The Week that Meta Employees Became Training Data

Meta reportedly wants to record employees’ clicks and keystrokes to train AI agents. It’s Taylorism for knowledge work: turn judgment and autonomy into measurable workflow data, then optimize it—and potentially automate the job away.

Pro Poker Player Says New Tax Laws Forced Him Into Semi-Retirement

You always have to account for tax drag.

A 2026 tax change lets gamblers deduct only 90% of losses. For high-volume pros, that can mean owing tax even when breaking even or even losing money.

Constellation Software’s Influences: An Overview of High Performing Conglomerates

Mark Leonard built Constellation by studying “high performing conglomerates.” and their decentralized acquisition discipline, high switching-cost niches, and relentless focus on returns on capital.

Acquired: Ferrari

I always joke “is it a Ferrari?” when someone says they have a surprise for me and listening to deep dive into the history of Ferrari makes me dream of one even more.

How Private Equity Ate Britain

Subscribe now

The Investment Case for Listed Railroads

Actual physical network effects with extremely high barriers to entry for competitors. The UNP & NSC merger should unlock a lot of savings and increase demand for their transcontinental system.

Long Broadridge Financial

Broadridge was not immune to the Saaspocalypse but it brought the stock price of Wall Street monopoly down to a great price.

Honoring Murray Stahl: Replay of a Fireside Chat with the Legendary Investor at Latticework 2018

“In a fireside chat, Murray Stahl discussed finding value in an increasingly indexed market, cryptocurrency, and unconventional compounders.”

Mitchell Green - Lessons from Cold Calling 10,000 Companies

An interesting angle on VC and growth investing.

2025 Tesla Cybertruck: Regular Car Reviews

Subscribe now

The Charlie Munger 200-week moving average screen is supposed to help find high-quality companies trading at a discount, where short-term sentiment does not reflect the business’s long-term earning power.

But like all value-focused screens, it will also surface value traps, or in this case, quality traps: businesses whose returns look great in the rearview mirror but whose future is slowly eroding.

The name that popped out on a recent run of that screen, and that struck me as a potential quality trap, is Morningstar (MORN) because it is facing two secular changes that should negatively affect its business.

The first has been grinding away for a decade: the migration from active mutual funds to passive ETFs, which cuts at the heart of Morningstar’s original franchise.

On a assets basis it looks gradual but when you look at flows, the secular shift is more pronounced.

Since 2015, long-term mutual funds have had net outflows in most years. Since 2020, those outflows have run in the hundreds of billions annually.

ETFs, meanwhile, have pulled in positive flows every year, hitting a record $1.1T in 2024.

Active mutual funds are a shrinking pool, losing $250–500B a year. The main things propping them up are market appreciation and 401(k) plans.

Morningstar is synonymous with mutual funds.

It famously started in 1984 at Joe Mansueto’s kitchen table to fill a void and provide institutional-quality research on mutual funds. Any business built on the rise of mutual funds should show signs of deterioration as mutual funds decline in popularity.

The second threat is generative AI and its potential to commoditize the kind of packaged research and data that Morningstar sells. That threat is the reason MORN used to trade at 36.6× EBITDA in early 2024 but now sits at ~10.5× today, a multiple last seen during the GFC.

I worked in labs throughout college because I originally planned to become a scientist. That obviously did not happen, but the scientific method is ingrained in my thinking. So, like any good scientific endeavor, the goal here is to try to disprove my hypothesis.

If I can’t, then the Munger screen might have surfaced exactly what it was supposed to: a high-quality company experiencing undue negative short-term sentiment relative to its long-term earning power.

The cleanest way to run this test is segment by segment. Morningstar reports five segments today — Morningstar Direct Platform, PitchBook, Morningstar Credit, Morningstar Wealth, and Morningstar Retirement — plus an Indexes business sitting inside “Corporate and All Other” that just got a lot bigger and will probably become Morningstar’s sixth segment.

Subscribe now

Morningstar Direct Platform

This is the segment people mean when they say “Morningstar.” It includes Morningstar Data, Morningstar Direct, and Advisor Workstation. These are the data feeds and analytical platforms that sit on the desktops of financial advisors and institutional analysts.

At roughly $820M of annual revenue, it is the largest segment. On paper, it is also the most exposed to both threats: the mutual fund database was built for an active-management world, and the GUI-based research workflow is exactly what an LLM should disrupt first.

So is it melting?

The first place to look is seats. Morningstar Direct licensed users are disclosed quarterly in the supplementals.

Direct licenses grew from ~17,421 at the end of 2021 to 18,562 by late 2023. Since then, growth has essentially gone flat. Over the next six quarters, users increased from 18,562 to 18,810.

If AI were going to strangle anything at Morningstar, it would start here, and the flattening licensed-user curve supports that story.

The AI threat is less about churn and existing users leaving to build their own solutions. It is more about slowing new customer acquisition, and the renewal rates support that.

In 2025, Morningstar Direct posted net retention of ~104% and Morningstar Data posted ~101%. Above 100% net retention means existing customers expanded contract value net of any churn.

Institutional data platforms in finance have always been workflow-sticky. Compliance-bound clients are slow to rip out incumbent systems, templates are embedded in downstream reporting, and Advisor Workstation in particular carries FINRA-reviewed audit trails that ChatGPT cannot replicate.

What the chart shows is Morningstar holding on tightly to existing customers while struggling to add new licensed seats. It is not just about AI. Last year was also a tough budget environment, and that showed up across several companies that serve the financial industry.

Right now, it is a decelerating business, not a melting one.

It is also what you would expect from a mature-category leader. Price protection and unit economics remain strong, but the top of the funnel gets harder.

Morningstar ≠ Mutual Funds

The idea that Morningstar is a mutual fund business misunderstands what Morningstar actually sells.

Morningstar’s customer is not the mutual fund investor. It is the financial professional who needs to analyze investments of any structure.

An advisor building a client portfolio in 2026 still has to compare expense ratios, holdings, factor exposures, tracking error, and historical performance. The advisor is agnostic to whether the asset is a mutual fund or an ETF.

It is easy to see why the decline of mutual funds did not coincide with the decline of Morningstar’s core business.

Morningstar Indexes

As we’ll see in the other segments, Morningstar’s management has a track record of recognizing where the business is headed, investing heavily in the right areas, and exiting the wrong ones.

They know ETFs are the future, and they have used their brand recognition to enter the index and ETF business.

So far, it has been a slow rollout. Morningstar Indexes is still reported inside “Corporate and All Other”.

But they just made a major investment with the acquisition of CRSP, and it should get its own reporting segment soon.

Morningstar Indexes revenue compounded from $37M in 2021 to ~$88M in 2025 before the CRSP acquisition. Then, in February 2026, Morningstar closed the $365M acquisition of CRSP, the University of Chicago-affiliated index provider whose US Total Market Index is the benchmark the Vanguard Total Stock Market Index Fund (~$2T AUM) is contractually required to track.

On a pro forma 2026E basis, the Indexes business jumps to roughly $143M, a ~60% revenue bump.

That growth is meaningful, but the acquisition also reshapes the competitive landscape.

Morningstar Indexes went from a subscale also-ran to #4 globally, with $4.2T of assets benchmarked.

Of that $4.2T, roughly $3T+ is already in ETFs, which is a higher ETF share than every other provider in the table except S&P Dow Jones.

The segment is not without risk.

Vanguard is now the single largest client relationship inside Morningstar Indexes, and it is a famously price-sensitive counterparty. Vanguard left S&P for CRSP in 2012 specifically over pricing. The renewed multi-year license is reassuring, but it is not permanent, and future Vanguard renegotiations will be the most-watched events in this segment.

Management knows the future is ETFs, and it needed to become a bigger player in the space. The acquisition of CRSP, and the resulting Vanguard relationship, is a huge win and a smart investment.

Granted, the immediate revenue contribution from Indexes post-CRSP acquisition is still small at 5.7%. But the margin contribution is much higher. MSCI’s index business runs 75% adjusted EBITDA margins, and if Morningstar’s Indexes is similar, that implies about $110M in segment profit and roughly 15% of total EBITDA.

PitchBook

PitchBook is the key segment that determines whether MORN is interesting or not. It’s the private-markets intelligence platform Morningstar bought in 2016.

Revenue grew from ~$207M in 2020 to $672M in 2025 — roughly 3.2× over five years — while adjusted operating margins moved from 26.6%, down to a 2022 investment-cycle trough of 16.0%, and back up to 31.3% in 2025.

Adjusted operating income went from $55M to $210M over the same window, nearly 4×. Even with growth decelerating, the profitability curve is moving in the right direction. Margins are expanding, not compressing, which is the opposite of what you’d expect from a business being commoditized.

The AI disruption narrative has the most superficial appeal with PitchBook because it is the closest thing Morningstar has to a “research portal” product.

But PitchBook’s value isn’t the UI. It’s the database behind it.

