The Anthropic Prospectus Arrives: My (but not the) IPO valuation!

    I am a dabbler, by nature, moving on from topic to topic, but this year, I have spent more time than usual on AI, but that is because it seems to be part of almost every business or economy related story. I tried to value SpaceX ahead of and after the filing of the prospectus for its initial public offering before moving on to the question of whether it and OpenAI and Anthropic should be part of the S&P 500 index. I started my analysis of how AI would affect the rest of us, from doomsday to do-nothing scenarios, largely in response to Citrini and Citadel reports on the topic. in August 2026, I took my first big picture look at AI as a business, which I titled its bar mitzvah moment, with a look at its potential market, business economics and moats. That look was a prelude to the initial public offerings in the works (then) for Anthropic and OpenAI, and I provided a simple way to reverse engineer how much investors would need in revenues from these companies to justify their sky-high pricing.  In early October, I looked at how the massive investments in AI are playing out in the aggregate earnings, cashflows and invested capital at US companies, and why it is contributing to holding the market intact, in the face of rising rates. Just last week, I argued that both Anthropic and OpenAI would be priced, rather than valued, for their initial public offerings, using what I called buzzy metrics (like annualized run rates) and promised that I would try to value the company, when it filed its prospectus. On October 26, Anthropic filed its prospectus, and at the risk of being called hopelessly wrong by both bulls and bears, I will try to value the company, using what I see as its most likely pathway, while also leaving the door open for you to disagree with me, and pick your own pathways to valuing the company.

The Prospectus

    Before I delve into the details of the S-1 filed by Anthropic, it is worth noting that at 416 pages, it is long and filled with distractions, as has become the case with almost every initial public offering in the last two decades. (I chronicled this trend in my post on the disclosure dilemma, arguing that while filings have become longer, they have become less informative). The Anthropic IPO has useful information that both bulls and bears will spin to justify their priors, and in the section below, I look at key disclosures, broken down by valuation input into revenue growth, profitability, risk and reinvestment categories, from both an optimist (glass half full) and pessimist (glass half empty) perspective: 

1. Revenue Growth and Composition: The good news in the revenue story is that the company has grown exponentially over the last year, with revenues in the first two quarters of this year about fourteen times as high as the revenues in the first two quarters of the last year, but the sobering news is that revenue growth slowed down in the third quarter of 2026, as OpenAI increased market share, compute capacity acted as a constraint and some enterprises moved to cheaper open source models. The good news on customer composition is that the number of enterprises that spend more than $1 million a year on Claude increased to 1900, from less than a hundred a year ago, but the bad news is that the company still has concentrated revenues, with its top five customers accounting for 25% of revenues. 
   
   
   
2. Operating and Net Profitability: In bad news, Anthropic reported an operating loss, corrected for non-cash and one-time charges, of over $ 8 billion, for  the first six months of the year. The actual loss reported was $80 billion, but $72 billion represented a non-cash charge, expensing the conversion option of the convertible bond. The good news, though, is that not only are the trend lines working in their favor, with the company claiming that it had an operating profit in the most recent quarter. The company is also delivering its revenues more efficiently over time, with all of its expenses decreasing as a percent of revenues across the quarters, with compute costs, its biggest cost,, decreasing from more than 250% of revenues to 48% of revenues, on a quarterly basis.
   
   
   
3. Reinvestment for growth: AI agents are computing power hogs, and AI companies invest immense amounts in compute power. The financial statements in Anthropic's prospectus actually do not show this investment at the moment, with neither the cap ex in the statement of cash flow nor the debt (and invested capital) in the balance sheet showing substantial investment in compute power. Before you accuse Anthropic of hiding the truth (or worse), the company is open about its future compute power commitments, reporting hundreds of billions in compute commitments in future years. The good news is that Anthropic is now catching up with OpenAI, which had the lead on compute power, but the bad news is that this catching up has not been cheap, in terms of cost commitments in future years. It i s also worth noting that compute power will need to be renewed and reinvested in, and while chip costs (a key component) may come down, data centers, facing opposition from communities, may become more expensive. 
   
