryptocurrencies’ revenue estimates of $153 million, reported $134 (-12%).

• Event contracts revenue & other revenue estimates of $122 million, reported $147 (+20%).

How are such large misses possible when Robinhood reports monthly results? The reason is that Robinhood only reports volumes every month; therefore, take-rates can’t be derived.

Equities were basically in line with estimates, which is because equity markets are mature and yields from the volume are pretty stable. There was a relative yield drop of 3% Q/Q, which resulted in a 1% revenue miss. However, in order to capture that 3%, we would have to break out four decimals; that’s how mature equity markets are.

Figure 1: Equities yield

The options yield dropped by 7%, which is also what the miss in options revenue amounted to. On a more drastic note, cryptocurrencies’ yield dropped a substantial amount, by 24% Q/Q. Analysts seem to have accounted for a drop in yield, since crypto revenues missed by 12%, but not enough of a drop. That is because ever since Robinhood introduced Bitstamp, yields have more than halved from the peaks of 0.55% in Q1 2025 and continue to dwindle. The bottom is unpredictable and difficult to forecast, which is likely why the discrepancy from analyst expectations was so big in Q1 2026.

If we apply Q4 2025 yields to Q1 2026 volumes, we would have options revenue in line, equity revenue in line, and a big 15% beat in cryptocurrencies. More importantly, transaction-based revenues would beat analyst expectations.

Figure 2: Cryptocurrency yield

The take rates (yields) were directly addressed by management during the Q1 2026 earnings conference call.

The CFO’s comments make it appear that gaining market share is a good trade-off for the drops in take-rate. In addition, it is implied that there’s more active trading occurring on the platform. However, Robinhood lost market share in three out of four trading segments, only gaining market share in cryptocurrencies, while losing in equities, options, and event contracts.

Figure 3: Trading market share summary

The CFO’s comments also imply more active trading, but the quarter sees fewer cryptocurrency daily active revenue trades (DARTs) Q/Q, fewer equity DARTs, and static options DARTs. When looking at the average trade sizes for cryptocurrencies and equities, it is roughly static compared to Q4. This means that the comments about winning market share and having more active trading on the platform are not supported by the publicly filed data.

Figure 4: Segmented average trade sizes

In general, I would agree that take-rates (yields) are a good trade-off in return for sharp gains in market share, but that did not happen this quarter to explain the drops across the board.

Turning to net interest revenue, a similar dynamic can be observed. Monthly data disclosures don’t include broken-out assets so that investors can measure how much of AUC is interest- and fee-bearing, but it does include total AUC. Interest-earning assets as a percentage of total AUC have been roughly the same over the trailing twelve months, both in Q4 2025 and Q1 2026, so analysts would have been more accurate assuming the yield was roughly in line. Both Q3 2025 and Q4 2025 had ~3% in annualized yield on interest-earning assets, a figure that dropped to 2.3% in Q1 2026.

Figure 5: Segmented interest yield

Assuming the same yield on interest-earning assets as in Q3 and Q4 of 2025, Q1 2026 net interest revenues would have come in at $457 million, beating expectations by 20%. Summing up transaction-based revenues and net interest revenues based on Q4 yields would have left Robinhood beating earnings expectations by 5% on total revenue.

Business is thriving

The rest of the research report is for paid subscribers only.

Please consider becoming a paid subscriber to help me do this full-time. That would allow me to deliver more frequent as well as higher-quality research.

Read more

The intrinsic value growth that this business has been exhibiting over the past twelve months is nothing short of extraordinary. Not only because Palantir keeps accelerating its pace of growth and expanding its margins, but because of how durable the growth is. This is not short-term; this business will accelerate for a long time into the future.

For the observant investors, the interest in Palantir is skyrocketing, and it’s not only sourced from Palantir’s own commentary. CEO Alex Karp and other executives have been saying that the volume of inbound they are receiving is more than they can keep up with for several quarters now, and that was reiterated once again.

To double-check the interest, I ran a screen on the word “Palantir” across all publicly traded companies in the U.S., and the results surprised me. There are 647 separate instances of Palantir being mentioned across different companies’ investor relations. The customer testimonials are exactly what Palantir has been saying; they are changing the unit economics of businesses at scale, and as Palantir’s numbers also suggest, the customers can’t get enough of Palantir.

The reason I highlight this particular customer is that the CEO was an AI skeptic. He does not follow hype on social media or delve into AI thoroughly. He thought that AI was a mere chatbot, with no tangible benefits to his business.

I realize that what I read and what we see talked about in the AI world and what Palantir’s operating system actually is, I can barely recognize the reality of what they’re doing on the ground with the talk that goes on around AI.

[…]you’re talking to a guy who’s not on social media, doesn’t know anything about it, doesn’t care to know anything about it. I was full grown when that came out. I skipped all of that. I thought AI and the chatbot and having conversations with an AI system was of the same ilk.

What I see, though, is the reality that Palantir creates a walled garden, if you will, where every piece of data in our network across all the systems and all of the handoffs and all of the spreadsheets and all of the memos and the hundreds of hours a week that we spend as individuals trying to explain and connect information from one system to another to another to then even be able to guess what our profitability is, let alone audit it.

Palantir’s Foundry system contains all of that information and drains the need for all of those systems and all of that activity. We’re talking tens of millions of dollars of benefit over the next three to five years annually in a business our size at only a $1.3 billion run rate.

Company profile

9 May, 2026 Follow-up coverage
Direction: Buy
Previous fair intrinsic value: $204.45, as of February 5, 2026

Symbol: PLTR, Exchange: NASDAQ
Sector: Technology, Industry: Software - Infrastructure
Theme: AI Software
Fair intrinsic value: $268.5 (94.85%), as of May 9, 2026
Market capitalization: $354 273 million
Pricing data: P/S 80x, P/E 220x

Previous coverage:

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

Note: Type-F sells an aggregated Excel sheet of Palantir’s historical financial data, containing ~1000 rows of both company-reported metrics, but also calculated KPIs, contract data, and more. The sheet also includes my valuation model, a charting tool (internal in Excel and on the web), and a dashboard overview of the business.

Available at: https://typefcapital.com, with free quarterly sheet updates.