Decades of accumulated deal data, fund-level returns, LP commitments, portfolio company financials, and league tables. Private-markets data is not on the open web. An LLM cannot scrape it because it was never posted. It has to be sourced, verified, and maintained the way PitchBook has been doing since the mid-2000s. So when a new AI-native application needs canonical private-markets data, whose data does the AI call?

Management has been active in making sure that call goes to PitchBook. They’ve signed data and distribution partnerships with OpenAI, Anthropic, Microsoft, and most recently, Perplexity.

PitchBook is also rolling out MCP servers and AI APIs for generative summaries and model connectivity.

Morningstar is also trying to turn PitchBook from just a private-markets database into a broader research platform by adding more asset classes and analytical layers.

They’ve added analyst-driven public equity research to the database. It also aggregates third-party research from nine outside partners.

Morningstar bought Leverage Commentary & Data (LCD) from S&P in 2022. This brings leveraged loan and high yield bond data into PitchBook Credit. They also bought Lumonic to improve portfolio monitoring of credit positions.

On top of distribution deals with top AI companies, PitchBook is adding its own AI research layer, PitchBook Navigator.

The problem is that all this investment and expansion has not changed PitchBook’s revenue deceleration.

PitchBook’s organic growth decelerated from 30%+ in 2021 to 8.5% in 2025, and Q4 2025 came in at just 5.9% organic, the lowest quarter in the dataset.

Licensed user count was essentially flat in 2025 at 113,447 versus 113,451 the year before. To be fair, there was a methodology change in how users were counted (internal active users were removed from the count), but it is still a caution sign.

The combined PitchBook platform & direct data renewal rate also ticked down from 108% in 2024 to 103% in 2025.

I see two reasons for this.

The platform expansion is aimed at deepening wallet share with core investor/advisor clients, which pushes out smaller users and increases churn.

The easy upsell to current clients is gone. This is, again, a new user acquisition issue.

Either PitchBook is settling into a normal mature-SaaS growth rate (high single digits), with AI partnerships reaccelerating it as the distribution deals compound, or PitchBook is structurally breaking and heading for sub-5% organic growth as AI alternatives pick at the edges.

The next four quarters are important.

If PitchBook reaccelerates even modestly, the whole MORN thesis gets materially easier. If it slips further, the bear case on the stock gets a lot more credible.

Subscribe now

Morningstar Credit (DBRS)

Morningstar Credit is essentially DBRS, the fourth-largest credit rating agency in the world, plus Morningstar Credit Analytics.

It is one of roughly ten SEC-registered nationally recognized statistical rating organizations (NRSROs), and one of only four credit rating agencies the ECB recognizes as an external credit assessment institution. The other three are S&P, Moody’s, and Fitch.

An LLM cannot assign an NRSRO-recognized credit rating. No issuer is going to structure a CLO around ChatGPT’s or Anthropic’s credit opinion, because the regulatory framework that governs CLO tranching, bank capital treatment, insurance-company investment mandates, and a dozen other downstream uses of credit ratings requires a recognized, licensed rating from an approved body. The cost of entry is enormous: multi-year track record requirements, jurisdictional approvals, and issuer inertia. The list of approved bodies is essentially closed.

And the business itself has just come ripping out of a cyclical trough.

Credit revenue peaked at $271M in 2021, then fell for two straight years as the Fed hiked rates and CMBS / RMBS issuance collapsed.

Then the cycle turned.

This segment grew revenue 35.1% in 2024 and another 27.9% in 2025. It was the fastest-growing Morningstar segment for two years running. The growth was driven by strength in corporate ratings globally and structured finance ratings in the US and Europe.

DBRS has been taking share steadily for years by positioning itself as a credible “diversity alternative” to the big three. Post-2008, issuers have a structural incentive to get a second or third rating from a non-big-three agency for capital-markets-access reasons.

DBRS is the least likely business division to be disrupted by AI.

Morningstar Wealth

Morningstar Wealth was historically a TAMP (turnkey asset management platform) business, offering model portfolios and SMAs for financial advisors.

Full-year 2025 revenue was roughly flat, and adjusted operating margins were in the low single digits.

This business is a melting ice cube.

Not because of AI or the secular shift to ETFs, but because it is a highly competitive industry where Morningstar has no meaningful advantage versus larger platforms like Envestnet.

Management recognizes this and is unwinding the business. Through 2024 and 2025, Morningstar has been sunsetting TAMP assets and booking some gains on sales.

Morningstar Wealth is small (~$255M of revenue, roughly 10% of the company), and eliminating the drag is accretive to consolidated margins. Removing the Wealth drag could add another 100–200 bps to Morningstar’s consolidated adjusted operating margin.

Morningstar Retirement

Morningstar Retirement is the smallest segment, at roughly $125M in revenue, with mid-40s adjusted operating margins and mid-single-digit growth.

Morningstar Retirement sells managed-account and fiduciary services to 401(k) plans. It sits on top of employer-sponsored retirement plans and provides personalized allocation advice to participants. Revenue is asset-based and scales with market levels and participant adoption.

There is not much of a disruption story here.

The business is protected by regulation (managed accounts carry fiduciary responsibility, which consumer AI tools do not), the customer base is sticky (plan sponsors rarely switch recordkeepers because it is operationally painful), and the growth driver is plan-level participant adoption that compounds quietly.

What Expectations are Priced In?

There are no bad assets, only bad prices.

- Howard Marks

The final piece is what expectations the market is pricing in. If expectations are too pessimistic, then Morningstar could be a buy even if some AI disruption occurs.

Read more

Who Is Satoshi Nakamoto? My Quest to Unmask Bitcoin’s Creator

John Carreyrou helped uncover the fraud and deceit at Theranos. Has he now unmasked Satoshi Nakamoto, the creator of Bitcoin, too?

Critical Review of SpaceX’s Financials

It’s part of a comment thread over at Hacer News, but it breaks down SapceX’s capital intensive nature, its Depreciation & Amortization, and how SpaceX is its own largest customer.

There Is Not Enough Claude

AI demand will keep increasing as models get more capable and spawn new “spillover” use cases. Like the internet, a general-purpose tech can keep compounding demand for decades. Leading to more compute scarcity.

Michael Mauboussin: Base Rates, AI Adoption, and Investing in the Intangible Economy

Michael joins Kai Wu to discuss his most recent white paper on base rates and growth expectations for AI companies like OpenAI and what investors should focus on.

Behind Chick-fil-A’s Unconventional Franchise Model

Subscribe now

“If all you ever did was buy high-quality stocks on the 200-week moving average, you would beat the S&P 500 by a large margin over time.”

- Charlie Munger

Taking Charlie’s advice I ran a screen for companies trading at or below their 200 week moving average.

I also added a Insider Buy column to see which companies had recent insider transactions that could act as a short-term catalyst.

The 22 companies with recent insider buys are listed below.

Some interesting names pop up like Microsoft, American Tower, and Broadridge Financial Solutions.

And then there are some other names that could be potential value/quality traps that I’d might explore in an upcoming issue.

The full screen of 122 companies is available below to paid subscribers.

Read more

The HALO Effect: Heavy Assets, Low Obsolescence

Capital-intensive stocks have outperformed capital-light peers by ~35% since 2025.

AI is compressing software moats while repricing hard-to-replicate physical assets like grids, pipelines, and industrial infrastructure upward.

ServiceNow: Deep Dive Analysis

ServiceNow is the workflow operating system for big enterprises with 98% renewals and strong free cash flow.

It too got caught up in the SaaSpocalypse but AI increases the need for a governed orchestration layer, not less.

Bayes and Base Rates: How History Can Guide Our Assessment of the Future

OpenAI’s $145B revenue target implies 108% CAGR over 5 years. Across ~18,900 observations of U.S. companies since 1950, zero have grown that fast from a $2-5B base.

Part Two: How Jeffrey Epstein Helped Build the Modern World

With the Epstein files slowly being released, we’re starting to see how truly connected he was. In part one and two, The Behind the Bastards crew look at how Epstein had a hand in the rise of Bitcoin and micro transactions in video games.

• Part one is more back story on the people, like the Mighty Duck actor, who became Epstein’s connections to the world of crypto currencies.

How Nike Lost Its Way

Subscribe now

On January 19, 2026, the New York Stock Exchange announced it is building a blockchain-based platform for trading and settling tokenized versions of U.S. stocks and ETFs.

24/7 trading with instant settlement, funded by stablecoins.

Two months later, NYSE named Securitize — the BlackRock-backed tokenization firm — as its first digital transfer agent, eligible to mint blockchain-native versions of stocks and ETFs on the platform. Three weeks before that, NYSE’s parent company, Intercontinental Exchange, invested roughly $200 million in crypto exchange OKX at a $25 billion valuation, securing a board seat and a deal that will let OKX’s 120 million users trade tokenized NYSE-listed securities.

Last week, ICE completed a $600 million cash investment in prediction market Polymarket, bringing its total commitment to $1.6 billion as part of a broader $2 billion arrangement.

What’s happening here is a coordinated infrastructure play by the operator of the world’s most iconic stock exchange to rebuild market plumbing for the next era of capital markets.

