Bloom’s Big Lie

Bloom claims it has “no China supply chain” for scandium, an essential fuel-cell input. This investigation traces multiple China-linked routes into Bloom’s materials and shows the scandium math may not support its growth story.

How Serial Acquirers Turn Into Multibaggers

Buy small private firms at low multiples, reinvest cash flow into the next deal, and compound for decades. Their edge is disciplined capital allocation, decentralization, and culture.

Incentives Work: Cocaine Edition

A policy meant to curb coca did the opposite. Colombia’s promised subsidies encouraged farmers to plant more, turbocharging cocaine supply. Lower production costs and new Ecuador shipping routes pushed more product to Europe, where seizures rose, and street prices fell.

Aircraft Leasing Lift-Off

Aircraft leasing and its capital cycle. And why earnings are set to inflect and a major reason why Air Lease was bought out recently.

Why American Houses Are So Flimsy

Subscribe now

On May 28, Anthropic closed a $65 billion round at a $965 billion post-money valuation, overtaking OpenAI’s $852 billion. Four days later, it filed confidentially for an IPO. Bankers expect a listing sometime between late 2026 and the first half of 2027.

At that valuation, it’s worth asking, does Anthropic have an economic moat?

Below, I run through switching costs, barriers to entry, scale advantages, and potential network effects to see whether there’s an economic moat, or at least the beginnings of one, that could justify a likely $1+ trillion public debut.

Subscribe now

Rented vs. Owned

As I evaluate Anthropic’s emerging competitive advantages, I keep coming back to a simple question: what is rented, and what is owned?

• Anthropic’s compute is rented (Nvidia, Amazon, Google).

• Its talent is rented, and the lease is short. Every day, employees get lucrative offers to leave.

• Its capital is rented. None of the leading LLMs are self-funding; they run on investors who eventually want their money back, plus a return.

• Its distribution runs through Bedrock and Vertex, rented from two companies that also ship rival models.

Almost every advantage on the list is leased.

That’s not a criticism. Many companies start with rented advantages. Boston Beer Co. (Sam Adams) rented brewing capacity (contract brewing) before it had the capital to own a brewery.

The question is whether the rented advantages convert into owned advantages before the lease runs out?

Switching costs

A while back, I wrote a post about the different switching costs and ranked them by strength. Here, I’ll run Anthropic through the same framework.

From weakest to strongest:

1. Financial

2. Time-based

3. Relationship

4. Technical/operational

5. Ecosystem-based

Financial costs

Financial switching costs are largely absent. It’s twenty dollars a month, no termination fee, no equipment. Competitors can discount aggressively or offer incentives to pull users away. This is the weakest rung, and for Anthropic, it barely exists.

Time-Based Costs

Time-based costs exist, but they range from minimal to annoying. A lot ports over: skills, memory, files. What doesn’t port cleanly is fine-tuning and the tacit craft that emerges after many iterations. The core transfers; the finer details often land close enough that you don’t notice what broke until you’re mid-workflow and the output is subtly wrong.

Then you’re stuck reviewing dozens of skills to find the failures. The initial hurdle can feel small, but the costs can spike after you’ve already switched. At that point, you face a sunk-cost choice: finish the changeover and re-tune, or retreat to the model you left.

Relationship Costs

Relationship costs don’t exist unless you’re one of the people who has succumbed to AI psychosis and fallen in love with a chatbot.

Technical/Operational Costs

The Inverted Learning Curve

The main technical hurdle is the learning-curve trap. The old saying applies: you can’t teach an old dog new tricks. Once you have mastered a piece of software (or your organization has), switching can be prohibitive because you have to relearn workflows and absorb a productivity shock.

LLMs invert this problem.

They are easy to start using. You can text or talk to them directly. You don’t need a training course to ask a question, and the product teaches you as you go. Not sure how to create a skill? Ask it. Not sure what the next step in a workflow is? Ask it.

There is still a learning curve in getting better at prompting, but it’s manageable. Once you master prompting in one LLM, much of that skill transfers to others.

The Enterprise Break

Everything changes once the user count goes above one.

A single person re-tuning a dozen skills is an annoying weekend. An organization revalidating four hundred skills is a project nobody wants. By then, the model often stops being a tool that a single employee is responsible for. It runs unattended, checking code before it ships and interfacing with internal systems through a web of connectors.

As the shared corporate context accumulates inside the model, it becomes harder to transfer. This is the classic data trap.

Ecosystem

The highest ranking switching cost is the ecosystem, and this is what Anthropic is targeting

Let’s look at what shipped last quarter.

Cowork launched in February. It sits on your desktop and can run all sorts of daily tasks for you unsupervised. Claude embedded itself inside Excel, PowerPoint, Word, Outlook, Slack, Gmail, and Google Drive. Everything you need to get your work done and to increase your productivity. Now Claude can follow one conversation across an email, a spreadsheet, and a deck without being re-briefed at every transition.

This shared context is their efforts to build an ecosystem.

For example, an analyst pulls comparables from an open workbook, builds the comps table in Excel, drops the valuation summary into the pitch deck, and drafts the note to their MD. The dataset gets explained once. Every application Claude touches makes the next one more useful, because the context carries across the workflow.

Then there’s the marketplace.

Anthropic gave enterprise administrators private plugin marketplaces. These are internal app stores, sourced from a company’s own code repositories, with admin control over what employees can install.

Anthropic shipped prebuilt plugin templates for HR, engineering, operations, financial analysis, investment banking, equity research, private equity, and wealth management. And it put Skills into the Excel and PowerPoint sidebars, so a firm can save its proprietary analysis or its approved slide template as a one-click action for every employee.

It’s like skills but scaled up to a firm-wide level, with an admin console bolted on. Employees don’t own the skill and can’t move it to their preferred LLM. They have to stay within the firm’s ecosystem built on Anthropic.

The Landlord Problem

But there is one problem with the strategy right now.

Claude in Excel is an Office add-in. Excel belongs to Microsoft. Microsoft is an investor in Anthropic and wants its investment to do well. But Microsoft also has Copilot, which competes directly with Claude.

The ecosystem is being built on apps it doesn’t own. The landlord is friendly, for now.

The plugin architecture is similar.

It’s API-based and broadly compatible with the specification everyone else already uses, so integrations built for other systems port over without much work. That fills a marketplace fast.

But it’s the MCP issue all over again.

The one piece Anthropic owns is the shared context layer. The thing that holds your workbook, your deck, and your thread in mind simultaneously. This is proprietary, and nobody else has shipped a competitor.

Yet.

Almost everything else I’ve described above is scaffolding. Skills, memory, connectors, plugins. It sits on top of the model, and it is the thinnest layer in the entire stack.

It is also exactly what a competitor’s migration tool would target first.

Network Effects

Similar to switching costs, I also ranked six types of network effects, from weakest to strongest.

Direct Network Effects

My initial thought was that Anthropic might benefit from direct network effects: more users create more training data, which improves the model, which creates more value for every new user.

But I think that’s wrong.

Frontier capability comes from compute, reinforcement learning, and synthetic data, not from increased user activity. The counterexample is DeepSeek. Its users and traffic are a rounding error relative to Anthropic’s, yet it still achieved meaningful capability. The same is true of Mistral and Qwen.

Anthropic can train on consumer conversations by default. But its commercial products (Claude for Work, the API, Bedrock, Vertex) are contractually excluded from training. The data Anthropic can learn from is mostly retail consumer chatter, which is comparatively low value.

The data that might actually create compounding effects, enterprise workflows and context, is walled off by the privacy guarantees that help Anthropic win those deals.

Data Network

At best, this looks like a data network, which ranks near the bottom. It’s also vulnerable: competitors with better technology can win quickly, and in AI, that happens often.

There isn’t a true direct network effect here. There is a narrower effect where usage generates failure telemetry and evaluation sets. That can improve the product, but it doesn’t reliably improve core capability, and it doesn’t compound the way search queries once improved Google’s index.