Growth is exceeding everyone’s expectations

Palantir has once again exceeded Wall Street’s expectations, my own expectations, and their own guidance in terms of topline growth. Q1 2026 saw 85% Y/Y revenue growth, but as always with Palantir, the mere number is not the impressive part; it’s the composition of the number that tells the whole narrative.

Figure 1: Segmented revenue

When a company is exhibiting accelerating growth to an extent that it becomes exponential in nature, the natural question is to ask “how long can this be sustained?” That is why I find it important to understand the pillars supporting the growth that the company is experiencing, which also allows investors to understand the scalability of the business. Looking at where the growth came from on a cohort basis, 50% of the overall growth came from existing customers, and 35% came from new customers. That means, even if Palantir saw no new customers in the past year, they still would have grown 50% in Q1 2026 purely through their existing customer base from a year ago.

Figure 2: Existing customer and new customer cohort growth

The accelerating trend in how much existing customers are spending on Palantir software is truly a marvel to behold, but it does not end there. Palnow now has over 1000 customers, and 2% of those (top 20) account for 41% of overall revenues. Over the trailing twelve months, the top 20 customers average $108 million in revenue per customer, while the average customer outside of the top 20 averages $3.1 million. The top 20 customers’ average spend is 3500% higher than the average of customers outside the top 20.

What’s astonishing is that these massive companies that are already spending huge amounts on Palantir software are actually growing their spending faster than the overall average of existing customers. The companies that already utilize Palantir on a large scale are still extracting so much value from Palantir that they find it necessary to increase their spending at an increasingly accelerating pace.

Figure 3: Top 20 customer cohort revenue growth (TTM), and percent of total revenue (TTM)

Those 3 charts are really all you need to tell the narrative of how Palantir is the only company in the enterprise software world that is actually delivering real, tangible value from its AI platform to its customers. The customers, who are already spending millions of dollars annually on Palantir software, are rapidly accelerating their pace of spending because they see how much value they are extracting from utilizing Palantir. In addition, the ones who are spending the most are the ones who want to also increase their spending the most.

While contract configuration seems to vary, we know from historical research that AIP has usage-based pricing. Meaning, implementing AIP to solve only a single use case may not yield a large increase in contract value, but those expanding AIP to power most of their business could yield a contract value increase of 2-3x, without a real ceiling. The more you employ Palantir, the more value you extract from it, and the more you want to keep implementing it. The reason Palantir can charge over $100 million annually for its software is that it generates more than $100 for the customers who are using it; meaning, if Palantir can save your business a hypothetical $100 billion in annual spending, you wouldn’t mind paying $50 billion per year for that software.

The scalable nature of the software is a core part of the sustainability of Palantir’s growth, but the more you dig, the more convincing it becomes. Existing customer cohorts and their accelerated growth are impressive, but there are many more financial results that point toward the sustainability and even acceleration of the growth that Palantir is seeing, for a long time into the future.

One of the more shocking comments from Palantir’s Q1 earnings call was Alex Karp mentioning that he will drive the company to double U.S. revenue in 2026 and also extend that to 2027. Two years of consecutive 100% growth in U.S. revenue, which is ~80% of total revenue, raises the intrinsic value of the business massively.

The comment can be analyzed from multiple angles. First, Palantir has continuously implied that they do not have the capacity to handle all the inbound demand.

So the question then becomes: why isn’t Palantir increasing its sales force to meet customer demand? The answer to why Palantir is intentionally not meeting demand, that I can think of, is quite attractive for Palantir.

The rest of the research report is for paid subscribers only.

Please consider becoming a paid subscriber to help me do this full-time. That would allow me to deliver more frequent as well as higher-quality research.

Read more

Near as we can tell, nobody’s paying for VantageScores. The bureaus send along the VantageScore for free when someone buys a FICO Score. You know, when you see the big VantageScore score volumes that VantageScore talks about, you should know that they’re largely unpaid for. You know, are they? Is anyone using them? Don’t know. Is anyone paying for them? Our sense is not much. You know, it’s pretty hard to triangulate on what their market share is. I mean, I think it’s trivial, is what I would say. I think you see that in our numbers, right? I mean, if we were losing market share, you’d see it in our numbers, and you don’t see any of that.

William Lansing, Chief Executive Officer
Fair Isaac Corporation, Q2 2026 Earnings Conference Call

CEO William Lansing reiterated what he said back in 2024: VantageScore is not a competitive threat. VantageScore is not taking any market share from FICO, and the large volumes are credited to the bureaus shipping a free VantageScore with each FICO pull. Even though VantageScore is being offered for free, no one is switching over.

As Lansing said in the Q2 earnings conference call, if FICO was losing market share, you’d see it in the numbers. What the numbers showed was a 60% Y/Y growth in Scores. The market has not yet woken up to the fact that FICO is undisruptible, presenting potential investors with a very attractive opportunity.

Company profile

May 2, 2026 Follow-up coverage
Direction: Buy
Previous fair intrinsic value: $2051.6, as of March 31, 2026

Symbol: FICO, Exchange: NYSE
Sector: Technology, Industry: Software - Application
Theme: High quality
Fair intrinsic value: $2116 (104%), as of May 2, 2026
Market capitalization: $25 433 million
Pricing data: P/S 11.3x, P/E 33.5x

Previous coverage:

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

The best business model in the world continues to dominate

In my previous research coverage of FICO, I detailed how FICO’s stock collapse is more or less fabricated. The FHFA Director and, most recently, Senator Hawley have both targeted FICO for its pricing power. The claim is that FICO’s price increases have been hurting borrowers. The reality of the situation is that the total production expense per loan started increasing sharply before FICO’s price hikes. After increasing the price to pull a score by 725%, the TPE/loan remained static compared to when FICO began increasing prices, proving that FICO scores have no meaningful impact on the borrower’s costs. A FICO pull makes up ~0.03% of TPE.

The real reason is probably that the optics are bad for FICO. A cold housing market needs a scapegoat, and pointing at FICO, which is exercising its pricing power, is the simplest approach to avoiding scrutiny in areas that are actually causing housing to be unaffordable. However, before FICO’s price increases, the bureaus and resellers had a 1844% markup, and after a 725% increase, the markup is still ~650%. FICO is, in my opinion, fully justified in increasing prices, as the muddied line item that traditionally comes with a FICO pull does not break down the excessive margin that bureaus are charging. FICO’s direct licensing program makes the process a lot more transparent for lenders, while also making it more difficult for others to charge excessive markups. For more details, see my previous piece on FICO.