For investors evaluating ICE as a long-term holding, this matters, not because tokenization will drive earnings next quarter, but because it represents asymmetric optionality that the market is currently pricing at zero.

What Is Tokenization & Why Should Equity Investors Care?

Tokenization is the process of representing ownership of a financial asset — a share of Apple stock, a Treasury bond, a unit in an ETF — as a digital token on a blockchain. The underlying asset doesn’t change. What changes is the infrastructure through which it’s recorded, traded, and settled.

Currently, when you buy a share of stock through your brokerage, the trade doesn’t truly settle for one business day (T+1). During that gap, a chain of intermediaries — clearing houses, depositories, custodians, transfer agents — reconcile records, move money, and update ownership.

Tokenized securities compress that entire process into near-instant, on-chain settlement. And they unlock capabilities the current system simply can’t offer:

24/7 trading. Global investors don’t stop wanting to trade U.S. equities at 4:00 PM Eastern. A tokenized market can operate continuously, capturing demand from every time zone.

Instant settlement. T+0 eliminates counterparty risk during the settlement window and frees up capital currently locked in the clearing process.

Fractional shares by design. Orders are sized in dollar amounts, not share quantities. A $50 investment in Berkshire Hathaway becomes trivially easy.

Stablecoin-based funding. Programmable digital cash enables settlement without dependence on traditional banking hours or wire transfer infrastructure.

The question isn’t if this will happen. It’s coming. The questions are how big will this be and, more importantly, who builds the infrastructure and gets to collect the tolls.

ICE’s Multi-Front Strategy

What distinguishes ICE’s approach from crypto-native tokenization experiments is that it’s working from a position with a structural advantage.

ICE already operates the matching engines, clearing houses, data networks, and regulatory relationships that underpin major global capital markets. It’s not building from scratch. It’s extending existing infrastructure onto new rails.

The strategy has four interlocking pieces:

1. The NYSE Digital Trading Platform

The core product. NYSE is building a new venue that supports trading of tokenized shares fungible with traditionally issued securities, as well as tokens natively issued as digital securities. It combines the Pillar matching engine — the same technology that handles billions in daily equity volume — with blockchain-based post-trade settlement across multiple chains.

Critically, tokenized shares retain full shareholder rights. This is not a synthetic derivative or a crypto wrapper. It’s a genuine share of stock, represented on-chain. That’s a fundamental distinction from every tokenized equity product currently available on offshore platforms.

ICE is also working with BNY and Citi to support tokenized deposits across its clearing houses, enabling margin obligations and funding transfers outside traditional banking hours. Launch is targeted for late 2026, pending SEC and FINRA approval.

2. Securitize: The Minting Partner

On March 24, NYSE signed an MOU with Securitize, naming it the first digital transfer agent eligible to create blockchain-native securities on the platform. Securitize — backed by BlackRock and Ark Invest, and SEC-registered — handled the issuance infrastructure for BlackRock’s tokenized fund (BUIDL). Its broker-dealer arm could participate in trading, giving it a role across both issuance and market activity.

This is where NYSE’s approach diverges from Nasdaq’s. Nasdaq recently received approval for tokenized trading, but plans to process it through traditional clearing infrastructure. NYSE is building a separate blockchain-native venue.

3. OKX: The Distribution Channel

ICE invested roughly $200 million in OKX at a $25 billion valuation, took a board seat, and established a partnership with two commercial dimensions.

First, ICE will license OKX’s spot crypto prices to launch regulated crypto futures which is a direct revenue opportunity for ICE’s exchange segment.

Second, and more strategically, OKX will offer its 120 million global users access to ICE’s U.S. futures and tokenized NYSE equities.

This solves ICE’s single biggest gap. The NYSE is the most iconic exchange brand in the world, but it has no direct distribution channel to international retail or crypto-native investors. OKX provides that channel overnight. If tokenized equities take off, ICE now has the matching engine, the regulatory framework, and the global retail distribution to capture it.

4. Polymarket: The Data Play

ICE’s $2 billion commitment to Polymarket makes it the global distributor of prediction market data to institutional investors.

In February, ICE launched the Polymarket Signals and Sentiment tool, normalizing real-time prediction market probabilities into structured data feeds.

This fits the broader strategy in two ways. First, prediction market data is a genuinely new data asset unavailable through conventional sources. Second, Polymarket runs on blockchain smart contracts, making it a natural adjacency for ICE’s on-chain infrastructure ambitions. ICE and Polymarket have explicitly agreed to collaborate on future tokenization initiatives.

What This Means for ICE Shareholders

Read more

In this episode, I break down each of Intuit’s four major business lines — Credit Karma, Mailchimp, TurboTax, and QuickBooks Online — and stress-test them against the AI disruption thesis. For each segment, I walk through the bear case (where AI agents pose a genuine threat) and the bull case (where Intuit’s data moats, switching costs, and platform strategy create durable advantages).

We cover Credit Karma’s conflict of interest and its new OpenAI partnership, why vibe-coding your own ESP is harder than it looks, TurboTax’s underrated structural advantages (including guarantees no LLM will match), and how QuickBooks could evolve from accounting software into a virtual business team powered by AI agents.

If you own Intuit, are thinking about buying it, or just want a framework for evaluating AI disruption risk in quality businesses — this one’s for you.

📖 Read the full article with valuation analysis and portfolio action plan: https://www.incrementalreturns.co/p/intuit-vs-ai-agents-threats-growth-engines

📬 Subscribe to Incremental Returns for deep dives on quality businesses with durable competitive advantages:

Incremental Returns

Insights that build your financial future one increment at a time. Subscribe for actionable strategies that transform small consistent efforts into substantial long-term returns.

From Glutted to Gutted

A Hormuz shutdown means the global economy is out 20M barrels/day. This loss is overwhelming quick supply responses, draining inventories, and pushing oil prices higher. Even stock releases and pipeline bypasses only soften the blow.

How To Tell There’s A Recession Coming, According To Sex Workers

Good luck explaining to your partner that you have to go to a brothel to gather some alternative economic coincident data.

There Are Only Two Paths Left for Software

Software’s middle lane is disappearing. CEOs have two credible paths: build AI-native products that add 10+ points of growth fast, or redesign the company for 40%+ true operating margins (including stock comp). “A little of both” won’t survive.

Fighting Words: The Energy Transition in 2026

Michael Cembalest covers the highlights from his team’s big 2026 energy report. It covers data center energy demands, the feasibility of small modular nuclear reactors, and obviously Iran.

WeWork - The $47 Billion Disaster

Subscribe now

In the spring of 1999, a dot-com darling did what a lot of dot-com darling did.

It hired a celebrity.

One of the most recognizable faces in American pop culture delivered an over-the-top pitch for a website that promised to revolutionize the way people booked travel. And for a brief moment, Wall Street believed. Its value soared to $24 billion.

The stock even rose more than 1,000% in a single month after its IPO.

Then the bubble burst.

By late 2002, the stock had fallen 99.3%. The market cap shrank from $24 billion to $250 million. The celebrity commercials disappeared and analysts stopped covering it.

The company became another monument to the dot-com excesses alongside the sock puppet.

But…

While the company’s stock was trading as if it were going out of business, a quiet dealmaker inside the organization made two small acquisitions in Europe. From the outside, both looked like the moves of an also-ran trying to stay relevant.

The combined price of the two acquisitions was less than $300 million.

At the time, the parent company was generating just $10 million in annual profit. It was trailing two larger, better-capitalized competitors by a wide margin. Nothing about it suggested a comeback.

But buried inside those two acquisitions was a business model built on strong two-sided network effects. A platform that got more valuable every time a hotel signed up, because it brought more travelers. And more valuable every time a traveler booked, because it gave hotels a reason to offer better rates. A compounding, self-reinforcing loop that didn’t require a celebrity endorsement.

Just supply, demand, and our company acting as a toll booth between them.

We are living through a similar moment right now.

Over the past two years, the rise of AI coding tools and autonomous agents has triggered a broad sell-off in software and internet businesses. The concern is straightforward: if AI can replicate what software does at a fraction of the cost, the revenue streams that once justified premium valuations become harder to defend. Multiples have compressed. Growth assumptions have been revised downward. The market is repricing the entire sector.

But what happened in 2002 is a pattern worth recognizing.

So far, the AI sell-off has not been as severe as the tech bubble bursting, when investors stopped paying attention to the entire sector and missed the companies that survived and emerged structurally stronger. Those companies sat at the center of two-sided marketplace networks, like Amazon.

But AI fears are once again causing investors to react to market prices without considering the underlying competitive advantages of the business.

This month’s recommendation operates a deeply entrenched global two-sided marketplace. It has survived every cycle — the dot-com bust, the financial crisis, a global pandemic that shut down travel entirely — and emerged from each one stronger and with more market share than it entered with.

The network effects that make it valuable today were invisible to most investors in 2002, and they look underappreciated again. And AI could accelerate the very dynamics that make this platform defensible.