Indirect Network Effects

An indirect network effect occurs when more usage of a product attracts more valuable complements, which then make the original product more valuable.

The classic example is Intuit’s QuickBooks Online. With a large installed base, it becomes worth ADP’s time to build payroll integrations, and PayPal’s and Square’s to build payments. Those complements make QBO more useful, which attracts more users, which attracts more complements.

AppFolio ran the same play. Open the API, let third-party tools integrate, and each new connection makes the platform harder to rip out.

The LLM version looks similar. Notion, Canva, Gmail, PitchBook, Slack, Stripe, and thousands of other vendors have shipped MCP servers that plug their software directly into models.

SDK downloads reportedly went from roughly 100,000 in the first month after launch to 97 million a month by March 2026. The complements are arriving exactly as the framework predicts.

But there’s a problem.

ADP built to QuickBooks’ API. That integration works with QuickBooks and nothing else.

MCP is not exclusive.

Anthropic published it in November 2024. OpenAI adopted it in March 2025, Google in April, and Microsoft announced support at Build. In December 2025, Anthropic donated the protocol to the Linux Foundation’s Agentic AI Foundation (backed by AWS, Google, Microsoft, OpenAI, Bloomberg, and Cloudflare).

So when a software company like Notion ships an MCP server, it works with Claude, ChatGPT, and Gemini at close to zero incremental cost to Notion.

Anthropic built the foundation for indirect network effects, then gave away the exclusivity that would have allowed those effects to accrue primarily to Anthropic.

I understand why.

In 2024, no one had enough market share to force a proprietary standard. An open protocol that everyone adopts beats a closed standard that no one uses. MCP helped induce demand for the whole industry.

The top three network effects(social, multi-party, two-sided marketplace) don’t apply here.

Anthropic’s best shot at network effects was a protocol network. But as discussed, it donated that advantage to accelerate broad adoption.

Barriers to Entry

The barrier everyone thinks of first is capital.

Leading AI labs are burning cash, and there is a finite pool of capital. Why wouldn’t the largest, most likely winners absorb the money and starve everyone else?

But capital is still flowing. Frontier Labs hasn’t had trouble raising. Neither has Mistral. Even xAI was able to raise before its merger with SpaceX.

The real barriers are less headline-friendly.

Compute Allocation

The constraint isn’t simply that a startup can’t afford chips. Supply is so tight that many teams can’t get chips at any price. Allocation is relationship-mediated, and it flows to counterparties with long-term commitments and a high likelihood of existing in five years. Suppliers ration to likely survivors. That becomes a self-reinforcing cycle.

Talent Concentration

The number of people who have personally run a frontier-scale training program is measured in the low thousands globally. Compensation to secure that talent is soaring. Meta reportedly paid one person $250 million.

Trust and Compliance

SOC 2, HIPAA, FedRAMP, indemnification, and model-behavior guarantees are boring, but hard to replicate quickly. Reputation for safety is built one company at a time.

Distribution

Bedrock and Vertex are the enterprise front door.

A counterargument is that these barriers mostly protect the frontier. That matters only if the bulk of value stays at the frontier. If “good enough” keeps getting cheaper, the key question is whether marginal customers will keep paying for frontier capability.

Scale Advantages

Scale today means a training run amortized over far more inference, better unit economics per token, and inference optimization that smaller competitors can’t match.

Tight compute supply also advantages the largest players. Suppliers prioritize the labs with the best odds of long-term survival and the most reliable revenue, reinforcing the cycle described above.

But again, Anthropic owns no fabs, chips, or data centers. It rents all of it from companies that also fund competitors and, in the case of Google and Amazon, sell rival models to the same customers.

Anthropic may have real scale advantages, but they sit on leased real estate.

Token costs are also falling, so it’s worth asking how durable that cost advantage is over time.

The Moat

My initial gut reaction was that Anthropic might develop an emerging moat built on high barriers to entry (mostly capital constraints), scale, and potentially both direct and indirect network effects.

After walking through the factors, what I’m left with is narrower: an emerging moat built primarily on switching costs in the B2B enterprise market, supported by scale and barriers to entry, but not the capital constraints I assumed would matter most.

We’ll learn more when the S-1 drops and regular financial disclosures begin.

Subscribe now

The S&P Global Spin-Off

S&P Global completed the spin-off of its Mobility segment on July 1. The new company is Mobility Global, and it trades on the NYSE under MBGL.

If you owned SPGI at the close on June 15, you received one share of MBGL for every share of SPGI you held—one for one. Fractional shares were sold and paid out in cash.

We didn’t sell anything. We still own the same S&P Global, and we now hold Mobility Global as well, trading around $20 a share.

The Business

Mobility Global is mostly Carfax.

Carfax is about 65% of revenue and 75% of adjusted EBITDA. The rest is a B2B segment selling data and software to automakers and dealers.

Carfax sells vehicle history reports. It pulls data from more than 177,000 sources—DMVs, police departments, auction houses, service shops, insurers—and holds more than 38 billion records. It took decades to build this database, and a new entrant can’t simply replicate that data with more funding.

Adjusted EBITDA margins are near 40%, and subscription revenue is more than 80% of the total. Because the business is tied to the used-car and dealer market, it holds up whether or not new cars are selling. Roughly 80% of revenue is recurring and subscription-based.

Why the Stock is Cheap

MBGL is a roughly $6 billion company.

That’s too small for the S&P 500, so funds that track the index can’t hold it. It also no longer belongs in the financial sector where S&P Global sits, so funds built around financial-sector benchmarks can’t hold it either. Both groups have to sell, regardless of what they think of the company, though selling from S&P 500 tracking funds is the bigger pressure of the two.

This is mechanical selling, and it explains most of the recent price decline post-spin-off.

My fair value estimate is around $28 per share, and some sell-side reports I’ve seen have target prices in the mid-$20s.

Risks

I see three main risks to Mobility’s business: one near-term and two longer-term.

The immediate risk is the balance sheet.

Mobility Global went public with about $2 billion of debt. That’s 2.8 times gross leverage against management’s 2.5 times target. The business generates enough cash to carry it, but it’s a meaningful debt load for a new standalone company, and it’s above where management wants to be.

The counterpoint is that S&P Global didn’t saddle the company with so much debt that it threatens its status as a going concern. It also means Mobility’s equity is a small sliver of its capital structure. When enterprise value improves, it can produce outsized percentage gains in equity value. This is essentially a leveraged recap play through a spin-off.

The longer-term risk is the concentration in Carfax. Experian’s AutoCheck is cheaper, and Carfax’s move into listing marketplaces gives Experian a bigger opening to compete there.

The even longer-term risk is self-driving cars and slowing population growth, shrinking the used-car market.

Why I’m Holding MBGL

The price is below most target prices, and—more importantly—below my estimate of fair value. It got cheap for reasons that don’t reflect the business.

While the business is not as durable or impregnable as its parent, S&P Global, it is still a dominant player with recurring, high-margin revenue that’s protected from cyclical car sales. It has a tangible data moat and intangible brand awareness.

I’m not adding yet. I want a couple of quarters of standalone reporting before I decide what I want to do, but it’s a good business trading at a cheap price.

The Honeywell Split

Honeywell completed the split of its aerospace business on June 29. The old company is now two: Honeywell Technologies (HON), the automation business, and Honeywell Aerospace (HONA).

For every two Honeywell shares you held on June 15, you received one share of Honeywell Aerospace.

Honeywell then ran a 1-for-2 reverse split on the shares you kept, so your HON count halved, and the price roughly doubled. Net of both steps: every two old Honeywell shares are now one Honeywell Technologies share plus one Honeywell Aerospace share. Fractional shares were paid out in cash.

This finishes the three-way breakup that started with Solstice Advanced Materials last October.