To combat FICO, FHFA has allowed GSEs to use VantageScore, a score that the industry has not switched over to, despite it being sent along for free with each FICO score. As Lansing said in various earnings calls (including Q2 2026), VantageScore’s volumes are fabricated, and in reality, VantageScore has ~2% market share or even less. The whole industry speaks FICO, and even free is too high a cost to make the switch to VantageScore. However, the market still views VantageScore as a real threat, despite FICO reporting 60% Y/Y growth in the segment.

Figure 1: Score revenue and score revenue growth

FICO, throughout the noise, has managed to find a way to increase pricing even further, while being perceived more lightly. The old performance model in the mortgage direct licensing program was priced at $4.95 per score, with a funding fee of $33. The new performance model is $0.99 per score, with a $65 funding fee, a 74% increase, while also addressing a point raised by Senator Hawley in his letter:

We already know from my last piece that the prices of FICO pulls are immaterial, but this way, the perception is improved, which should help with scrutiny over FICO’s pricing power.

Looking at mortgage originations revenue growth, we saw another massive quarter, growing 127% Y/Y. That is another nail in the coffin against the argument that FICO’s price increases are actively hurting the mortgage market. With the performance model reducing average per-score fees by 50% or 90% (for classic FICO and FICO Score 10T, respectively), and the per-score model being the same price on average as in 2025, FICO was confident enough to increase the revenue guidance for the fiscal year to $2.45 billion from $2.33 billion.

Figure 2: Mortgage originations revenue growth

The direct license program is not widely available yet, as evidenced by the increase in bureaus as a % of overall revenue and from management commentary. That figure should drop once the program is live and adoption increases, and it will also impact the timing of score revenue. The performance-based model would have a trailing funding fee compared to straight up-front costs, which would hurt the score’s growth in the initial phases (very short term), but would catch up as the funding fees start being received.

Figure 3: Credit bureau revenue as a percentage of total revenue

Scores are not the only exciting part of the business anymore, as software had a fantastic quarter as well. Scores grew 60% Y/Y in Q2, while software grew 7% Y/Y. Not many people would get excited at such a headline, but those are the same investors who don’t look deeper at the dynamic unfolding.

Figure 4: Segmented revenue

The dynamic that many investors unfamiliar with FICO may be missing is that a majority of FICO’s software revenue is legacy, and actively being phased out for a modern, cloud-based platform solution. The mix is quickly shifting towards the modern platform, a segment that is seeing impressive scalability (akin to Palantir’s). Platform revenue grew 54% Y/Y, which, while not 60% like scores, is completely organic revenue. It is important for FICO as a business to have organic growth, as scores are cyclical in nature. Non-platform is losing revenue each passing quarter, acting as a drag on overall software revenue. However, as the mix shifts, the software segment will increasingly become harder and harder to ignore.

Figure 5: Segmented software growth

A year ago, platform revenue made up ~30% of overall software revenue, a figure that is now 44% as of Q2 2026, rapidly approaching that 50% majority line. What’s impressive is not the mere top-line growth figure, but the expansion among existing customers. The dollar-based net retention rate (DBNRR) increased to 36%, meaning that a majority of the growth in the segment is attributed to existing customers increasing their usage of the service. This is a dynamic similar to what Palantir is seeing, where it is important to simply land inside a business, to then showcase value, after which existing customers increase their spending. I am not expecting the segment to grow at the rapid pace Palantir is, but the increasing DBNRR is showcasing the value of the analytics and decision-making platform.

Figure 6: Segmented software dollar-based net retention rate

The rest of the research report is for paid subscribers only.

Please consider becoming a paid subscriber to help me do this full-time. That would allow me to deliver more frequent as well as higher-quality research.

Read more

Read more

Company profile

Theme: Flywheel, Direction: Buy

Symbol: AMZN, Exchange: NASDAQ
Sector: Consumer Discretionary, Industry: Internet Retail
Fair intrinsic value: $467.42 (88.8%), as of April 20, 2026
Market capitalization: $2 680 549 million
Pricing data: P/S 3.7x, P/E 34.5x

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

Approaching a $ trillion in annual revenue, but it’s still early

Amazon’s ability to enter seemingly any vertical is impressive. It is not clear if Amazon will always emerge as a winner in any given vertical, but it is almost certain that everyone else loses. Over the course of Amazon’s history, shareholders have gotten used to Amazon bleeding capital, staying patient, and eventually achieving market dominance in the vertical. Today’s Amazon is a multifaceted powerhouse, the company that generates the most revenue in the world, and a company that is actively competing in the world’s most important race: the AI race.

Despite approaching a trillion dollars in annual revenue, Amazon remains in the early innings of several secular shifts, including AI, e-commerce, and grocery digitization, cloud migration, automation, and robotics. It is hard to imagine a company of this scale continuing to grow in the double digits, but all signs point towards such an outcome.

Figure 1: Segmented revenue

The core to being able to continue compounding growth is to have business segments that complement each other in order to create a robust flywheel effect. Amazon has managed to do that in each of its lines of business, tying the whole business together and empowering each segment into an interconnected ecosystem where the high-margin segments subsidize and empower the low-margin segments.

The original Virtuous Cycle (retail flywheel)

The retail flywheel, as famously sketched by founder Jeff Bezos on a napkin in 2001, is Amazon’s core self-reinforcing growth engine. The focus is on scale, efficiency, and driving customer obsession.

• Offering the lowest prices, the broadest selection, and faster delivery leads to a better customer experience.

• A good customer experience drives traffic, purchases, and loyalty to the platform.

• Third-party sellers are attracted because they want access to the traffic.

• More sellers mean a wider selection of products, further elevating the customer experience.

• As the cycle spins, Amazon achieves a massive scale, leading to lower economies of scale in logistics and increased purchasing power.

• Reinvest the savings into a better customer experience and repeat the cycle.

Low prices and better selection → Increased traffic → More sellers → Expanded selection → Cost-structure loop → Low prices and better selection.