Read more

In this episode, I break down why Broadridge has dominated US financial infrastructure for over two decades, processing more than 80% of outstanding shares in the country while maintaining revenue retention rates of 97–98%. This isn’t a story about a flashy product or a disruptive technology — it’s a story about a business embedded so deeply into compliance-critical workflows that no one wants to replace it, and no competitor can easily replicate it.

We cover:

• Why “street name” securities create a structural moat that’s almost impossible to dislodge

• How five-to-seven year contracts and low strategic priority for broker/dealers cement Broadridge’s incumbency

• The scale advantages that compound over time — including householding and postal optimization

• Why the NYSE-regulated fee structure actually removes the pricing lever a disruptor would normally use

• The subtle but real network effects at work across issuers and broker/dealers

• Why the direct indexing boom is a quiet volume multiplier for Broadridge’s core revenue

This is the kind of business that doesn’t make headlines but quietly sits at the center of how US equity ownership actually functions.

📖 Read the full article: https://www.incrementalreturns.co/p/the-hidden-monopoly-financial-infrastructure-broadridge-financial

For more quality investing analysis, subscribe to Incremental Returns: www.incrementalreturns.co

The New Age of Investing

With the rise of the internet and now AI, traditional information advantages are disappearing and the edge is shifting to processes and mindset.

The Price of Having Fun

Growth investing is like buying a lottery ticket. A few extreme outliers but mostly losses.

Costco’s Kirkland Brand Is Bigger Than Nike—and It’s About to Get Even Bigger

Kirkland/Costco shows how a private label can become a trusted mega-brand across multiple categories of products.

Matt Belloni on the Oscars, the Ellisons, and Hollywood’s Next Chapter

From the Ellison takeover of Warner Bros. & Paramount to the Oscars moving to YouTube. What $80B in debt, Gulf sovereign money, and the creator economy mean for the future of film & TV.

The Academy needs money. YouTube was stepping up to maintain the license fee that the Academy has become accustomed to... In 2029, are a lot of people going to recognize and realize that you’ve got to download the YouTube app, put it on your TV or phone and watch it there? I don’t know that the older audience for the Oscars is going to do that.

I Taught AI Warren Buffett’s Strategy. It Picked Better Stocks and Still Lost

Subscribe now

Markets are supposed to be efficient, but efficiency depends on the quality of the data market participants use to make decisions.

Invisible Women by Caroline Criado Perez is a public policy and social advocacy book, but its underlying thread is the systematic failure to collect and analyze data on half the population: women.

The result is that products, policies, infrastructure, and algorithms built on this incomplete data are systematically wrong. Products do not fit women’s bodies. Medical treatments are tested only on men. AI systems are trained on male-default datasets. Urban infrastructure is designed around male commuting patterns.

Every one of these issues affects investing because “systematically wrong” often means “systematically mispriced.” The gender data gap is a structural market inefficiency, and inefficiencies create opportunities.

I am going to try to translate the book’s central thesis into a framework that public market investors can use to identify mispriced opportunities, hidden risks, and underappreciated secular trends.

Subscribe now

The $12 Trillion Misallocation: Unpaid Labor as Invisible GDP

McKinsey estimated in 2015 that closing the gender gap in labor force participation could add $12 trillion to global GDP.

Between 1970 and 2009, the addition of 38 million women to the US labor force made GDP 25% larger than it otherwise would have been. Every percentage-point increase in women’s employment yields a disproportionate increase in GDP.

Any country that wants to increase incremental GDP growth should tap into this economic reservoir, but the major hurdle is unpaid care work.

Globally, women do 75% of unpaid care work, spending between three and six hours per day on it compared to men’s thirty minutes to two hours. This unpaid labor subsidizes the entire economy, yet it is invisible in traditional economic measurement. When governments cut public services, demand does not disappear. The work simply shifts onto women, reducing their participation in paid labor and weighing on GDP.

The companies and countries that enable women’s participation in the paid workforce through childcare infrastructure, flexible work, and parental leave are not going “woke,” and they are not shifting the burden of childcare costs from people who want kids to people who do not want kids, they’re unlocking another reservoir of economic growth.

Studies cited in the book showed that a decrease in the time British women spend doing unpaid work from five to three hours correlated with a 10% increase in their paid labor-force participation and higher participation rates corresponds to higher GDP growth that benefits all.

The Care Economy as Infrastructure Investment

The book makes the case that the care economy — childcare, elder care, domestic services — is genuine economic infrastructure.

Early childhood education was found to have a greater positive impact on long-term economic growth than business subsidies, adding an estimated 3.5% to GDP by 2080. Adults who received ECE were more likely to be employed (76% vs. 62%), to own homes, and to have savings accounts. In New York, before the introduction of universal preschool, more than a third of families waitlisted for childcare assistance lost their jobs or were unable to work.

Investing 2% of GDP in the caring industries would create nearly 13 million new jobs in the US, compared to 7.5 million from the same investment in construction — the traditional “infrastructure” play. The UK’s Women’s Budget Group found that investing 2% of GDP in public care services would create almost as many jobs for men as investing in construction, but up to four times as many jobs for women.

Investors typically associate “infrastructure” with roads, bridges, and broadband. But sometimes you need to do the cost-benefit analysis with better data. Maybe it’s cheaper in the long run to invest in social infrastructure if it reduces lost labor costs, improves workforce participation, and drives GDP growth.

The Venture Capital Blind Spot

93% of venture capitalists are men. And men back men.

The Boston Consulting Group’s (BCG) research found that female-owned startups generate 78 cents of revenue per dollar of funding, compared to 31 cents for male-owned startups. They also generate 10% more cumulative revenue over five years.

This is one of the starkest examples of a market inefficiency driven by a data gap. The entire venture capital and growth equity pipeline is systematically mispricing female-led companies because the capital allocators evaluating them are pattern-matching against a default founder profile that looks like themselves.

For public market investors, this matters because the mispricing doesn’t stop at Series A. Companies that were underfunded in their early stages carry that disadvantage into the public markets. And the same pattern-matching bias that causes VCs to underweight female founders likely causes public market analysts to underweight female leadership more broadly.

Diverse Leadership as Mispriced Alpha

S&P Global found that within 24 months of taking office, female CEOs saw a 20% increase in stock price compared to their male counterparts. A portfolio simulation tracking companies led by women CEOs from 2002 to 2014 yielded a 348% return. The S&P 500 returned 122% over the same period. S&P 500 firms currently led by women CEOs have, on average, a 53% higher market cap and 68% higher net income than their male-led counterparts.

The CFO data is equally compelling. OneStream’s 2025 “Glass Chair” study, which analyzed over 1,100 finance leaders, found that underperforming companies that appointed a woman as CFO saw total shareholder return improve by an average of 10% relative to industry benchmarks. Publicly listed companies with women CFOs in the US, UK, and Europe delivered an average annualized shareholder return of 4.5% during their tenures, outperforming industry benchmarks by approximately 0.2% per year. That may sound small, but compounded over a decade it’s material.

Why the outperformance?

Research consistently shows that companies with female CFOs exhibit more conservative accounting, fewer financial restatements, lower levels of corporate overinvestment, and reduced stock-price crash risk. They tend to manage earnings more conservatively and maintain stronger internal controls.

The outperformance is especially pronounced during periods of stress. During financial crises — including the 2008 financial crisis and COVID-19 — firms with female leadership tended to achieve better returns, higher stability, and lower risk.

On the portfolio management side, Fidelity research found that women tend to trade less and hold firm during volatility, leading to higher risk-adjusted returns.

Another BCG study found that companies with above-average management diversity reported innovation revenue 19 percentage points higher than companies with below-average diversity. Diverse companies generated 45% of their total revenue from innovation, compared to just 26% for less diverse companies. McKinsey’s research, updated through 2023, found that companies in the top quartile for gender diversity on executive teams are now 39% more likely to outperform on profitability — up from 25% in their earlier studies.

Diverse teams are better informed about their actual customers. For pretty much every product out there, half of your customers are women. And when half your end-users are women, having women in leadership isn’t a diversity initiative. It’s a necessity. It’s the difference between designing products that actually fit your market and designing products that fit the assumptions of a homogeneous leadership team.

And yet, women still hold only 9% of CEO roles in the S&P 500. If the market were efficient at pricing leadership quality, this gap between performance data and representation would have closed long ago. The fact that it hasn’t suggests that the same male-default thinking Criado Perez documents in product design, urban planning, and medical research is also operating in how investors evaluate management teams. Female leadership remains structurally undervalued.

Product Markets Designed for Half the Population

The book is filled with examples of products that fail women because they were designed by and for men. Smartphones too large for women’s hands. Voice-recognition software 70% more accurate for male speech. Car crash-test dummies based exclusively on male bodies, which makes women 47% more likely to be seriously injured and 17% more likely to die in a car crash.

Every one of these design failures represents an addressable market.

Consider the breast pump market, estimated at $700 million with room to grow, dominated by a single company (Medela) selling products that users widely find inadequate. More women buy iPhones than men — so why are iPhones still designed more for men’s hands? The women’s health tech (femtech) space covers pelvic-floor health, menstruation tracking, and fertility — conditions affecting 50% of the population — yet the sector remains chronically underfunded.