The Businesses

Honeywell Aerospace

Honeywell Aerospace (HONA) sells mission-critical systems that aircraft manufacturers can’t build planes without. The cleanest example—and its most important franchise—is the Auxiliary Power Unit (APU).

The APU is roughly a $1 million part on a $100 million jet. Replacing it would take years and require a new FAA certification. Once an APU is designed into a plane, switching becomes extremely costly.

As mentioned above, each APU is deeply integrated and interfaces with the flight management systems, environmental control systems, and engine starting mechanisms. Switching to a different APU manufacturer would require extensive re-engineering of these systems for the aircraft manufacturer.

Then they would need to run extensive and expensive tests on the new APU system. Every aircraft must obtain a Type Certificate from the FAA demonstrating that the aircraft design meets all safety standards. The APU is part of that design.

Switching from a Honeywell APU to a new APU requires major modifications to the Type Certificate and a new round of testing, documentation, and regulatory approval. This process could take years and cost millions of dollars.

Then on top of this sits a razor-and-razor blade aftermarket: sell the equipment once, service it for decades.

Life-limited parts like the APU must be replaced after a certain number of cycles, flights, hours, or years, and it’s almost impossible to source them from anyone other than the OEM.

The FAA requires that any replacement part on a certified aircraft come from the OEM or have its own Parts Manufacturer Approval (PMA). To obtain a PMA, a company must demonstrate that the part is identical to or better than the original. This requires exhaustive testing to prove compatibility and integration with the rest of the system—a process that takes significant time and money.

“Generic” aircraft replacement part companies like Heico tend to reverse engineer the parts they compete against, which is far easier for simpler components. Life Limited Parts are usually the most technically sophisticated components and rely on proprietary information that OEMs won’t share.

The business earns around 30% on invested capital, has a disciplined capital-allocation record, and has the kind of moat this newsletter exists to own.

It’s trading around $237 as I write this, versus my updated estimate of fair value of $280 per share.

One more point worth noting: HONA is joining the S&P 500 and the S&P 100, which makes index funds forced buyers.

Honeywell Technologies

Honeywell Technologies (HON) is the automation company left behind. It covers building automation, process technology, and industrial automation. These are fine businesses, but this is the half whose uneven results dragged down the company for years, and it carries the goodwill from a long, mixed acquisition history.

It trades around $224 as I write this, which is above my updated estimate of fair value of $160 per share.

Why We Bought Honeywell

Read more

AI Moats and Value Traps

A low multiple is not the same as cheap when a business is being disrupted because the price has already moved but the fundamentals haven’t yet. So the current multiple is measured against earnings that are about to erode.

How Serial Acquirers Turn Into Multibaggers

Serial acquirers can be a repeatable “multibagger” model. Buy many small private businesses using strong free cash flow, improve them, and reinvest the cash into the next deal. The great ones, the multi baggers, share common traits.

Stop Buying Memory Stocks

HBM demand for AI is creating a temporary packaging/yield shortage that markets are pricing as permanent. But as packaging yields improve and $100B+ of new capacity comes online, memory’s commodity cycle returns, and today’s peak-cycle valuations can unwind fast.

Rory Johnston on Why His $200 Oil Prediction Didn’t Turn Out Right

TL;DR China.

WeWork - The $47 Billion Disaster

Subscribe now

In the spring of 1848, Sam Brannan walked through the streets of San Francisco holding a small vial of gold dust above his head, shouting that there was gold in the American River.

Brannan wasn’t a prospector and he wasn’t sharing this detail out of the kindness of his heart.

He wanted to spark gold fever because weeks earlier, he bought up every shovel, pan, and tin of provisions he could find in the region.

He purchased metal pans for twenty cents and then sold them in his store for $15 as the California Gold Rush took off.

The miners chased the gold and Brannan sold them the tools to chase it with. Only one of them reliably made money.

Brannan became California’s first gold-rush millionaire without ever working a claim.

You could say we’re living through a similar moment right now. There is the obvious AI gold rush and we own a picks and shovel play in this. But there is also a biological gold rush that’s been overshadowed by the AI story.

Over the past decade, capital has poured into a new generation of medicine: engineered antibodies, mRNA, cell and gene therapies, proteins brewed in bioreactors instead of synthesized in a lab.

The promise is enormous but the success rate is not.

For every blockbuster biologic, a graveyard of programs runs out of money before they reach a patient.

But the key thing to remember about a gold rush is you don’t have to know which miner will strike it rich to make money. You just have to be the one selling them the shovels.

This month’s recommendation is a company that sells the shovels to the biologics industry.

It supplies the resins, filters, media, instruments, and single-use kits that nearly every biologic drug — approved or experimental — has to flow through to get made.

And right now it’s trading about 30% below my base case estimate of fair value because the market is still fixated on its post-pandemic inventory hangover that is clearing.

Read more

Inside TCI’s Conference: The Anti-AI Portfolio

Under AI, long-standing software/service moats may erode faster, making winner/loser selection harder and pushing TCI to be more cautious. TCI has shifted toward “defensive” businesses with hard assets and low disruption risk (e.g., aerospace with multi-decade backlogs).

It’s Time To Rotate Into Platform Businesses

AI-linked names command large index weightings and premium multiples vs. the rest of the market. But instead of going to cash, it is time to rotate into durable platform businesses whose valuations have boom even more attractive.

10 Signs of Exuberance

Leveraged theme ETFs, 13F-driven spikes, vibe-based price targets, and nonstop YTD flexing to name a few.

Part One: The Rise of the “Activist” Investor

A different angle on the rise of Bill Ackman by the Behind the Bastards team

How a Secretive Trading Empire Is Taking Over Wall Street

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 33 companies with recent insider buys are listed below.

Jack Henry (JKHY) is an interesting new name making the list. It too is caught up in the AI induced SaaSpocalypse but it serves credit unions and regional banks, two very conservative customer bases, with built in high operational and regulatory switching costs.

The full screen is available below to paid subscribers.

Read more

The Whirlwind is Upon Us

We haven’t seen stock dispersion like this since 1999.

Ensign: The Nursing Home Empire Built on Fatal Neglect

Is Ensign (ENSG) boosting profits by staffing nursing facilities below what sicker patients need while still collecting public reimbursements?

Organs Through a Subscription!?

TransMedics turned a transplant device into “Transplant as a Service”: fly teams and organs on its own jets, charge ~$120k per case, and lock in hospitals with logistics switching costs.

He Wrote the Book on Bubbles | Edward Chancellor on If AI is Different

Why transformative technologies can still produce terrible returns for investors.

Why Fast Food Got So Expensive

Subscribe now

LPL Financial ($LPLA) is the largest independent broker-dealer in the United States, supporting more than 32,000 financial advisors and roughly 11 million customer accounts, with about $2.4 trillion in advisory and brokerage assets on its platform at the end of 2025.

It is not an asset manager.

It provides the plumbing to independent advisors to run a modern wealth management practice: technology, clearing, custody, compliance, research, and practice-management support.

The bulk of its profits come from advisory fees, commissions, and interest on client cash.

Its dominant position is built on three main factors: immense switching costs, scale advantages, and barriers to entry.

Subscribe now

Switching Costs

Switching costs are LPL Financial’s single strongest competitive advantage.

LPL has retained roughly 97% of platform assets on an average annual basis over the past decade. This translates into a useful life of client assets around 35 years.

Advisors and their clients, once on the platform, simply do not leave.

Why don’t they leave?

Because leaving is genuinely expensive, on several fronts.

The first cost is client attrition.

Industry data from Cerulli suggests that advisors who change firms lose upwards of 20% of their assets in the transition—clients who don’t want to make another move and accounts that fall through the cracks during repapering.

When your revenue and earnings are based on assets under management, losing clients is a large haircut to earnings and it could take years to replace those lost assets.

The second is the economics of the relationship.