The flywheel is evident when looking at retail sales over a long period of time. The operating margins are seemingly kept low on purpose while compounding scale. The operating margins seem like they are improving as of 2018 and onward, but that is not attributed to the retail sales; it’s attributed to the Amazon ads platform becoming material. While still relatively tiny compared to overall retail revenue, as of 2017, ads contributed more to operating profit than core product sales, according to Amazon. Ads’ operating margins are not broken out, even though revenue now is (as of 2022), but it is likely the reason retail operating margins are growing.

Figure 2: Retail revenue and retail operating margin, the virtuous cycle

Working the math, I still assume that the retail segment operates on negative margins, and I’ll break it down further in this article. With time, due to the sheer scale, robotics and automation through AI will turn a 30-year flywheel from a structural loss-maker to a mid-single-digit margin business. That may not sound like a lot, but at a run-rate that is approaching $1 trillion, each basis point increase in margin is massive. That is all while keeping the flywheel intact, with continuous scale, and while building out more profitable segments.

The Infrastructure Engine (AI and cloud flywheel)

On the AI end, Amazon’s strategy is a vertically integrated loop where each layer feeds the next, which in turn creates a compounding cost and demand advantage. Much of the attention in AI has been directed towards the labs (Anthropic, OpenAI, etc.), or the pure infrastructure plays (neoclouds, energy companies), but the nearsightedness undermines the advantages that companies such as Amazon and Google have.

• Custom silicon (Trainium, Graviton, and Nitro) gives AWS better price-performance than merchant GPUs and CPUs. By changing the economics of AWS, margins are elevated.

• Lower inference costs make it more economical to deploy AI more broadly inside Amazon’s services.

• Broader AI deployment generates more data, producing training signals that other cloud providers can’t access.

• Improved training data improves cash generation, which is then redeployed (dwarfing most competitors’ spending).

• Using capital that was generated in the loop, Amazon can build the next generation of silicon and capacity.

Custom silicon → Better price & performance → Improved AI services (Bedrock, SageMaker) → Lower inference costs → AI application build-out → More proprietary data and usage → Cash flow funding capex for reinvestment → Next generation of custom silicon and capacity.

The flywheel is structurally different from most other AI companies, which tend to run a single loop, while Amazon runs all five. While Amazon has its own chips to improve AI economics, competitors have to pay Nvidia margins. While Amazon builds out its own hyperscale cloud, competitors have to rent compute. Amazon generates billions of dollars in cash flow from its AI ventures, while competitors burn venture capital. The massive proprietary data from retail, logistics chains, and Prime lead to more cash generation, while competitors scrape data. Amazon has large amounts of in-house demand that de-risks the infrastructure build, while competitors must find external demand.

Figure 3: AWS revenue and AWS growth

The Amazon super flywheel

By now, you may have noticed that the retail and AI flywheels share a lot of common characteristics, and that is because they should not be seen as separate systems. Both flywheels share customers, data, and capital, and each feeds into the other at various junctions throughout the flywheel. The retail side of the business generates signals (data) and demand, while the AI side generates margin and capability, accelerating one another. Amazon is not a pure commerce player or a pure AI player; instead, it manages to leverage both to be a dominant force in any segment.

• AWS capital generation funds the whole ecosystem.

  • Funds next-gen silicon, which lowers AWS costs and expands AWS margins.

  • Subsidizes the retail flywheel’s lower-price engine.

• Retail data trains and grounds the AI

  • Every Prime subscription, every query, every delivery decision, etc, creates a proprietary data training signal.

This segment actually has its own mini flywheel inside: Better signals → better Rufus, seller agents, ad tools (making retail smarter) → better recommendations, more efficient ads and fulfillment → better experience → better signals.

• Advertising serves as a bridge between businesses

  • Ads sit on top of retail, but have unit economics that even surpass AWS. It’s also powered by both retail intent data and AI optimization. It is perhaps the clearest example of the two flywheels producing something neither flywheel could produce alone. Retail serves the audience, AI provides targeting, and the cash that is generated funds both separate loops.

• Prime subscriptions and logistics provide the physical layer

  • Prime helps provide an identity (shopper, viewer, reader, AI-assistant user), while logistics (robots, delivery networks, infrastructure) is the moat in the physical world. AI alone can’t build either, but can optimize both.

Each piece of the puzzle is important and helps build a robust business that is not replicable. However, each of the flywheels on its own can be matched and replicated.

1. Walmart can match the retail scale.

2. Microsoft and Google can match the AI infrastructure.

3. Meta and Google can match the advertising segment.

4. FedEx and UPS can match logistics.

However, no competitor can run the synergistic flywheel across all four gears simultaneously, with the same customer base and with the same cash flow outlook. Investors are worried seeing $200 billion in capital expenditures being forecasted by management for 2026, but it is necessary to keep the integrity of the flywheel. If AWS slows down, the retail cross-subsidy would weaken. If retail slows, the AI segment would lose its data and demand edge. The capital expenditures being spent on AI infrastructure are necessary expenses to accelerate the flywheel, to grow scale, and improve unit economics.

Growth ahead

Despite the already massive scale, the growth opportunities for the business are quite vast across each segment.

Retail segment

Starting off with retail, as Andy Jassy mentioned, ~80% of commerce is still happening physically. There is a lot of growth ahead as economies move towards e-commerce, and of that pie, Amazon is still relatively tiny despite being the largest e-commerce market globally.

To calculate the market share, we have to first make some assumptions since Amazon does not report its gross merchandise value (GMV). Its first-party sales translate directly to GMV as Amazon owns the inventory and records the full transaction as revenue, but the third-party revenue is subject to a take rate, which is where we have to make assumptions in order to back out the GMV. The third-party seller services line item is revenue that Amazon collects on third-party sales, not GMV itself. I bundle referral fees, FBA fulfillment fees, FBA storage fees, and other miscellaneous line items such as ad placement.

To do so, first, we need to grab the third-party unit share, which is 62% for FY2025.