Companies that close these product-design gaps don’t just capture underserved demand. They build switching costs from genuine consumer loyalty.

AI and Algorithms: Compounding The Gender Data Gap

The most important part of the book for investors and market participants going forward is the section on technology, algorithms, and AI. It comes down to a simple question: who controls and codes the algorithm, and what data do they use?

The book’s argument is that AI systems trained on historically biased data don’t just replicate the gender data gap, they amplify it at scale.

The compounding works like this: historical data is male-default because medical trials, employment records, speech corpora, and image datasets were all built with male subjects as the norm. Algorithms trained on this data encode the bias as objective truth.

The outputs then shape future data collection, creating a self-reinforcing feedback loop. Each cycle widens the gap.

The most glaring examples exposed in recent history involved hiring algorithms at Gild and Amazon.

Gild built hiring algorithms that inadvertently disadvantaged women by treating male behavioral patterns, like contributing to open-source code on GitHub at nights and on weekends, as proxies for talent. Women with unpaid care responsibilities are less able to code at midnight on a Saturday. The algorithm did not account for this because it was built predominantly by men who did not have experience living and working as a woman. The intent was not to be sexist, but training on biased data was enough to reinforce the bias.

Amazon’s now-infamous AI hiring tool ran into the same problem and had to be scrapped. Built to score candidates by learning from a decade of past resumes, it effectively learned that “male” signifiers were a proxy for “qualified” and began downgrading women’s applications for technical roles. Engineers tried to patch the model by stripping out explicitly gendered terms, but the bias kept resurfacing in subtler patterns, and Amazon ultimately abandoned the project rather than ship a system that automated discrimination.

In general, 72% of US CVs never reach human eyes. Interview algorithms are being trained on the posture, facial expressions, and vocal tone of “top-performing employees.” Were those top performers gender and ethnically diverse? Does the algorithm account for socialized gender differences in tone and facial expression? We don’t know, because the companies developing these products don’t share their algorithms.

And not knowing what data they are being trained on or what data gaps exist can be dire especially when making medical decisions

The book documents a medical knowledge base heavily skewed toward the male body. With AI diagnostics being built on that foundation, there’s a real risk that these systems make diagnoses for women worse, not better. The second most common adverse drug reaction in women is that the drug simply doesn’t work. Less than half of publicly available prescription drugs have been analyzed for sex differences. Building AI on top of this data without correcting for it is scaling a known deficiency.

I see the following investment implications.

Regulatory risk. As governments scrutinize algorithmic bias — the EU AI Act being the most prominent example — companies deploying biased AI face significant legal and reputational liability. Investors will need to evaluate AI-dependent companies not just on model performance but on training data composition, bias auditing practices, and transparency. The companies that can’t answer these questions are carrying an undisclosed risk.

Companies that solve for this bias can capture enormous value. Whoever builds medical AI that works for women, recruiting algorithms that select for talent rather than male behavioral defaults, or other AI-driven innovation trained on data competitors are missing will have a structural advantage. Today’s AI ecosystem is potentially leaving half the market underserved, which creates a large addressable market.

There is also a dangerous information feedback loop for investors. If the data and analytical tools investors use are built on the same male-default assumptions, then investment analysis itself inherits the gender data gap. Quantitative strategies that rely on NLP, sentiment analysis, or alternative data should be scrutinized for the same biases. If today’s investment and trading algorithms are making decisions on incomplete data, then the firm that incorporates the gender data gap into its decisions can gain a meaningful competitive advantage.

Closing the Data Gap Is an Alpha-Generating Strategy

Again, the book uses data, or more importantly, the lack of sex-segregated data, as a social advocacy tool to push for better policies to improve the lives of half the world’s population and to make economies and governments work better for everyone.

But from an investor viewpoint, I see a massive, structural market inefficiency.

The inefficiency persists for the same reason most market inefficiencies persist: the people making the decisions do not see it. Male-default thinking is so deeply embedded that it reads as neutral. Products designed for men feel like products designed for everyone. Data collected on men feels like data collected on humans. Algorithms trained on male behavior feel like algorithms trained on merit.

But neutrality requires evidence. And the evidence, once you look for it, overwhelmingly shows that the default is not neutral at all.

Companies and investors who recognize this can find mispriced opportunities, avoid hidden risks, and position themselves ahead of regulatory and demographic shifts.

Invisible Women: Data Bias in a World Designed for Men

Chartbook 436 Unseasonal war. How the US-Israeli war on Iran threatens the global agricultural cycle.

LNG disruptions and shipping chokepoints are spiking urea prices right when fertilizer needs to move. The biggest risk is not for rich countries, but for smallholder farming economies at the end of the chain.

Can Advanced AI Lead to Negative Economic Growth?

AI can raise potential output while lowering realized GDP if automation shifts income away from high-spending workers toward low-spending capital owners, causing demand to collapse

5 Lessons I Learned from Li Lu, the “Chinese Warren Buffett”

Long-term investing edge comes from thinking like an owner, staying temperamentally patient, and waiting to act decisively when a rare high-conviction opportunity offers a wide margin of safety.

Rory Johnston on How oil Could Surge to Over $200 a Barrel

A good Odd Lots episode breaking down all the different scenarios regarding the oil market. How the different oil markets and products heavily dependent on oil break down. And who benefits the most and who is harmed the most the longer the Strait of Hormuz is closed.

The Trade So Bad, He Faked His Own Death

Subscribe now

Fair Isaac shares were a wild ride this week. Declining over 23% this week.

The catalyst?

On the morning of March 9, TransUnion announced it was pricing mortgage VantageScores at $0.99—down from $4 previously. Hours later, Equifax and Experian followed suit.

Adding fuel to the fire, TransUnion’s CEO disclosed at a March 10 investor day that one of the GSEs had conducted a pilot of securitizing a loan portfolio using VantageScore. At a March 11 conference, TransUnion declined to specify whether the pilot was exclusive to VantageScore or if it also included FICO scores, but he described it as “very modest.”

On the surface, it looks like FICO’s moat is cracking. A competitive score is being priced at pennies on the dollar. A GSE is experimenting with it. The credit bureaus are mobilizing. And the stock just got hammered.

I see it differently, and this selloff is giving investors another opportunity to buy a high-quality company with tremendous pricing power at a significant discount to intrinsic value.

Why $0.99 VantageScores Are a Sign of Weakness

Back in October, TransUnion introduced VantageScore pricing at $4 per mortgage score, a 60% discount to FICO’s $10 sticker price. TransUnion also offered free VantageScores through 2026 for any customer purchasing a FICO score. Equifax and Experian followed with similar offers in the $4–$5 range.

Five months later, TransUnion slashed that $4 price to $0.99. Equifax and Experian immediately matched.

What that means to me is the $4 price—already a steep discount—wasn’t working. Lenders weren’t switching. The bureaus weren’t gaining traction. So they cut the price by another 75%.

This is a pattern I’ve seen repeatedly when trying to compete with companies with strong network effects.

When a challenger has to give the product away to generate adoption, it’s a signal that price isn’t the barrier; the network itself is. The incumbent’s advantage is structural, not cost-based. The ecosystem of users, investors, regulators, and processes built around the incumbent are far stickier than a price advantage could ever overcome.

VantageScore has been around since 2006. It has had nearly two decades to gain traction. Despite being jointly owned and promoted by all three credit bureaus. VantageScore holds a sub-5% market share in the US asset-backed securities market, which is the highest-value use case for credit scores and where the network effects are strongest.

The Credit Rating Agency Analogy

I think the closest historical parallel is what happened with credit rating agencies after the 2006 Credit Rating Agency Reform Act. It opened the door for new entrants to emerge—Kroll, Egan-Jones, and others—but the Big Three (S&P, Moody’s, and Fitch) retained over 90% of industry revenues.

The government designation started S&P, Moody’s, and Fitch’s economic moats, but it was no longer their greatest strength. The network effect among issuers, investors, and regulators was.

I believe the same dynamic applies here. The FHFA’s approval of VantageScore doesn’t dismantle the network that underpins FICO’s dominance any more than the NRSRO reforms dismantled Moody’s or S&P Global’s.

The Securitization Pilot

Read more

Booking Holdings: Why the AI Bear Case Has It Backwards

AI may reshape how travelers find hotels, but Booking’s moat is closing the booking: payments, trust, cancellations, and 24/7 multilingual support for millions of small properties.

Past “disintermediation” attempts (Google, TripAdvisor) failed; AI likely becomes another funnel.

The Wrap – A (Blue) Owl In the Credit Mine?

Blue Owl admits negative headlines drove higher redemptions in private-credit BDCs, showing “headline risk” can become liquidity risk fast.

Probably, why Blackrock also halted redemptions in its private credit funds.

Matthew Ball: State of Video Games

Gaming hit record revenue in 2025 (~$196B), yet private funding collapsed 55% and US player counts fell below pre-pandemic levels.

The paradox: growth is increasingly driven by Chinese publishers and price hikes, not more players.