LPL’s average advisor manages about $74 million in assets and generates roughly $470,000 in annual revenue. At LPL’s payout rate of nearly 87%, that advisor keeps around $416,000 before their own overhead. If that same advisor jumped to a wirehouse, where captive payouts cluster near 45%, they’d take home closer to $188,000. The independent model pays more than twice as much and a competitor would struggle to replicate those economics.

The third cost is operational entanglement.

LPL layers on roughly $21,000 per advisor of a la carte services. This includes marketing support, bookkeeping, paraplanning, integrated technology for running a practice. Leaving means an advisor would lose immediate access to this support harming how the business is ran and then they would have to spend time and money finding sourcing replacements from multiple vendors.

Leaving means an advisor would lose immediate access to this support, harming how the business is run, and then they would have to spend time and money sourcing replacements from multiple vendors.

And the only off-ramp that increases payouts is going fully independent as a registered investment advisor. But this means SEC registration and building a compliance and risk infrastructure from scratch. This is a daunting proposition for the typical advisor.

LPL even takes care of the most natural form of attrition, retirement. LPL has a liquidity and succession program that transitions retiring advisors’ books to other advisors and keeps the assets in-house.

All these stacked together create immense switching costs for an advisor.

Scale and Cost Advantage

As the largest independent broker-dealer in the country — after absorbing Commonwealth Financial Network — LPL spreads its fixed costs (technology, clearing infrastructure, back office) across a far larger base than any direct competitor.

Measured the same way a custodian like Schwab would report it, LPL’s expense on client assets runs around 0.20%, well below large banks such as Morgan Stanley (0.33%) and Bank of America (0.41%). Schwab’s is 0.11%. LPL is, in its own words, the lowest-cost solution for independent advisors on the market.

Taking a play book from Costco, LPL chooses to pass much of that scale benefit back to advisors in the form of industry-high payouts rather than keeping it as margin.

This is their flywheel: greater scale lowers the unit cost to serve each advisor, which funds the highest payouts in the business, which attracts and retains advisors, which adds scale.

The result over the past decade is an advisor headcount that has compounded at about 8.7% annually. Industry-wide advisor growth is roughly 0.3%. This means LPL is not only growing with the market but taking market share from competitors.

Passing these savings on through to its advisors instead of maximizing near-term profit, leaves little room for a well-capitalized competitor to undercut LPL. You might have to destroy your own economics to beat LPL on price.

Barriers to Entry

The barriers to entry in this business are high and rising.

There are a lot of regulations in this business. Operating a self-clearing broker-dealer and an RIA at national scale requires registration, net-capital reserves, and a compliance apparatus that is expensive to build and expensive to maintain.

Then there are financial barriers to compete. LPL reinvests north of $465 million a year into its technology platform, smaller independent broker-dealers cannot match this spend without eroding their profitability.

Probably the largest barrier is the economics of its advisor deals.

To lure LPL’s advisors away, a challenger would need to match LPL’s nearly 90% payout and its low cost base and its breadth of services. Smaller independent peers can’t afford the payout; wirehouses won’t cannibalize their captive model; and pure-play custodians like Schwab and Fidelity are aimed at a different customer. Meaningful encroachment on LPL’s core market looks unlikely, which is exactly why the firm keeps compounding share.

Network Effects

It’s tempting to describe LPL’s flywheel as a network effect, and you’ll see it called one, but it isn’t. The distinction matters for assessing the strength of its competitive factors.

A true network effect compounds because each new user makes the platform more valuable to other users. Think of a marketplace or an exchange, where more buyers attract more sellers, which attracts more buyers.

LPL doesn’t work that way. One advisor joining LPL doesn’t make the platform meaningfully more valuable to another advisor; advisors don’t transact with each other.

What LPL has instead is a scale flywheel: more advisors lower the per-advisor cost of running the platform, which funds better economics, which attracts more advisors. The loop has the same self-reinforcing shape as a network effect, but a different engine driving it: cost economics, not cross-user value.

That’s why LPL’s strongest competitive factor is its switching costs plus cost advantages, not network effects.

Putting It Together

Scale produces the cost advantage; the cost advantage funds the payouts and services that build switching costs; switching costs lock in the assets that sustain the scale.

This flywheel is why LPL’s adjusted ROIC has averaged around 20% over the last five years and consistently exceeds its cost of capital.

Also, the independent advisor industry is fragmented and consolidating. Being the lowest-cost platform with scale advantages makes LPL an attractive home for advisors—and an acquirer of other competitors (LPL just bought Commonwealth for $2.7B)—to plug into its flywheel. That gives the company a long runway of growth and reinvestment to strengthen and widen its “moat.”

Subscribe now

6 Charts on SpaceX’s Pre-IPO Financials

Next week is a big IPO week for SpaceX. Morningstar put together 6 charts on how the company makes makes, how it loses massive amounts of money, and how much key insiders cash in from the IPO.

How Turkey Hacked the Hair Transplant Industry

I’m sure you’ve seen the pics and memes about Turkish Hair Lines. But Turkey’s dominance in the hair transplant medical tourism came from a series of calculated moves and technology improvements.

The Rikishi Moment

Why the Nasdaq changing its rules for index exclusion to add SpaceX is like the Pete Rose getting repeatedly smashed in the face by the gigantic ass of Rikishi at Wrestlemania.

Think and Grow Rich - If Books Could Kill

Who would have thought that the granddaddy of grifting self-help, get-rich books was written by a lifelong crook?

The Inference Shift | Stratechery by Ben Thompson

Subscribe now

It was the best of quarters. It was the worst of quarters.

It was the best of quarters because Assisted Tax, aka TurboTax Live, grew 36%, Intuit/QuickBooks’ mid-market expansion grew nearly 40%, and the money platform grew 30%. Management raised full-year guidance for the second time this fiscal year, bought back a record amount of stock, and raised the dividend 15%.

It was the worst of quarters because CEO Sasan Goodarzi admitted Intuit “lost on price” at the low end of the consumer tax base, total IRS filings shrank for the season, the company announced a 17% reduction in force, and the $12 billion Mailchimp acquisition was demoted to cash management.

The market overwhelmingly sided with the second story.

The stock fell roughly 20% on the print and now sits around $302 per share. Far below its 52-week high of $813.70

Intuit’s fiscal third quarter contained two companies. A quality business solidifying its market position and growing by moving upmarket, and a business under AI-and-pricing pressure at the low end.

I’ve written before that I think the market is overstating the AI disruption risk to Intuit and underweighting AI’s potential to reinforce its position. I added to the position on the way down, bringing it up to a two-thirds weight. Then they released their Q3 2026 earnings and the stock dropped another 20%. Goldman also recently downgraded it to Sell, and Intuit now sits roughly 62% below its 52-week high of $813.70.

So this is a thesis check — a deep dive into the two different companies that exist inside Intuit. It’s also a reexamination of what I wrote in March, and a reverse DCF valuation update.

This is not a trade decision. Not yet, but one will have to be made very soon.

The Best of Times Company

If Intuit were only the company described in the first half of its own press release, the 56% drawdown would be a tremendous gift. Three engines are firing, and all three are the ones I argued AI would strengthen.

Assisted Tax = Freestyle Chess

TurboTax Live revenue grew 36% to roughly $2.8 billion. More than double management’s own long-term 15–20% framing. Customers were up 38% and new customers (excluding one-time offers) were up 29%.

Live is now 53% of total TurboTax revenue, up eleven points year over year, and retention ticked up two points.

Upselling to TurboTax Live has been a margin enhancer and now it is also looking like a strong acquisition channel.

In March I compared this to freestyle chess.

The AI does the calculation, the human owns the strategy and the liability. Goodarzi essentially said the same thing on the call: customers in this segment “buy confidence, not code,” and they spend roughly seven times more on experts than on software alone.

Mid-Market Expansion

Combined online-ecosystem revenue for QuickBooks Online Advanced plus the Intuit Enterprise Suite grew about 38% in the quarter, and IES contracts grew 37% sequentially.