However, that calculation assumes that first-party and third-party sales have identical average unit prices, which is unlikely. Third-party selling prices are likely ~30-35% higher than first-party prices since first-party skews toward low average selling price household essentials (i.e., Amazon Basic), while third-party skews toward branded items and higher ASP goods. If we factor that in, the calculation changes (using 32% premium factor):

This roughly translates to a 30% take rate. Peeking over at guesses from industry analysts, Amazon’s third-party take rates are in the estimated ranges of 27-31% (lower results in higher GMV). That gives us a total of $843.2 in GMV for FY2025, and I use the assumption of 30% take rate for historical periods as well. Now that we can estimate Amazon’s GMV, we can measure it against the global e-commerce market and get a market share estimate.

Figure 4: Amazon’s global e-commerce market share

The rest of the research report is for paid subscribers only.

Please consider becoming a paid subscriber to help me do this full-time. That would allow me to deliver more frequent as well as higher-quality research.

Read more

dLocal operates a payments ecosystem in (primarily) emerging markets where traditional finance networks have low penetration (e.g., card networks such as Visa, Mastercard, etc.). dLocal handles hundreds of alternative payment methods (APMs), currency conversions, compliance infrastructure, fraud prevention, tax integrations, and settlement services. In essence, they eliminate the need for merchants to build local integrations in emerging markets, which are often highly complex and fragmented. The business is inherently sticky, as evidenced by its very impressive customer retention rate.

Company profile

Theme: EM Fintech, Direction: Buy

Symbol: DLO, Exchange: NASDAQ
Sector: Technology, Industry: Software - Infrastructure
Fair intrinsic value: $33.07 (152.02%), as of April 12, 2026
Market capitalization: $3 959 million
Pricing data: P/S 3.6x, P/E 20.1x

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

One man’s trash is another man’s treasure

 dLocal operates across many EMs, but to truly grasp how money flows through dLocal’s platform, it is best to visualize the payment volumes across a grid. One axis represents the geographies, while the other axis represents the direction of money. For the geographies, the payment volume is either cross-border or local-to-local. For the direction, money is either being acquired (pay-ins) or disbursed (pay-outs). These are not mutually exclusive, since each transaction is classified across both segments simultaneously. For example, a pay-in can be both cross-border and local-to-local.

• Cross-border

  • A transaction where dLocal collects funds in one currency and settles them into a different currency (and typically a different geography).

• Local-to-local

  • A transaction where dLocal collects and settles funds in the same currency.

The reason global merchants use dLocal in a geography where they have their own corporate entities is because of dLocal’s already present infrastructure. Despite having a local entity, having to integrate with various local payment methods, dealing with regulations, and obtaining licenses is a headache. Instead, they utilize dLocal’s “One API” platform, which is already built and operational.

Both segments are currently at a 50% split in terms of mix.

Figure 1: Geographically segmented TPV and TPV growth

• Pay-ins

  • dLocal’s merchant customers receive payment from their end customers.

• Pay-outs

  • dLocal disburses money on behalf of its merchant customers to the relevant business partners, suppliers, contractors, drivers, or customers.

Pay-ins have historically accounted for ~70% of the mix in the directional segment.

Figure 2: Directional segmented TPV and TPV growth

dLocal sees structural tailwinds and moats connected to local-to-local and pay-outs specifically, since those flows are harder for global players such as Stripe and Adyen to replicate without dLocal’s local licenses, APM network, and regulatory expertise. Shifting the mix towards that combination will result in stickier local flows and is a core part of dLocal’s “one dLocal” flywheel and differentiation in the EM space.

However, that is not the most profitable combination, which would be cross-border pay-ins. While management does not break out the specific take-rates tied to the TPV, they do provide brief commentary regarding positive drivers for net take-rates. Cross-border pay-ins benefit from the most monetizable variables across the TPV grid since it includes the inherently sticky nature of pay-ins, together with processing complicated FX spreads.

Local-to-local pay-outs, while over time will be a core driver, likely carry a lower margin since it is a focus segment. That does not mean “growth at all costs” where dLocal has to sacrifice margins extensively, but it means that they have to take a big enough market share among domestic rails, and also carry more processing and liquidity costs.

The payment volume itself is highly diversified across all parts of the economy, both across consumer categories and corporate categories. That means that TPV benefits from many parts of the economy growing in the EM’s, and the diversification means that there is less risk associated with the overall growth of the payment volumes. The only segment that has been actively growing its share across the TPV mix over the reportable periods is commerce, from 15% of overall TPV in 2022 to 34% in 2025.

Figure 3: Directional segmented TPV and TPV growth

If we look at what all the payment volume translates to in terms of revenue, it has been an impressive growth story, and a story that is not about to end anytime soon. Since 2019, revenue has grown at a compounded annual growth rate of 64%, which can be attributed to a variety of tailwinds and execution for the business.

Figure 4: Gross take rate and total revenue

Seeing the gross take rate decline over time may seem alarming and a sign of competitive pressures. However, in dLocal’s case, it is a sign of scale rather than competition, unlike what may be observed in global payment service providers like Stripe, Adyen, and PayPal. The reason cited by management is that they favor payment volume growth over raw take rates, as it would impact prospects.

For large and scaling customers, there are volume discounts. As volume scales, the large merchants negotiate lower pricing or move to volume-tiered pricing, which is also favorable for dLocal since a large majority of both TPV and revenue growth is from existing merchants. Similarly, as the TPV mix shifts more toward pay-outs, it will have an impact on the take rates and margins due to being inherently less monetizable than pay-ins. The tradeoff is that it comes with stickiness and volume, which is still a favorable proposition for dLocal, and one that management embraces.

However, the take rate compression should start to slow down significantly as the mix starts to incorporate more profitable services such as BNPL, stablecoins, tokenized cards, AI routing, and more. In summary, deeper merchant relationships drive stickiness, which drives more volume, and in turn, provides more scale advantages. Winning the infrastructure layer in EM’s at a massive scale is more important than unit economics at this stage. That becomes very evident when viewing the source of growth, which is predominantly driven by existing merchants.

Retention is a core driver and a risk

To get a sense of how important it is to nurture existing merchants and scale the relationship with them, we don’t need to look further than the revenue bridge between 2024 and 2025. 96% of the added revenue in the period was from existing merchants.

Figure 5: Revenue bridge (2024 - 2025)

Looking at the net revenue retention over the past years, we can see that an overwhelming majority of growth is attributed to existing customers. By breaking it down into two cohorts of existing and new customers, the chart tells a clear picture of the importance of nurturing relationships with existing customers in order to drive growth for the business.