Netflix Walks, Paramount Wins, and the Ellisons Take Hollywood

An emergency pod update on Paramount winning the bidding war for Warner Brothers and what it means for the future of Hollywood and crown jewel assets like HBO.

• Bill Simmons and Matt Belloni also had a good episode on the winners and losers in this deal.

Why Every App is Getting Worse On Purpose

Subscribe now

Every time a shareholder votes a proxy, receives a fund prospectus, or opens a trade confirmation from their broker, there is an overwhelming chance that Broadridge Financial Solutions processed it. Most investors have never heard of the company. That obscurity is part of what makes it so interesting.

Broadridge is the dominant provider of investor communications and securities processing solutions to broker/dealers, asset managers, and corporate issuers. It has held this position for more than 20 years. Its proxy and regulatory communications business processes over 80% of outstanding shares in the United States. Revenue retention rates sit at roughly 97–98%, and over 90% of fee revenue is recurring.

The obvious question is why this kind of dominance has proven so durable.

Broadridge is not a consumer brand, and its services rarely show up in an investor’s decision-making. Yet it sits at the center of workflows that are compliance-critical, operationally messy, and deeply intertwined with the incentives of broker/dealers and issuers. Those characteristics create a set of competitive advantages that look less like a typical software playbook and more like entrenched financial infrastructure.

Subscribe now

Switching Costs: The Most Powerful Layer

Switching costs are Broadridge’s strongest competitive feature, and they operate across multiple levels of the business simultaneously.

The Proxy Relationship is Structurally Embedded

Most US securities are held in “street name” — meaning the brokerage holds them on the investor’s behalf rather than the investor holding them directly. This creates a coordination problem. Issuers cannot reach their beneficial owners without going through the broker/dealer. The broker/dealer is legally required to forward proxy materials and collect votes. And virtually every broker/dealer outsources this function to Broadridge.

To replace Broadridge, a broker/dealer would need to rebuild these workflows from scratch or migrate to an alternative provider with no meaningful track record at scale. The operational complexity of coordinating communications between issuers, broker/dealers, and millions of beneficial owners — across corporate proxies, mutual fund interims, and ETF filings — makes this a project no compliance team wants to initiate.

Contracts Are Long, and Attention is Low.

Broker/dealer contracts typically run five to seven years. But even beyond contract length, the nature of the service reinforces inertia. Proxy communications are a regulatory necessity, not a source of competitive differentiation for broker/dealers. No one wins or loses client assets by switching proxy agents. When a service is compliance-critical but strategically low-priority, incumbents rarely lose.

The Technology and Operations Segment Adds a Second Layer of Friction.

Beyond the proxy business, Broadridge provides securities processing solutions covering data aggregation, performance reporting, reconciliation, order management, and wealth management platforms for capital markets and asset management firms.

These are BPO-like engagements where switching costs stem from process disruption, integration complexity, and retraining.

The timeline speaks for itself: Broadridge signed its wealth management platform deal with UBS in late 2018 but did not recognize the bulk of that revenue until fiscal 2024. Implementation cycles of that length mean that once a client is live, the friction of moving is enormous.

Add-On Services Deepen the Moat

Broadridge has layered additional products onto its core relationships over time — customer communications and fulfillment for broker/dealers, virtual shareholder meeting tools, analytics, and registered shareholder solutions for corporate issuers. Each additional service increases the scope of re-platforming required to replace Broadridge and makes the overall relationship stickier.

Scale Advantages

Broadridge processes more than 80% of outstanding shares in the United States. That concentration of volume creates a structural cost advantage that makes the business increasingly difficult to attack from below.

The proxy and regulatory communications business has high fixed costs — printing infrastructure, postal optimization, data processing technology, compliance workflows. Broadridge spreads those costs across an enormous base of issuers and broker/dealers. A smaller competitor cannot achieve the same unit economics.

This scale also gives Broadridge the ability to offer an attractive revenue-sharing arrangement to broker/dealers, who receive a portion of the fees that issuers pay. Smaller rivals cannot match this economics, which makes it harder to win broker/dealer relationships in the first place.

And with more broker/dealer clients comes an additional financial benefit: householding.

When multiple accounts at the same household receive the same proxy materials, Broadridge can bundle them into a single mailing and capture additional fees for doing so. That benefit only exists at scale.

The result is a cost structure that reinforces market share, and a market share that reinforces the cost structure.

Barriers to Entry

Even if a well-capitalized competitor wanted to challenge Broadridge’s core proxy business, the market structure presents fundamental obstacles.

The service is a compliance necessity. Errors, missed deadlines, or failed vote collections create regulatory and reputational risk for broker/dealers that far exceeds any cost savings a competitor might offer on fees. When the downside of a vendor failure is an SEC inquiry, institutions do not take chances on unproven providers.

The chicken-and-egg problem is also severe.

To attract broker/dealers, a new entrant needs to demonstrate reliability at scale. To demonstrate reliability at scale, it needs broker/dealer clients. Broadridge has operated this infrastructure without interruption for over two decades. That track record cannot be purchased or replicated quickly.

The regulated fee structure adds another dimension. Maximum proxy delivery fees are set by the NYSE proxy working group with SEC oversight. Because issuers are required to reimburse broker/dealers at the maximum rate, there is no incentive for broker/dealers to negotiate below-market pricing with a new entrant. The fee structure removes the primary lever a disruptor would normally use to gain entry.

The only meaningful competitors in the core proxy business are Mediant Communications, which serves a small number of broker/dealers, and Say Technologies, which Robinhood acquired for its own platform. Neither has made a material dent in Broadridge’s dominant position.

Network Effects

Broadridge’s network effects are more subtle than those of a two-sided marketplace, but they are real and compounding.

As more broker/dealers standardize on Broadridge’s rails, issuers have a stronger incentive to use the same infrastructure to reach beneficial owners. As more issuers route their communications through Broadridge, it becomes the natural default for any new broker/dealer entering the market. Each side reinforces the other’s dependence on the incumbent.

This coordination dynamic is distinct from switching costs, though the two are related. Switching costs explain why individual clients are reluctant to leave. The network effect explains why Broadridge’s value to each client grows as its overall market share increases. A broker/dealer on Broadridge’s platform reaches issuers whose materials Broadridge already processes. An issuer using Broadridge reaches beneficial owners held at every broker/dealer without managing separate integrations. The infrastructure becomes more useful as it becomes more universal.

There is also a secondary feedback loop worth noting. Processing billions of proxy and fund communication deliveries over decades generates operational data that can improve deliverability, reduce return rates, optimize householding, and refine postal logistics. The scale of the data asset reinforces the quality of the service, which reinforces market share, which reinforces the data asset.

Direct Indexing Catalyst

Direct indexing — where an investor owns the individual securities that replicate an index rather than a fund — has been growing rapidly.

Managed account equity position growth has averaged 10–15% annually in recent years. This matters for Broadridge because proxy fees are based on the number of positions, not the dollar value of assets. A $500,000 direct indexing account replicating the S&P 500 generates roughly 500 separate equity positions, each of which requires proxy communications. A traditional S&P 500 ETF holding the same dollar amount generates one position.

As direct indexing scales across wealth management platforms, it acts as a multiplier on Broadridge’s core revenue driver without requiring the company to acquire new clients or expand into new markets. Broadridge simply sits on the infrastructure through which all of these communications must flow.

Why This Moat Holds

What makes Broadridge’s business so durable is that — like most high-quality companies — its competitive advantages reinforce each other.

Switching costs keep clients from leaving.

Scale creates cost advantages that make entry economically unattractive for rivals.

The regulated fee structure removes the pricing lever that a disruptor would normally use.

The market structure ensures that new broker/dealers and issuers default to Broadridge because everyone else already does.

And the network effect means Broadridge’s value to each participant increases as the platform becomes more universal.

Most businesses with durable advantages have one or two of these properties. Broadridge has all of them operating simultaneously in a market that processes the ownership records and shareholder communications for the majority of US equities.

Over two decades of reinforcing each other has created an embedded infrastructure that no one wants to replace and that no competitor can easily replicate.

Subscribe now

Intuit is down almost 50% since initiating a starter position, and then down another 36% after I added to the position and increased its weight to a two-thirds position, based on prior price weakness and my trading rules.

In The Art of Execution, the assassins would have sold out. Given the current market structure and the rise of large pod shops, plenty of assassins are doing exactly that.

I’m not an assassin.

Being the fastest on the button is not my edge, and it is not one I want to cultivate.

I want my edge to be behavioral: the ability to do the hard thing when the market is telling me I’m wrong.

But I also don’t want to be stubborn like the rabbits in the book, doubling down on a losing position that keeps losing because I believe I’m right and the market is always wrong.

Wisdom in crowds exists, but so does madness.

So which is it with Intuit? Is the market right, or is it caught up in short-term AI hysteria?

Either way, Intuit’s recent price action calls for a fundamental review. I want to go through its four major business lines to understand the key AI risks and the counterarguments. Then, using a DCF and a reverse DCF, I want to determine whether current expectations present an opportunity to add to a high-quality business at a discount, or whether the market is right and Intuit is a quality trap?