The direct sales force is scaling up 30%, and in August they’ll launch consumption-based pricing for AI agents paired with existing subscriptions. Potentially a new catalyst for FY27.

Recently, a fractional CFO stated that they’ve seen a 200% year-over-year increase in IES adoption, and in some instances reverse migration, where growing clients came back to the Intuit/QuickBook ecosystem from the more expensive ERPs like NetSuite and Sage.

Money Platform

The money platform and consumer cross-sell are scaling too.

Total online payment volume grew 30% (18% excluding bill pay), with bill-pay adoption the standout.

On the consumer side, the cross-monetization is working too.

Customers using both TurboTax and Credit Karma carry roughly 30–35% higher ARPU than single-product users, more than 35% of TurboTax customers now adopt fast-money offerings, and TurboTax filers who start in Credit Karma are guided to grow 54% year over year. Credit Karma itself grew 15%, with the full-year guide raised to +19%.

Increased Capital Returns

Intuit repurchased $1.6 billion of its stock in Q3. That’s more than double from the same quarter a year-ago. Over the last nine-month, buybacks are up over 60% to $3.34 billion. Management also authorized a new $8 billion buyback plan, and raised the dividend 15%.

The Worst of Times Company

Then there is the other company.

Low End Disruption

The most exposed part of TurboTax’s customer base is price-sensitive DIY filers, who typically earn under $50,000 per year.

Earlier, management guided for total paying TurboTax Online units to grow 2%, with ARPU and mix doing most of the work. But total online units, including the pay-nothing base, were revised to −2%. That pay-nothing base fell from 8 million to about 7 million, and TurboTax’s share of e-files declined by a point. The headline +7% TurboTax revenue growth came from price and mix, not units.

The strategic response is to shift from complexity-based SKUs (one price for a W-2, more for a W-2 plus charitable donations) to value-based pricing: let simple filers in cheaply or for free, then monetize them through Credit Karma and the money platform.

The main question is whether this guidance change is due to price (management’s story), AI substitution (part of Goldman’s Sell thesis), or broader macro forces.

On the macro side, total IRS filings are down about 30 bps versus an expected +1%. That equates to about 2 million “missing” filers, mostly manual filers. According to Goodarzi, this is the worst category contraction since the post-COVID season.

Management is adamant it is price and macro, but that is what you would expect them to say. They are not going to come out and admit to AI disruption; it would hammer the stock even more than it already has.

Either way, it means fewer new customer acquisitions that could later be upsold as incomes rise and filings become more complicated. Management can overcome pricing and macro with the right strategy. Low end disruption by AI means these customers are never coming back.

Mailchimp Overpay Confirmed

Back in March I argued the real Mailchimp risk was never AI killing email; it was Intuit overpaying and then aggressively raising prices to justify the $12 billion purchase price, pushing current and future customers away.

Q3 made it official.

Mailchimp revenue is down slightly year over year, and the asset has been openly demoted to cash management.

In Aujla’s words, its cash-flow profile “will generate more value for Intuit than a third party is likely to pay for that asset in the current environment.”

Translation: Intuit can’t sell Mailchimp without taking a big loss so they’ll run it for cash and minimize reinvestment.

17% RIF

Intuit is cutting about 3,000 employees, roughly 17% of the workforce.

Goodarzi was explicit that the cuts are not AI-driven, breaking them into four buckets: fewer management layers, fewer coordination roles, removal of TurboTax/Credit Karma integration duplication, and right-sizing Mailchimp. Maybe that’s it?

But this is the second reduction in two years, and the 2024 framing that Intuit doesn’t “do layoffs to cut costs” didn’t last past 2026.

And Aujla’s clarification that the majority of savings will flow to the bottom line (versus 2024, when savings were reinvested) is management stating that they are now optimizing for cash and EPS rather than reinvestment like they said back in 2024.

Mid Market Ceiling

The same IES that’s winning the low-to-mid market reportedly keeps losing the high-mid. Once a client needs serious multi-entity consolidation and audit readiness, evaluations tip to NetSuite, Sage, or SAP. “Moving upstream” is a big growth lever but upstream growth may have a ceiling.

Reconciling the Two

Does monetizing beyond tax — Credit Karma plus the money platform, a 30–35% ARPU lift — more than offset price compression at the low end?

Can lower share and/or lower ARPU at the low end be more than offset by a higher mix in Assisted Tax?

Intuit currently has only about 10% of the ~$23B assisted tax prep market, and filers in this category generate, on average, 2.6x the ARPU vs DIY.

Intuit could lose at the low end and still grow if it takes assisted share from less technology-forward competitors faster than its low-end base erodes.

Because this is unknowable for several quarters, it increases uncertainty around the name, and market participants do not like uncertainty. They demand a lower multiple to compensate for the wider range of outcomes. Intuit’s current stock price reflects this rerating.

My framework from March still holds up. Who bears the liability when the AI is wrong, what regulatory barriers exist, how real switching costs are, and how proprietary the data is. But the answers have moved unevenly since March.

Credit Karma’s lead-gen model remains the highest structural risk. Even though TurboTax concerns garner all the attention.

TurboTax got worse at the low end and better in Assisted.

Mailchimp is being run for cash and not a top of funnel priority now.

And QuickBooks’ core is intact, and making great inroads up market.

Valuation

I updated the reverse DCF and my Bear, Base, and Bull cases around its new numbers.

Read more

TCI cut Microsoft from 10% of the portfolio to 1% in the first quarter, and it’s probably at 0% by now.

Egerton Capital, a woefully underfollowed fund, fully exited its position too.

Meanwhile, Bill Ackman initiated a brand-new ~$2.1 billion MSFT position.

Three concentrated, high-conviction stock pickers came to opposite conclusions in the same quarter.

TCI and Egerton see Microsoft losing the AI race. Ackman sees a fortress that’s mispriced.

The Bear Case: “AI Introduces Uncertainty Over Microsoft’s Competitive Position”

Chris Hohn was explicit in TCI’s Q1 2026 investor letter:

“We reduced our investment in Microsoft because the rapid progress in AI introduces uncertainty over Microsoft’s competitive position in the future.”

He singled out two specific concerns: the Office productivity franchise, and what he called “some risks in Azure.”

TCI then redeployed the capital from MSFT into Alphabet. Raising it to 5% of the portfolio and making it their largest tech position.

TCI rode Microsoft through a near-400% return over nine years. So when a high-conviction, long-term holder sells the position completely, you have to stop and ask: what do they see? Especially when MSFT is part of your own portfolio.

Productivity

The bear case on Office isn’t that ChatGPT, Claude, or other LLMs replace Word or Excel. It’s that AI-native productivity tools chip away at the Office bundle one workflow at a time. The bundle has been Microsoft’s moat for 25+ years. If AI-native tools fragment the workflow, its not one death blow but a death by a thousand cuts.

Recent data points have been unpleasant.

Microsoft 365 Copilot’s Net Promoter Score landed at -19.8 in early 2026 and it’s market share among paid AI subscribers collapsed from 18.8% to 11.5% in six months. And after more than 18 months on the market, Copilot sits at roughly 20 million paid seats. That’s about 4% penetration of the M365 installed base. Seat additions are accelerating. The most recent quarter was the fastest net-add quarter since launch, but the absolute share losses is showing up in the data.

“Just bolt AI onto the legacy suite” isn’t a winning product strategy. And I think that’s what Hohn believes MSFT’s strategy is right now.

“Some Risks” in Azure

Hohn didn’t elaborate on Azure, but it might be the growth rate and the quality of its CRPO backlog.

Google Cloud grew 63% year-over-year in Q1 and accelerated 16 points sequentially. AWS grew roughly 21% and accelerated 4 points. Azure grew 40% in constant currency and accelerated 1 point. All three hyperscalers are running faster, but Azure is running the slowest right now.