Figure 6: Cohort revenue growth

Historically, dLocal released merchant cohort data, which allowed us to track the “land and expand” go-to-market strategy. The 2023 cohort grew its TPV by a massive 719% from 2023 to 2024, and the 2018 cohort has grown at a 120% CAGR. However, as impressive as 120% CAGR is, the 2020 cohort has grown at a 154% CAGR. When viewed through this lens, it makes a bit more sense that take rates are taking a back seat in favor of scaling the payment volumes and deepening merchant relationships (making them stickier).

Figure 7: Total payment volume by merchant cohort, in $ millions

The rest of the research report is for paid subscribers only.

Please consider becoming a paid subscriber to help me do this full-time. That would allow me to deliver more frequent as well as higher-quality research.

Read more

Read more

Read more

Read more

That is also the problem right now: FICO has too much pricing power in the eyes of regulators. I previously detailed FHFA Director Pulte’s comments about FICO being a core reason for the unsustainable housing market. This time, it is Senator Josh Hawley opening an investigation into FICO through a letter to FICO CEO William Lansing. The letter reads as a bullish investment case into the company, and cites many of the core reasons I am very excited about the company. However, much like with what Director Pulte brought forward, the concerns raised are misdirected.

Screenshot 1: Senator Hawley’s oversight letter to FICO, highlighted bullish remarks

Company profile

March 31, 2026 Follow-up coverage
Direction: Buy
Previous fair intrinsic value: $2081.66, as of January 12, 2026

Symbol: FICO, Exchange: NYSE
Sector: Technology, Industry: Software - Application
Theme: High quality
Fair intrinsic value: $2051.6 (95.33%), as of March 31, 2026
Market capitalization: $25 168 million
Pricing data: P/S 12.2x, P/E 38.3x

Previous coverage:

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

The best business model in the world continues to dominate

I typically start off my coverage of a company by breaking down the operations and the financials, but this time I want to deal with the elephant in the room first. Both Director Pulte and Senator Hawley both cite that FICO has raised the prices for its royalties per score pull significantly, and both cite concerns over how FICO’s pricing power may impact the consumer. Let’s break it down by looking at the number.

First, it’s important to understand that closing costs vary depending on the lender, the borrower, the state, and many other variables. However, what drives closing costs and the overall consumer impact is the cost of producing a loan. The cost to produce a loan must be recovered through a rate, a point of origination (direct up-front cost for the borrower), or a mix of both. For example, a $12,000 production expense could be charged to the borrower through $3,000 in closing fees, and the rest through a rate spread. Another lender might charge $0 in fees and price everything into the rate. In any case, the cost and more will be recouped for an issued loan. The Mortgage Bankers Association (MBA) tracks the production expense per loan, and we can use this as a relative measure to see if it correlates with the raise of FICO scores royalties to gauge the impact.

Figure 1: FICO Royalty per Score pull, total production expense per loan, cumulative increase %

We can see that the TPE/loan increased by 45% in 2022 compared to 2020, while FICO’s royalties remained at $0.60 per pull, a 0% increase in price. From 2022 to 2025, FICO increased the royalty per score pull by 725%, but the TPE/loan remains relatively static. By 2022, the TPE/loan had increased by 45% cumulatively from 2020, then 55%, 46%, and 47% for 2023, 2024, and 2025, respectively. The TPE/loan is not correlated to FICO’s price increases, and there is a good reason for that.

The FICO royalty makes up 0.04% of the total production expense per loan. In 2020, that was 0.01%, both equally non-meaningful to overall production expenses. However, Senator Hawley comments on this directly with the following:

If the concern is indeed about the average borrower feeling the pain from FICO’s price increases, and in turn, is hurting the mortgage market, then it is most definitely relevant to look at whether FICO’s royalties constitute a meaningful share of overall closing costs or not. It clearly does not, given the objective data, and the probe should be ended with that. However, I think Senator Hawley is correct to ask the question about FICO exercising monopolistic pricing power. We can dig into that data as well.

Is FICO right to raise its royalties per score pull?

I believe FICO is justified in its price increases, and it has definitely been underpriced historically. In fact, I still think it is underpriced, and they should increase the price even further. Why do I say that? Because the markup from bureaus and resellers over the past five years has been as high as 1844%. While FICO only charged $0.60 per score pull, or $1.80 on a tri-merge basis, the total estimated cost per pull was $35. That implies a full 1844% markup from bureaus and resellers. The mathematical model of the credit report is the integral and centerpiece, so why should it be a mere 5% of the overall cost per pull? The line item presented to banks and borrowers is a lump sum with bureau data fees and CRA fees, not allowing for transparency of who is making it out like a bandit.

Since 2020, FICO has increased its cost per score pull by 725%; however, its share of the estimated total cost per pull only increased from 5% to 14%. FICO, the core and integral part of a FICO score, is still by far the minority in terms of overall costs for a score pull.

Figure 2: Estimated costs per credit score pull, markup from bureaus and resellers

FICO launched the mortgage direct licensing program to cut out a lot of the markups associated with pulling a FICO credit score, and in turn, also provide a lot more transparency. Senator Hawley claims that FICO doubled its prices from 2025 to 2026, from ~$5 to $10, while not providing a meaningfully better product. While that is partially true, the $10 per score is, on average, the same price that resellers paid for FICO scores in 2025, which gives a clear picture of how muddied the bundled line item was.

The other pricing option for FICO scores, by going directly to lenders through resellers, is a performance model. In the performance model, the price per score remains the same at $4.95, with a $33 funding fee on closed loans. This directly combats the statement made by Senator Hawley in his letter that FICO is actively hurting first-time borrowers who typically have their credit scores pulled multiple times. This is effectively a 50% reduction in average per-score fees compared to what resellers paid for FICO scores in 2025.

I don’t see how FICO should be blamed for various types of lenders having to undergo multiple credit score pulls. If Senator Hawley wants to combat this, FICO, which only provides the mathematical scoring models, has no impact on the practice.

One final point that Senator Hawley raises is that FICO’s dominance is the result of a regulatory framework that required lenders selling to GSEs to exclusively use FICO scores for over three decades.