Subscribe now

Credit Karma

Credit Karma is a lead generation machine.

All of its consumer financial services are free. Credit Karma uses your financial information to provide targeted product recommendations. They earn revenue when you accept a recommendation and sign up with a lender.

Credit Karma doesn’t show random ads. Using their proprietary recommendation engine, Lightbox, they pre-qualify each user for each product they recommend. You know before clicking that you’ll likely be approved. That’s why Credit Karma’s recommendations have approval rates above 95%.

There’s an inherent conflict of interest in Credit Karma’s business model. They don’t get paid for giving you the best advice. They get paid when you take action on a product from one of their partners. If Chase pays a higher referral fee than Capital One, Credit Karma has an incentive to show you the Chase card more prominently—even if the Capital One card has better terms for you.

A personal finance AI agent doesn’t have this bias. It could compare every credit card, loan, or insurance product on the market and give you the best one based on your financial circumstances. If the AI agent could also aggregate each bank’s pre-qualification tools, it would cut out the middleman entirely.

Of all the potential AI disruptions, Credit Karma’s recommendation business seems most immediately at risk.

Intuit recognizes this. They’re looking to turn AI into a growth lever.

OpenAI

Intuit just partnered with OpenAI and is embedding their recommendation engine inside ChatGPT.

If you ask ChatGPT “what credit card should I get?”, the answer could be powered by Credit Karma’s Lightbox engine and proprietary data.

Intuit earns the referral fee regardless of which interface you use.

To be fair, this sounds a little like letting the fox guard the henhouse.

One Consumer Platform

Intuit merged Credit Karma, TurboTax, and ProTax into a single consumer business reporting segment.

The idea is these businesses should cross-sell each other and strengthen each other’s value proposition.

TurboTax captures your income and tax data once a year. Credit Karma uses that data year-round to make better financial product recommendations.

Credit Karma’s year-round engagement helps drive existing and new users to the TurboTax platform during tax season.

This self-reinforcing cycle helped contribute to a full point of tax revenue growth in 2025.

Intuit wants to add AI features to Credit Karma like Debt Assistant that helps craft a pay-down plan and Tax Assistant that pre-fills your tax returns based on year-round activity. The more financial tools that are personalized to you, the stickier the platform becomes and the stronger the cross-selling becomes.

Data Moat

Intuit holds roughly 70,000 tax and financial attributes per consumer and generates 60 billion machine learning predictions daily.

A general-purpose AI can give you generic advice about credit cards. Credit Karma’s AI, powered by your actual income, spending habits, credit history, and tax situation, can give you advice that no outside model can match without that same data access.

Credit Karma’s Lightbox technology already drives approval rates above 95% on recommended products. That kind of precision is only possible because of the underlying data. And that data advantage compounds with every new consumer interaction.

Access to your financial data comes down to trust. Credit Karma has already built this trust. Layering AI on top of their current services and your data can help them become a better recommendation platform and help you with your financial health.

The question is: how long until you feel comfortable giving OpenAI or Claude access to your financial data? Given our history of trusting new technology way before we should, I’d say this will happen quickly.

Mailchimp

Intuit’s $12 billion acquisition of Mailchimp was controversial from the start and remains so today. Critics called the purchase price excessive and a clear misallocation of capital.

So why would Intuit pay that much for an email service provider?

The strategic rationale made sense. Mailchimp was the largest ESP with a massive small business customer base. It could serve as a top-of-funnel acquisition channel for Intuit and QuickBooks Online. Many small businesses start collecting customer emails before they need serious accounting software. Mailchimp could capture these businesses early, then cross-sell QuickBooks Online to users who didn’t yet have an email marketing strategy.

Intuit publicly maintains a 15% IRR hurdle rate for acquisitions. Someone at Intuit believed they could clear that bar even at a $12 billion price tag. With the fed funds rate still at 0% when the deal closed, that probably helped the math work.

But that $12 billion price is now haunting Intuit as it raises prices and cuts back its formerly generous free plan while new AI risks emerge.

The AI Threat to ESPs

You can’t send mass emails directly from Claude.

But you can use AI tools to generate email campaigns, write copy, design templates, and analyze customer data. And you can vibe code your own ESP, host it on AWS, and use an API call to Amazon SES or Sendgrid to handle email delivery.

You’ll save money doing this.

What this cost breakdown doesn’t factor in is the legal and regulatory hurdles, the ongoing maintenance commitments and costs, and the operational “insurance” a dedicated ESP provides.

Regulatory Compliance

The regulatory hurdles aren’t insurmountable, but they require careful attention.

The two major regulations you need to comply with are CAN-SPAM in the U.S. and GDPR in Europe.

CAN-SPAM violations can cost up to $50,120 per offending email.

GDPR violations could cost up to 4% of your global revenue.

Then there are other anti-spam marketing laws you need to build into your system.

This isn’t impossible. Any competent developer with AI assistance should be able to build a compliant system. It just takes time and attention to detail.

Ongoing Maintenance

Once you’ve built your ESP, you need to maintain it.

You need to handle both soft and hard email bounces, manage complaints from major email providers, and keep your list clean by processing unsubscribes.

The biggest ongoing job is maintaining deliverability. What’s the point of building your own ESP if your emails never reach subscribers? The three main factors affecting deliverability are infrastructure, content, and reputation.

Infrastructure is your email’s ID card. When you send an email, the recipient’s email provider (Gmail, Outlook, etc.) wants to verify you are who you claim to be.

Content is how recipients react to your emails. Do they open them? Reply to them? Click links? Pull your emails out of spam? If email providers see poor engagement with your emails, they’re more likely to automatically route them to spam.

Reputation is how email providers score you based on infrastructure and content. High score means inbox. Low score means spam.

Building a good reputation from scratch takes a long time. Each email provider has its own scoring system. You have to check each one separately and adjust your system based on their feedback. Then you have to adjust again when they update their spam filters or when new privacy regulations take effect. Maintaining deliverability could easily become a full-time job by itself.

For many small businesses, Mailchimp is just their ESP. They built their website on a different platform like WordPress and integrated Mailchimp to collect emails. They’d have to vibe code and maintain these integrations along with any other third-party connections. More integrations mean more ongoing maintenance to ensure everything works and syncs properly.

The big risk with a DIY system: if something goes wrong and you get your domain blacklisted, all this work was for nothing and you have to start over.

The Value Proposition of Mailchimp

Yes, you pay more for Mailchimp. But you’re paying them to save you time, to handle all compliance issues, all maintenance issues, all integrations. And most importantly, you’re paying them to ensure your emails consistently get delivered from a trusted source. You pay for access to their development team so everything works behind the scenes and you simply log in, compose a marketing email, and hit send.

The Real Risk

The risk isn’t every business vibe coding their own ESP to save money. The ESP business is already a commodity. The risk isn’t current Mailchimp subscribers leaving either. The data and operational switching costs are too high.

The biggest risk is that Intuit—trying to justify their $12 billion purchase—raises prices too high too fast and pushes potential future customers to other platforms. This slows customer acquisition and future revenue when these customers eventually need to upgrade their subscriptions. Which is exactly what’s played out in recent quarterly calls.

The Premium Repositioning Strategy

But Intuit could also be repositioning Mailchimp as a premium business-focused ESP. They’re moving Mailchimp away from its roots as an ESP for everyone and positioning it as a business growth tool. They’ve also been pushing harder into mid-market companies who have bigger marketing needs and are willing to pay up for more features.

Layering AI into these services could help Mailchimp retain and grow more business clients. Everything I discussed about what AI can do for email marketing currently has to happen outside Mailchimp. Then you have to go into Mailchimp and implement it manually. If Mailchimp had an AI agent layer, in theory you could have it build your campaign, set up automations, and manage your entire email marketing strategy.

Because small business owners are pressed for time, they’re willing to pay up for features that save them time, reduce operational costs (fewer employees needed), and help increase revenue.

Turbo Tax

The AI Threat to TurboTax

Tax filing is fundamentally a structured data problem. You gather documents, extract numbers, apply rules, and submit forms. Claude has demonstrated it can read PDFs, extract structured data, organize it, and perform multi-step workflows. A capable AI agent connected to your bank accounts and W-2 documents could prepare a return without any software. This threatens TurboTax’s value proposition to the basic DIY market—historically its volume driver.

50% of US tax filers still pay someone $200-500+ to prepare their taxes. Each year TurboTax tries to grab market share in this lucrative “assisted” segment. TurboTax Live has helped them move upmarket because many assisted filers just have one or two concerns they want professional help with. TurboTax Live provides that help for a fraction of the cost of full-service preparation.

But now AI is a competitor in this space too. Potential growth into this lucrative market is either greatly reduced or gone altogether.

TurboTax is Intuit’s second largest business with about $4.4 billion in revenue. Market share loss in the DIY segment plus slower gains in the assisted market would significantly impact Intuit’s current and future value.