For roughly a year, Azure has been range-bound between 35–39% constant-currency growth—which, don’t get me wrong, is still great. But the narrative was always “Azure is the AI cloud.” If that’s the case, then its growth rate should’ve shown up stronger this quarter when all the other hyperscalers’ growth rates picked up. But it didn’t.

Microsoft’s guidance for the next quarter calls for modest acceleration, supported by capacity coming online roughly six weeks earlier than expected at the Fairwater, Wisconsin facility. So they are still capacity constrained, not demand constrained. But this is the third quarter where MSFT has said “we couldn’t keep up with demand,” and at some point investors stop accepting it.

The quality of MSFT’s backlog could also be a concern.

Microsoft’s Commercial Remaining Performance Obligation is $627 billion, up 99% year-over-year and bigger than Google Cloud’s $462 billion backlog. But size isn’t the same as quality. Roughly 45% of Microsoft’s CRPO is concentrated in OpenAI commitments. Google’s backlog is more broadly distributed across enterprise customers.

The workload mix is probably Microsoft’s most defensible point.

Google Cloud is winning AI training workloads thanks to TPUs and the native Gemini stack. Azure is winning enterprise AI inference layered onto existing M365 and identity infrastructure. This should be stickier with a higher-margin profile.

The growth gap is an interesting short-term data point, but Azure is also growing 40% off a base roughly 2.5x larger than Google Cloud’s. This is a risk to monitor, but not a thesis-breaker, yet. If Azure dips below 35% while GCP stays above 50% for two consecutive quarters, it becomes worrisome.

The Bull Case: The Fortress Is Mispriced

Ackman built his position in MSFT during the Q2 selloff. This continues a recent pattern: Alphabet during its “AI loser” panic, Amazon after Liberation Day, Meta during its capex panic, and now Microsoft on AI skepticism.

MSFT’s valuation is at a multi-year low.

At roughly 21x forward earnings, Microsoft trades in line with the S&P 500. This is below its own three-year average and at the bottom of its NTM EV/EBITDA range.

Buying the most dominant enterprise software franchise in the world at a market multiple has historically been a profitable trade.

The core of Ackman’s thesis is that M365 is a fortress.

Microsoft 365 isn’t a single application; it’s a bundle tied together with Microsoft’s security, compliance, and governance infrastructure. He describes it as “tightly integrated into the daily workflow of nearly every large enterprise.”

An AI startup can replicate Word. It cannot replicate years of embedded enterprise identity, authentication, and data-governance infrastructure. M365 and Office are the industry standard, the common language across businesses. Take Excel: it’s the common denominator across all of finance. LLMs like Claude aren’t trying to replace it; they’re trying to become a valuable tool within it.

The Copilot share-loss data is real, but it has mostly come from individual consumers, not enterprise, which really drives the economics. Claude and Excel work great together, and it’s what Copilot should’ve been from the start. It’s also why MSFT formally designated Anthropic as a sanctioned sub-processor for MSFT’s Online Services in early 2026.

Ackman also believes the market is mispricing MSFT’s OpenAI stake.

MSFT’s ~27% economic interest in OpenAI is worth roughly $200 billion at OpenAI’s most recent funding-round valuation. That’s roughly 7% of Microsoft’s current market cap. The total Microsoft investment across several rounds was roughly $13 billion. The stake is now worth approximately $228 billion. That’s a 17.6x return.

Ackman’s counter to recent worries about capex and cash flow is that MSFT’s current capex is in a classic J-curve.

Hyperscaler capex keeps ratcheting upward and MSFT’s FY26 capex is now expected to be $190 billion. But that’s $190B against $627B of contracted demand and an AI business running at a $37B annual run rate, up 123% year over year. Though frighteningly high, the capex looks like growth investment against visible demand, not speculative overbuild. The thesis breaks if demand declines.

Where I Land

Read more

How Big is Space?

SpaceX filed for an IPO at a $1.75T target valuation on $18.6B of sales with a claimed $28.5T total addressable market that approaches total U.S. GDP.

Griffin’s Doors

Ken Griffin built one of the most profitable hedge funds ever. When AI first came out his firm had little use for it killing 195 of 200 projects. But now that has changed and Ken is slightly worried.

The Coming Mega-IPO Flow & Fudging Problem of 2026

2026 could bring a historic IPO wave (SpaceX, OpenAI, Anthropic). Trillions in new stock must be financed by selling what investors already own, creating pressure across markets.

Ben Carlson: Exploring Risk and Reward

A good interview with Ben Carlson on his new book about the history of risk and reward in markets and how the lessons of the past can help make you a better long-term investor.

• & the book Risk & Reward: How to Handle Market Volatility and Build Long-term Wealth

How Private Equity Ate Britain

Subscribe now

Read more

Meta is Dying. It’s About Time.

Meta may be entering its “zombie era”: user growth is slipping, spending is debt-fueled, and near-term monetization moves (more ads, higher prices) can accelerate long-term decline.

Agentic AI and the Future of Enterprise Software

Public markets are pricing AI as a threat to software. Vista’s 2026 Outlook argues the opposite that a switch from seat based pricing to outcome based pricing will unlock value and cause enterprise SaAss multiple to rerate higher.

The Van Halen Test

Van Halen’s “no brown M&Ms” clause was a compliance check. Failure signaled bigger risks with complex, safety-critical stage setups. How you can build small process checks to catch sloppiness by you or your team.

How to Find 100 Bagger Stocks

Matt Ankrum breaks down everything he has learned from studying 100 baggers and how to indentify potential ones and how he is building a “coffee can portfolio” with them.

• And the book if you’re looking to pick it up The Coffee Can Investor: A Stock-Picker’s journey to Build Generational Wealth

Why Youth Sports Are Out Of Control

Subscribe now

Back in January Anthropic shipped Claude Cowork.

Within hours, software stocks started bleeding, and by the end of the month, $170 billion in market cap had evaporated.

It was The SaaS-pocalypse.

If an agentic AI can write code, draft emails, summarize documents, and handle structured workflows at near-zero marginal cost, then the application layer is at risk. Anything that was or looked like a “SaaS” company was vulnerable until proven otherwise.

Paychex got dragged in with them.

Down around 40% from its most recent 52-week high.

But Paychex isn’t a SaaS company.

It’s a payroll processor with a PEO business stapled on that has a regulatory moat built in.

The AI-replacement story that wrecked SaaS companies doesn’t map cleanly onto a business whose customers pay it to not get audited.

But there is an AI adjacent element to the bear case and AI disruption is the loudest story in the market and the one getting the most attention even for companies like Paychex.

Subscribe now

The Four Bear Narratives

The bear case against Paychex is composed of four narratives.

AI and Advisory Pricing

Paychex sells payroll at the core, but the margin lives in the HR consulting, compliance guidance, and PEO advisory wrapped around it. That’s where the take rate sits.

The concern is that agentic AI eats away at the HR, compliance, and PEO advisory business.

Routine HR questions like “what’s the I-9 deadline?”, “how do I classify this 1099?”, “what’s the FLSA threshold this year?” are exactly the kind of structured, rule-based queries large language models handle. You don’t need 80% of Paychex’s advisory function to get replaced. 30% of its highest-margin layer being competed away is enough to change the future value of the business.

Even though AI disruption is the most salient risk right now across the market. I don’t see it as an immediate risk to Paychex’s advisory business because of behavioral inertia and major legal barriers.

If an SMB is using Paychex and it solves all their payroll and/or compliance needs, then why switch? Yes, there are cheaper payroll companies, which I’ll talk about below, but then you’ll also need to replace your HR and compliance. It’s a hassle, and unless a small business is really at their breaking point with Paychex, they’re unlikely to get distracted by a new project that doesn’t grow their business.

For the AI disruption narrative, the biggest hurdle is liability.

Paychex’s HR and PEO business is not just selling a service; it’s selling indemnification.