The framing Senator Hawley puts forward overstates the role of regulation. There was no federal statute, FHFA rule, or HUD regulation that mandated GSEs to adopt or exclusively use FICO scores. The GSEs themselves chose to incorporate FICO into their underwriting and delivery guidelines in 1995 because it was already widely used by lenders and seen as a gold standard. GSE guidelines are contractual and business requirements that lenders follow to sell loans, not regulations imposed by a government agency. In fact, in 2018, Congress passed the 2018 Credit Score Competition Act specifically to require FHFA to create a validation process for alternative credit score models. FICO was the standard through GSE policy, not through a binding regulatory mandate.

In summary, FICO score royalties have no impact on the production expense for loans. FICO exercising its pricing power does not hurt the end consumer in a meaningful way, as its overall royalty constitutes a negligible part of overall closing costs, even after a 725% increase. A FICO royalty only constitutes 0.04% on total production expenses in 2025, and of an estimated cost per pull on a tri-merge basis, the markup from bureaus and resellers is still 641% (down from 1844% pre-increases). These investigations into FICO scores will not help the mortgage market or borrowers. FICO is an easy target to go after, as it creates headlines with the pricing increases. However, the math behind the increases justifies them and does not hurt the overall mortgage market, and as such, I don’t see a scenario where FICO is meaningfully hurt over these probes.

I see FICO retaining complete monopolistic dominance as it is too deeply embedded in the finance ecosystem. For a more detailed breakdown on FICO’s monopoly, see my previous FICO coverage.

Continued dominance

The rest of the research report is for paid subscribers only.

Please consider becoming a paid subscriber to help me do this full-time. That would allow me to deliver more frequent as well as higher-quality research.

Read more

Download the PDF by clicking here.

Company profile

Theme: SuperApp, Direction: Buy

Symbol: GRAB Exchange: NASDAQ
Sector: Technology, Industry: Software - Application
Fair intrinsic value: $7.25 (+96.42%), as of March 24, 2026
Market capitalization: $15 520 million
Pricing data: P/S 4.61x, P/E 57.91x

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

Download the PDF above for a better viewing experience.

Read more

The stock has cratered from an already extremely cheap quote, the CEO has been fired, and uncertainty is the highest it has ever been. However, the stock is not the only thing that has cratered; so has the intrinsic value. Is there any reason remaining to own stock as of right now?

Company profile

March 12, 2026, Follow-up coverage
Direction: Hold
Previous fair intrinsic value: $145.90, as of November 2, 2025

Symbol: PYPL, Exchange: NASDAQ
Sector: Financials, Industry: Credit Services
Theme: Deep value
Fair intrinsic value: $68.73 (+52.66%), as of March 12, 2025
Market capitalization: $43 219 million
Pricing data: P/S 1.30x, P/E 8.26x

Previous coverage:

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

Uncertainty gets punished

On a surface level, PayPal was truly set up to capitalize on the agentic wave that we are seeing across the AI industry. They have partnerships with OpenAI, Google, Mastercard, Perplexity, and just about every other major player in the space in order to enable user purchases within AI platforms. In addition, they have the single greatest source of actual purchase data relating to e-commerce, so why has the ads business not been launched and been a success yet? The questions relating to execution have been piling on, and investors’ patience has seemingly run out.

In my coverage of PayPal, I have deemed it a deep value opportunity, since the stock quote implied that the stock would not only stop growing, but actually decline. Where the stock quote is currently trading, it implies about a -7% CAGR at 13% FCFF margins over the next 10 years. Such a narrative has not materialized, and PayPal, while not as undervalued as I initially thought, is still trading very undervalued. However, the undervaluation could be justified given that they are quickly decelerating the business in an overall industry outlook that is set to grow in the mid-teens.

Former CEO Alex Chriss took over a sinking ship, and famously mentioned that PayPal was going to “shock the world” in January of 2024, and with a solid track record at Intuit, many investors believed him. What followed was two years of no material progress towards the goals of capturing the AI commerce market, launching a successful and prominent ads business, or meaningful innovation.

What followed in the Q4 2025 results was a fired CEO and horrific guidance for the fiscal year 2026 of declining transaction margin dollars and EPS. PayPal pulled their revenue guidance several periods ago, and instead, they give us transaction margin dollars guidance as a way to track their path towards “profitable growth”. To reach profitable growth, they sacrificed the performance of core drivers to exit the race to the bottom that all payment solution providers are participating in (resulting in a decline in take rate for all participants). Now, they’re still losing take rate, but even worse, there’s no trade-off in rapid payment volume growth. It’s a lose-lose situation in that regard.

Searching for a positive note with a lantern

While PayPal didn’t make material progress towards becoming the premier commerce orchestrator as promised, not all is doom and gloom. As mentioned, the company still has growth, and is not showing indications of ~8% annual declines as the quote implies. However, the growth in revenue for FY2025 was a mere 4.3% Y/Y, while being participants in a digital payments market that grew ~20% Y/Y, and is expected to compound annually at a rate of ~20% over the coming 7-10 years.

Figure 1: Segmented revenues

The core segment, transaction revenues, only grew 3.3%. The other value-added services (OVAS) grew 14.2%, showing signs of life after a flat 2024. The 14.2% increase signals that the business is gaining traction across credit and interest products, as well as subscription and partnership fees.

Figure 2: Other value-added services revenues

All the segmented total payment volume segments grew more than 3.3%, which means that the take rate decreased for the transactions. The whole purpose of having an off-year to reposition the business was to stop the bleeding take rate at the expense of growth, but as mentioned, PayPal lost on both ends.

Figure 3: Take rates

The highest margin segment, branded checkout, only increased 4% on a currency-neutral basis for the year, and 1% in Q4 2025. Peer-to-peer (P2P) is accelerating as a result of Venmo’s growth, but the real pain is felt in the payment service provider (PSP) solution, which accounts for a majority of the TPV.

Figure 4: Segmented TPV growth rates

Venmo is one of the strongest financial services in the U.S., and is consistently at the top of most downloaded apps in the Apple and Android app stores. However, for a long time, it has not been properly monetized or given enough attention. One of the few developments under Alex Chriss’s management was improved substantially. Historically, even though Venmo was owned by PayPal, the two operated in silos without communication, essentially operating independently. That has changed, and Venmo is now starting to see innovation, growth, and monetization. However, even Venmo, while gaining traction as of late, is still stuck at where it was three years ago. In Q1 of 2023, Venmo made up 18% of total TPV. In Q4 of 2025, it makes up 18% of total TPV; essentially, there hasn’t been any progress.