Why TurboTax Retains Its Value

Preparing taxes is stressful and time-consuming depending on your financial situation. You want them done quickly and painlessly. Most people will pay for this. That’s why TurboTax is so successful and why the assisted tax prep market is so large.

TurboTax has built switching costs into a product that shouldn’t really have any. They do this by increasing the speed of tax completion without sacrificing quality or accuracy. A key part is TurboTax’s ability to import last year’s returns and pre-fill your basic information.

Current TurboTax customers who are used to the interface, speed, and efficiency are unlikely to build their own process in an LLM. Why would they want to spend more time on taxes? And how would they know it’s accurate?

TurboTax has a team ensuring its systems comply with current federal and state tax codes and work properly. How do you know the LLM’s training is up to date on recent tax law changes? LLMs also hallucinate and make things up when given noisy data. How can a filer trust AI not to make a costly mistake?

TurboTax’s Guarantees and Support

Intuit backs TurboTax accuracy with real money.

It offers an Accurate Calculations Guarantee that pays any IRS penalties from calculation errors. ChatGPT and Claude can’t do that, nor would they want the liability exposure.

TurboTax also provides Audit Defense. If you get audited, TurboTax provides enrolled agents who know the tax code to represent you before the IRS. If you use an AI chatbot, you’re on your own.

How AI Can Enhance TurboTax

AI implementation can also help TurboTax increase its speed, accuracy, and personalization. It could include real-time accuracy checks, enhanced prompts for better deductions and credits based on what similar taxpayers have claimed, and an assistant to answer each filer’s specific questions.

TurboTax Live has been successful. As of Q4 2025, TurboTax Live grew 47%—well ahead of management’s 15-20% revenue growth expectations. AI should have the largest positive impact on TurboTax Live.

It’s hard to scale one-on-one professional help, which is why TurboTax has made only small inroads into the assisted tax prep market. But AI tools can reduce the time experts spend preparing a return by about 20%. Every hour saved per return means more returns per expert, which means Intuit can serve more of the assisted market without proportionally growing headcount.

This also allows Intuit to lower TurboTax Live’s price without sacrificing margins, expanding market share and profits.

The example I keep referencing with AI is freestyle chess—human and computer working together. The computer does the calculations and the human oversees the strategy.

Intuit offers Expert Full Service for far less than traditional assisted tax preparation. Using AI to scale this service can help TurboTax gain more assisted market share, increasing revenue.

The E-Filing Advantage

Finally, no LLM is an authorized e-filer. You still have to print your taxes and mail them in. TurboTax is an IRS-registered e-filer.

Will LLMs eventually try to register as e-filers? Maybe, if the rules change. Right now only a real person can apply. But if they do, they’re taking on increased compliance costs, data security costs, anti-fraud costs, and overall liability. Do they want that, or would they prefer to sell their LLM to a company that can layer it on top of existing compliance infrastructure?

QuickBooks Online (QBO)

QuickBooks is Intuit’s biggest revenue source, generating about $12 billion.

Like TurboTax, QuickBooks is built around the basics: categorizing transactions, reconciling bank feeds, generating profit-and-loss statements, creating invoices, and tracking expenses. These are exactly the kinds of structured data workflows LLMs and AI agents are good at.

The concern is not that QBO gets replaced overnight. Switching costs are too high for a small business already running on QBO to risk an unproven, vibe-coded DIY system just to save $30–$200 a month.

The Real Threat: Future Customer Acquisition

If you’re a solopreneur starting a landscaping business, do you really need a $30–$200 monthly subscription just to categorize fuel purchases and truck payments?

Claude or ChatGPT can read a bank CSV, categorize each transaction, and generate a clean P&L in minutes. You can even build a dynamic financial statement in Excel with Claude’s extension.

So Intuit either fails to win these new customers and loses future growth, or it has to spend more to win them, which compresses margins.

There are an estimated 30 million micro businesses in the U.S., including solopreneurs, freelancers, gig workers, and side hustles. Even $30 a month for QBO Simple can be a meaningful expense.

QuickBooks is utilitarian. It meets most needs for most businesses, but it will not meet every need for every customer. Technically savvy business owners could build custom financial management tools. Or a developer could build a consulting business that helps small businesses spin up an accounting system that pulls bank data via Plaid, uses AI to categorize transactions, and generates reports tailored to specific needs.

That is not most business owners today, but the population capable of doing this will grow as the tools get easier to use.

The Accountant Channel: Strength and Vulnerability

QBO’s most powerful distribution channel is the accountant channel.

Approximately 600,000 accountants in the U.S. are trained on QuickBooks and recommend it to clients. Businesses use QBO because their accountant prefers it, and accountants prefer QBO because most clients use it.

But the Illinois CPA Society believes bookkeeping is likely to be one of the most disrupted functions due to AI in the coming years. Again, double-entry accounting is a set of rules applied to structured data.

Fewer accountants means fewer QBO recommendations to small business owners. Without a nudge from an accountant, a small business might choose a QBO competitor or handle it themselves with AI’s assistance. Then QBO’s two-sided network, one of its strongest competitive advantages, begins to unravel.

Domain-Specific Intelligence: QBO’s Structural Edge

QBO has domain-specific intelligence. The platform is trained on hundreds of millions of small business financial transactions across every industry and vertical. AI trained on this data will enhance its value proposition.

For example, a general contractor pays United Rentals $4,500. A general AI sees an equipment company and categorizes it as an equipment expense, but the treatment depends on the job. A backhoe for a specific client job is a direct job cost and should be labeled as such to properly analyze job profitability. But a generator for the contractor’s office trailer is an overhead expense. Same vendor, same transaction type, different accounting treatment.

Or consider an e-commerce brand that receives a deposit from Shopify Payments for $12,450. A general AI may log it as revenue. But that $12,450 is gross sales minus Shopify’s processing fees. It may also include sales tax collected that is not revenue at all. It is a liability owed to state tax authorities. Correctly entering the deposit requires decomposing it into gross revenue, processing fees (an expense), and sales tax payable (a liability). QBO integrates with Shopify and is trained on millions of these transactions. It knows to automatically unbundle the deposit and route each component to the correct account.

From Accounting Software to Virtual Business Team

With its vast trove of data, QBO with an AI agent can transform from accounting software into a virtual business team.

The agent spans multiple functions: Accounting Agent (reconciliation, anomaly detection, P&L generation), Payments Agent (cash-flow optimization, invoice tracking), Finance Agent (forecasting, scenario planning), Customer Agent (lead tracking, pipeline management), and Project Management. It operates autonomously and executes multi-step workflows without intervention.

The 2.6 million customers already using these agents have reported saving 12 hours per week and getting paid five days faster. A shorter cash conversion cycle can make a business more agile and give the owner more room to invest quickly and take advantage of growth opportunities before competitors.

QBO’s pitch shifts from accounting software to a team of AI specialists that runs and streamlines the back office, so owners can spend more time on the work that grows the business.

Moving Upstream: The Mid-Market Opportunity

Layering AI agents onto the platform should also help Intuit move upstream into mid-market businesses. Mid-market customers need more complex solutions and come with a higher compliance burden. But they also have higher switching costs, higher revenue, and longer lifetime value.

Intuit estimates there are 1.7 million such businesses, defined as those with revenue between $2 million and $100 million. 800,000 already use QuickBooks.

By using AI to customize Intuit Enterprise Suite for specific verticals, Intuit hopes to capture more share within the mid-market. It recently launched an ERP specifically for construction, a $2 trillion industry.

And rather than weakening accountant distribution, AI could strengthen it. QuickBooks with AI can make accountants more productive and allow them to handle 2–3x more clients. Accountants are unlikely to shift away from the platform that drives their business and increases profitability. They would lean further into promoting it, which increases the number of small businesses using QBO.

Finally, Intuit’s development teams will use AI to increase productivity and ship features faster. QuickBooks will not stand still. It will improve faster and become easier to customize for each business’s needs, strengthening its value proposition and cementing its role as a central nervous system for small and mid-market businesses.

Please upgrade your subscription to paid if you would like to see my updated valuation assessment and portfolio action plan.

Valuation & Market Expectations

Read more

LLMs vs Marketplaces

Marketplace defensibility depends on three factors: how hard supply is to aggregate, how managed the marketplace is, and the nature of customer engagement (frequency and consideration).

Response to Centrini and Stablecoin Impact

Payments are a trust network, not just a rail.

Stablecoins may speed settlement, but Visa/Mastercard still win on global acceptance, rules, fraud + dispute handling, and issuer/merchant incentives.

It’s February, and I am writing about… heat waves

An interesting way to look at the economic costs of climate change with the rise of sick leave for those with pre-existing conditions during heatwaves.

“a three-day heatwave costs the German economy an estimated €32.2m in ‘lost wages’ and increases health care costs by 0.22% for each day in the heatwave.”

Interview with David Kim a.k.a Scuttleblurb (2026)

I always enjoy these annual interviews with Scuttleblurb on the state of the research driven newsletter business.

Space Data Centers Are Dumb

Subscribe now

Read more