Someone has to sign the 941.

Someone has to be the named defendant when an employee files an EEOC complaint.

Someone has to hold the workers’ comp master policy and carry the ERISA fiduciary duty on the 401(k).

The IRS is not going to accept “the AI told me to” as a defense. AI cannot be sued, cannot be deposed, cannot be an insurance counterparty, cannot owe fiduciary obligations to a plan participant. These are structural properties of being a legal person with assets.

With PEO or co-employment, Paychex becomes the employer of record on the W-2. The IRS certified them as one of fewer than 150 CPEOs in the country, which requires quarterly bonding, audited financials, and confers sole liability for federal employment taxes on the PEO itself.

Sure, an LLM can answer basic questions and handle basic consulting, but it can’t sign the return, post bond, hold the policy, and absorb a lawsuit.

Increased Payroll Competition

Payroll, Paychex’s original business, has been commoditized, and you can argue it is losing ground to newer software-native companies like Rippling, Gusto, and Deel. That is one reason Paychex bought Paycor: to strengthen its position at the upper end of the market. But at the low end, the new-business-formation cohort is increasingly turning to modern, tech-native payroll platforms with better UIs.

The AI disruption narrative is also at play here, with the assumption that SMBs will vibe-code their own system.

Again, this skips over the legal and liability hurdles.

Every state or local jurisdiction has its own filing frequency, deposit schedule, withholding tables, unemployment insurance rate, and rules for sick leave accruals, tip credits, and supplemental wage rates. New York City alone has roughly 40 distinct payroll-related compliance touchpoints. The IRS assesses about $13 billion annually in payroll tax penalties. Roughly 40% of small businesses get hit with at least one payroll-related penalty per year if they self-file.

Paychex’s major selling point is that it will eat the penalty if it messes up. Again, it’s selling SMBs indemnification. A vibe-coded solution opens up the SMB to increased liability.

But a cheaper software native solution that provides the same payroll efficiency and liability protection is a bigger immediate concern.

Management Solutions ARPU growth, the closest available proxy for advisory pricing per client, decelerated from +6.9% in FY23 to +2.3% in FY24, before the Paycor deal closed.

Paychex does not disclose ARPU directly, but it does disclose total revenue and approximate client count. The three proxies break PAYX’s estimated ARPU into three parts: total revenue per client (broadest, includes float), service revenue per client (ex-float), and Management Solutions per client (the closest thing to an advisory ARPU). The acquisition lines flag Lessor, Oasis, and Paycor, where the denominator gets rebased.

Paychex is not dying.

The PEO business is growing.

Payroll tax compliance across 50 states is not getting easier.

But the ARPU deceleration suggests its pricing power is fading.

PAYX’s acquisition strategy is about buying growth and moving upmarket because the organic flywheel is slowing at the lower end. It’s a defensive move to combat the encroachment of software-native competitors.

The Small Business Jobs Index Is Below 100

Paychex publishes its own Small Business Jobs Index. The SBJI baseline is 100.

It printed 99.30 in January 2026 and 98.81 in March. Below 100 means SMB employment is contracting on a relative basis.

Paychex revenue is partly priced per check. Fewer checks per client means less revenue. Yes, AI could increase layoffs at SMBs but the more immediate threat is a general economic contraction.

The SBJI is Paychex’s own data and based on their customers, but this is Paychex publicly warning that small businesses are getting hit by the current administration’s economic policies: tariffs, inflation, and high gas prices crimping the US consumer.

Things could change in the back half of the year, if tariff uncertainty dissipates, oil starts flowing through the Strait of Hormuz again, and inflation subsides enough to warrant further rate cuts.

Interest Rates

Paychex holds client funds between collection from the employer and disbursement to employees. Around $4.5 billion. Paychex gets to earn interest off this float at near-zero cost. It flows right to PAYX’s bottom line.

In FY25, interest was $167 million. In FY26, on management’s guidance, it’s tracking toward $205 million. Roughly 3% of total revenue and more than 3% of operating income.

Float income is rate-sensitive. Every 100 basis points of Fed cuts compresses the yield Paychex earns on those funds by approximately the same amount. Rates fall, float income declines, and earnings decline.

The operating-business EPS and the reported-EPS line have moved together for four years because rates have been a tailwind. But that has changed recently, and if the new Fed chief Warsh can get the board to lower rates further, then that is another earnings drag.

This is cyclical, and Paychex has been through many different rate environments, but a rate-cutting cycle occurring when market participants are pricing in business disruption increases the pressure on PAYX’s stock price.

What’s Priced In?

When I run a basic reverse DCF on Paychex at its current price, with a 9% discount rate and a 2.5% terminal growth rate, the market is pricing in roughly a 3.7% revenue CAGR with flat margins. Or if I keep revenue growth flat, it’s pricing in 650 basis points of margin compression.

Either way, the multiple has rerated to a slow-growing cyclical with float headwind and modest competitive erosion business.

The bears arguing AI is killing Paychex are over-arguing the case. The market doesn’t need that to be true. It only needs small business contraction and falling interest rates to play out, which looks to be the case, to justify the rerating of Paychex.

The bull counter is in the next 6-12 months SBJI inflects above 100 and the Fed pauses cuts, this could cause the multiple to snap back because the regulatory moat is durable. The PEO cross-sell flywheel kicks back in post Paycor acquisition. And now that 3.7% implied growth is too low.

The Real Threat

The SaaSpocalypse and AI disruption narrative is the loudest narrative, and it’s affecting Paychex’s stock price, but it’s not the immediate threat to Paychex’s core business.

It’s increased payroll processing competition on the low end, economic pressure on Paychex’s core small business customers, and declining float income in a rate-cutting cycle.

Subscribe now

How Fast Can Gulf Production Recover After Reopening?

The base case is a few months but the longer the wells stay put in the harder it gets to bring them back online. The worst case right now is a multi quarter process.

FICO Part 3: Update & Framework

The reality behind the wild market moving news surrounding FICO, the GSEs, and the FHFA.

Wealthfront Corporation (WLTH)

A look at the capital light software driven investment management platform. With 10-20% growth, >40% EBITDA margins, and 1/3rd of its market cap in cash and no debt.

SpaceX IPO: Why This Hedge Fund Manager is Fading the Hype | Moez Kassam

Although he’ll not short SpaceX directly, there are plenty of me too space companies to look at and who’s flows may reverse and go to SpaceX when it comes to m market.

Why Saudi Arabia’s $5B LIV Golf Experiment Failed | WSJ What Went Wrong

Subscribe now

In 1887, the German mathematician Carl Gustav Jacob Jacobi gave a piece of advice that Charlie Munger would repeat throughout his life.

Man muss immer umkehren.

Invert. Always invert.

Jacobi was talking about hard math problems that were resistant to the usual direct assault. But when you flip them around and ask a different question, the problem becomes much easier to solve.

Munger took the same idea and applied it to investing.

Don’t ask how to get rich. Ask how to stay poor, and then avoid that.

Don’t ask which stocks will win. Ask which stocks tend to lose, and start by not owning those.

It’s a powerful reframing.

What to Avoid

In 2008 Cooper, Gulen, and Schill published their paper looking into the asset growth anomaly. What they found was firms that grow their total assets aggressively go on to underperform the market by roughly 8 percentage points a year on a value-weight basis and this is sustained over five years.

The anomaly builds on two other anomalies. Pontiff and Woodgate found that net stock issuance predicts for returns. Then Titman, Wei, and Xie showed the same for abnormal increases in capital expenditures (Capex).

Companies that grow their asset base aggressively are usually doing one of three things: issuing equity at a rich multiple to fund growth they can’t fund internally, acquiring assets at poor prices, or increasing Capex into a trend they think is secular but is, in fact, cyclical (or less robust than they thought).