Figure 5: Venmo TPV

Looking at the core focus for 2025, transaction margin dollars, we saw an increase of 6%. This is expected to decrease in 2026, per management guidance as of Q4 2025. On a positive note, the transaction margins have been increasing for two years in a row, after dropping from 65.26% in 2013 to a low of 46.03% in 2023. As of 2025, the transaction margin is 46.62%, implying that the bleeding has ceased for the time being.

Figure 6: Transaction margin dollars

Looking at the operational statistics, it is a bit of a mixed bag. For the first time since 2022, PayPal managed to grow the average active accounts by ~7 million, representing 1.8% growth Y/Y.

Despite 7 million additional active accounts, the number of payment transactions decreased by ~1 million compared to 2024. For reference, it is the first time in reported history that PayPal failed to grow the number of payment transactions Y/Y. Historically, payment transactions have been growing on average by ~20% CAGR since 2013, and that includes the -3.7% in 2025.

Since total TPV grew by 6.7% Y/Y (10% FXN) while total payment transactions decreased, it means that the average payment value increased, crossing $70 for the first time (~11% growth).

Figure 7: Average payment value per transaction

The saving grace and investment thesis

The rest of the research report is for paid subscribers only.

Please consider becoming a paid subscriber to help me do this full-time. That would allow me to deliver more frequent as well as higher-quality research.

Read more

Updates are provided once a month with commentary, changes and a breakdown.

Returns since inception (May 26, 2025):
Type-F Thematic: AI: 21.73%
S&P 500: 14.63%
Performance against index: +11.86%

As of: 10 March, 2026

Read more

Updates are provided once a month with commentary, changes and a breakdown.

Returns since inception (July 31, 2025):
Type-F Thematic: Energy: 16.92%
S&P 500: 9.05%
Performance against index: +7.87%

As of: 10 March, 2026

Read more

Updates are provided once a month with commentary, changes and a breakdown.

Returns since inception (May 26, 2025):
Type-F Thematic: Barrier To Entry: 16.71%
S&P 500: 14.74%
Performance against index: 1.97%

As of: 10 March, 2026

Read more

Updates are provided once a month with commentary, changes and a breakdown.

Returns since inception (May 27, 2025):
Type-F Thematic: Financials: -6.49%
S&P 500: 14.74%
Performance against index: -21,23%%

As of: 10 March, 2026

Read more

Illustration 1: PayPay’s ecosystem

Company profile

Theme: Financial SuperApp, Direction: Buy (within fair value range)

Symbol: PAYP, Exchange: NASDAQ
Sector: Financials, Industry: Credit Services
Fair intrinsic value: $18 927 billion, as of March 6, 2026
Targeted IPO valuation: $14 000 - $20 000 billion
Major underwriters: Goldman Sachs, J.P. Morgan, Mizuho, Morgan Stanley

TYPEFCAPITAL.COM
Consider following me on X and YouTube.

This article is for informational purposes only and does not constitute investment advice or an offer to buy or sell securities.

Not just another Fintech

The Tokyo Stock Exchange has been largely stagnant for 25 years; however, that has now changed. The Nikkei index is up 53% over the past year and has had a massive resurgence as of late. Many Japanese companies have seen extreme upside in a relatively short amount of time, as outside investor interest has skyrocketed. An interest that SoftBank wants to leverage by expanding overseas with its PayPay subsidiary.

PayPay was initially expected to float in December, but due to the U.S. government shutdown, it was postponed. In mid-February 2026, PayPay Corp filed for a U.S. IPO and is expected to float as soon as this month (March).

Another fintech company may not be too exciting for Western investors, as it is regarded as a highly competitive space with eroding margin profiles. That is not true for Japan. Japan’s unique culture and core demographies turn that narrative on its head, as the country is predominantly cash-heavy and has an aging population. A digital payments transformation is a secular tailwind, and there are not many players in the space vying for market share.

In 2018, Japan’s government set a goal of reaching 40% cashless payments by 2025. Japan was lagging behind the rest of the developed countries. For reference, at the start of 2018, Japan sat at 21% cashless, while South Korea’s cashless ratio was ~90%, China’s was over 60%, and the U.S. around 50%. The government realized that maintaining a physical cash infrastructure was costing the economy trillions of yen annually. The same year that the goal was outlined by the government, PayPay was launched through a joint venture between SoftBank Group and Yahoo Japan.

As a stark contrast to the cashless adoption, Japan has the highest smartphone penetration rate globally of adults under 50 (~100%). That is an impressive feat, considering how late Japan was to adopt smartphones, as they relied on an advanced ecosystem of domestic feature phones (flip phones) that already had features such as mobile TV, internet, email, and digital wallets. All that was needed to start the cashless revolution was a catalyst and someone to capitalize on the opportunity. The catalyst was the Japanese government, and the company to take advantage of the opportunity was PayPay. The 2025 goal of reaching 40% cashless, laid out by the Japanese government, was effectively reached by the end of 2023.

Figure 1: Japan’s cashless payment ratio, end of period

How PayPay dominates

Depending on which third-party estimates of Japan’s smartphone penetration you use, about two-thirds of all Japanese smartphone users are registered for a PayPay account. In the most optimistic assumption, implying that 98% of all Japanese use a smartphone in 2025, PayPay has a ~60% registry rate. A rate that is still trending upwards aggressively.

Figure 2: PayPay registered users and market share of Japanese smartphone users

The PayPay usage rate translates to an even higher overall market share of the primary service, which is code-based in-store payments. Over the past four years, of the ¥64.56 trillion in-store QR code payments volume, around 67% was paid through PayPay. Calculated using the following formula:

Figure 3: PayPay in-store code payments market share

The playbook that PayPay employed to capture a majority of the market combines network effects with a powerful flywheel effect. The key elements of the strategy can be outlined as follows:

The rest of the research report is for paid subscribers only.

Please consider becoming a paid subscriber to help me do this full-time. That would allow me to deliver more frequent as well as higher-quality research.

Read more