These companies disappoint because overzealous acquisitions get written down, the incremental returns on increased Capex trend toward the cost of capital (producing little excess value for shareholders), or the high-flying stock price used to fund growth re-rates lower, stunting future growth.

Taking Munger’s advice and inverting the investment puzzle, the companies and stocks we want to avoid are the aggressive overinvestors.

The Screen

I started with every name in the S&P 500 and 400 (midcap) and then filtered for companies with a market cap greater than $5 billion at the rebalance date.

I then applied three filters. First, organic asset growth, total assets ex-acquisitions, in the top quintile. Second, external financing as a percentage of prior-period assets in the top quintile. Third, EV/Sales in the top quintile.

The screen produces ~30 names per year. Each name had to be flagged using only data that would have been known at that moment. The basket was equal-weighted, rebalanced annually at year-end, and held for 12 months.

The backtest included names that subsequently delisted, held at last observed price, zeroed out if they went away entirely.

The look back window was the last nine years, from year-end 2016 through year-end 2025.

What actually happened

Over the last nine years, the Aggressive Overinvestors basket grew 2.66x vs 3.50x for SPY.

Annualized, that’s 11.5% for the basket against 14.9% for the market. The basket lagged SPY by about 3.4 percentage points a year.

And this is even during an extremely favorable period for aggressive corporate investing with AI infrastructure, GLP-1 buildouts, defense rearmament, and energy expansion.

And there were some strong years for aggressive overinvestors during this period.

The basket put up two enormous years: +65% in 2020, which encompassed the stay-at-home boom with Roku, Wayfair, Carvana, and Chewy leading the way. Then again at the start of the AI infrastructure boom in 2023-2024.

The basket’s worst year was 2022, when it lost 33% against SPY’s 18% drawdown.

The 2022 basket included Carvana, AppLovin, Wayfair, Roku, Block, Toast, Warby Parker, Qualtrics, Tandem Diabetes, Edgio, and Vicor — twelve names that lost between 48% and 98%.

The Disasters

Carvana, down 98%. AppLovin, down 89%. Wayfair, down 83%. Roku, down 82%. Block, Toast, Warby Parker, Qualtrics — all down between 48% and 80% in a single year. Almost all of them in 2022.

These were among some of the most-discussed, most-recommended, most-owned growth stocks of the prior cycle. They appeared in dozens of newsletter portfolios. They were core holdings of high-flying funds and the narrative for each of them was well articulated

But every single one of them tripped the same set of mechanical filters in the year before they collapsed. Aggressive asset growth, funded by external capital, at a rich revenue multiple.

The screen is not without its drawbacks.

The same screen also flagged Eli Lilly, Broadcom, Palantir, and several other names that performed well through the period. The screen is designed to identify a cohort that underperforms on average. It does not specifically tell you which individual name will collapse.

The goal is to prevent you from owning these dramatic blowups. From permanently impairing your capital.

The goal is to incrementally improve the long-term returns of your portfolio.

But if you’re using this screen for potential shorts, you can’t rely solely on it. It’s a starting point. You will need to dig deeper and create a wider mosaic to try to identify the potential blowups.

Behavioral Checkpoint

I’m not saying this is some binary choice between trying to pick winners or avoiding losers. It’s not a competition because both approaches complement each other.

They can protect you against different mistakes.

Let’s say you think you found your quality compounder. Returns on capital are high, the moat seems real, the management team has a credible plan.

But before you initiate a position you see if it makes the screen for aggressive overinvestors. Is this name in the top quintile of organic asset growth? Is it being funded primarily by external capital? Is it trading at the kind of revenue multiple where the math only works if everything goes right?

If the answer is yes to any of them, you don’t necessarily need to kill the idea. As I said above, it flagged Eli Lilly, Broadcom, and Palantir too. But you need to ask some harder questions and understand why the company is making this screen. Maybe you size the position smaller or have a tighter stop if your thesis is not playing out.

Munger’s advice to invert is essentially a behavioral checkpoint. It asks you to slow down and look at your next great investment a different way because maybe in your excitement about how great this investment is, you overlooked a crucial detail.

Creating a checkpoint slows our thinking down and causes us to engage our more rational thought processes.

Key Lessons

I see three key lessons from Munger’s inversion advice.

The first is that inversion is useful and widely underused. Most investors run elaborate processes to find what to buy with no formal process to identify what to avoid.

The second is that the screen’s real value isn’t as a standalone strategy. Its value resides in being a behavioral checkpoint.

The third is about portfolio management. The point is not to be right about every stock you avoid. It’s about being wrong less often about the stocks you own.

A quality book is a concentrated bet on a small number of companies. The cost of accidentally including a blow-up is enormous. We want to reduce those chances for the long-term health of our portfolio.

What’s in the Basket Today?

The current Aggressive Overinvestors basket is concentrated in three themes: AI infrastructure, GLP-1 manufacturing scale-up, and the post-IRA (Inflation Reduction Act) industrial build-out.

This doesn’t mean all the names in the basket will be losers. Some of them could turn out to be great compounders — especially if their asset growth is in highly productive Capex that generates returns above their cost of capital. But a few of them just might make the list of worst performers.

The full list of names is below for paid subscribers.

Read more

Making the Marlboro Man

The Marlboro Man is one of the most successful rebranding in history. It changed a niche cigarette for women into a global behemoth.

Half of Google’s and Amazon’s Blowout ‘AI profits’ Came From Anthropic—Not Their Core Business

If you own Google and Amazon, then you own a piece of Anthropic too.

Google and Amazon’s “AI profits” weren’t mostly from selling more ads or cloud—they were largely paper gains from marking up their Anthropic stakes after new funding rounds.

GLP-1s, Peptides, and The Trillion-Dollar Health Revolution

The GLP‑1 race isn’t about the most dramatic weight‑loss headline. It’s about the drug people will actually stay on. Lower friction, better tolerability, and a cheaper, easier format can unlock far more adoption than a “stronger” molecule.

AI Can Make Software Now. That Changes Everything, with Paul Ford

Despite the title it’s a much needed middle of the road discussion on AI, vibe coding, and the risks to software jobs.

Why Did Italian Football Fall?

Subscribe now

In 1992, most small businesses in America were still doing their books on paper.

Some had graduated to spreadsheets.

An enterprising few repurposed Quicken, a personal finance tracking software, as a business accounting tool.

They were already using Quicken to better understand their personal cash flow and balance sheet and immediately recognized the benefit of using an accounting program in their businesses. The problem was that none existed, and Quicken became their workaround—even though it wasn’t the best fit for business accounting.

Intuit recognized the opportunity and launched QuickBooks to immediate success. It soon became the platform for small business accounting. Accountants standardized their practice around it, and their clients—small businesses—followed them onto the platform.

Intuit continued to build upon its platform, entrenching its users further. They added payroll, payments, tax solutions, and third-party integrations. Each new addition made the ecosystem more valuable to its users and harder to leave.

Intuit digitized small business accounting and became the industry standard.

Thirty-four years later, I see a similar pattern and opportunity forming in a $13 trillion industry.

Construction is 13% of global GDP and McKinsey ranks it the second-least-digitized industry, behind only agriculture.

Project schedules live in spreadsheets. Requests for information (RFIs) sit unorganized in email inboxes. Budgets get reconciled manually across disconnected point solutions, sometimes weeks after decisions have already been made in the field.

The industry isn’t under-digitized because it doesn’t need software. It’s under-digitized because no one has solved the operating system yet, like Intuit did with Quickbooks.

Until this month’s selection.

The company has spent the better part of two decades making the same bet Intuit made in 1992.

And like Intuit, the core product is just the base to build incremental layers of operational value for its customers, increasing the value of the platform and increasing the costs to switch, while also making it hard for owners, contractors, and general contractors to operate their businesses without it.

The company is still in the early stages of building its competitive moat and becoming the industry standard. This creates a large opportunity, but also increased risks and a wider variance of potential outcomes.

Read more