   
   
4. Operating and Regulatory risk: Anthropic's argument is that its customers, once they adopt Claude models within their enterprises, will find that they become so integral to operations that they will not abandon them (making revenues sticky and the business less risky). That has not been fully tested yet, and it is entirely possible that spending on AI agents is cyclical and that companies will pull back in bad times. On the financing front, committing to hundreds of billons in compute, purchase and lease commitments exposes the firm to default risk (with a bond rating of BB+), but in response the company will argue that its access to new capital, as a publicly traded large market-cap company, will make this risk recede. In addition, its cash balance, already at $114 billion, and likely to be bulked up with another $100 billion from the IPO offering, will give them more than enough cash to cover commitments in the next few years. 

   

   Source for AI risks: https://airisk.mit.edu 

If I were to summarize the information that I obtained in the prospectus that is relevant to valuing the company, I would come up with the following (and neither Anthropic bulls nor bears will be happy with me):

1. The company's revenues are growing quickly, and for at least of the subset of users that are its core customers, it is delivering on its promise. That said, the revenue growth rate is clearly coming down, as scale, competition and compute power constraints all work against the company, and the fact that Anthropic is so interconnected in both operating and financing practices with other big players in the AI game (the circularity that is often noted by critics) does lead to the potential for double counting and opacity.
2. The expenses for Anthropic, across categories, are growing slower than revenues and that is helping reduce operating losses and putting the company on a glide path to not just profitability, but significant profitability. That said, notwithstanding one quarter's reported profits, the company is not quite there yet, and will require significant revenue growth to keep ahead of the expenses (from the compute power commitments) that will hit the income statement in future periods.
3. The immense compute power commitments at Anthropic have not only allowed the company to lock in what it needs to deliver growth but are also largely responsible for the high gross profit margins reported by the company, because they have moved company's biggest expense from the variable to the fixed operating cost column. That has come at a cost, since this not only increases the company's operating leverage, but also will result in further commitments down the road, as compute power needs to be replenished and grown.
4. Anthropic has built a buffer against risk in the form of a very large cash balance, which will be augmented after the IPO, and it is is possible that Anthropic's customers will become reliant on its products, making its revenue stickier.  However, it is exposed to risk on three dimensions - in operations, from high operating leverage, in financing, because of its contractual content cost commitments and on survival, since it is dependent on growth to meet its obligations.

Over the next few days, don't be surprised to see people make cases for and against Anthropic using their spin of the data in the prospectus. Neither side will be convinced, or convincing, and will emerge with their priors about Anthropic intact and reinforced. I think that too much time and effort is being spent on going through this prospectus, as if it contains the answers to every question you may have on AI, when, in truth, it is a snapshot of a young company that is evolving. This prospectus, at least in my view, is analogous to reviewing a report card for a kindergartner (a precocious and talented one perhaps, but still a kindergartner) and trying to determine how much, if any, to set into that child's college fund. 

    My one piece of advice, no matter which side of the optimist/pessimist divide you may fall on, is that you do not get sucked into conspiracy stories revolving around AI company cabals (where the hyper-scalers, chip maker and LLMs are all working at creating a mirage of AI success to fool investors) or accounting fraud. I agree that self interest is driving many of the actions that you observe in this space and that accountants sometimes use the discretion in accounting rules to make the numbers look more favorable for companies than they really are, but those are common problems across all companies, not just AI and Anthropic, and falling into the conspiracy rabbit holes effectively blocks you from even considering alternative explanations.

The Anthropic Valuation - Story and Numbers

    The prospectus, with all of its disclosures, is still a document rooted in the past, and the value of a business lies in your expectations for the future. In this section, I am planning to leave the safety of the historical disclosures and embark on forecasting the future, knowing fully well that I am going to be wrong, and perhaps even massively wrong, in my attempts. At each step of the story, I will offer you a chance to disagree with my take, and provide your own, with the resulting change in inputs. I know that you are busy, but if you can humor me and play alone, we may both benefit from this exercise.

1. The Scaling Up Story

    The value of Anthropic, and any of the other AI product and service companies that will come later, ultimately rests on how big you believe the AI products and services market will become, and what portion of that market you company will inhabit. I started on this discussion a couple of months ago in my assessment of AI as a business, and argued that the total market for AI products and services will depend on four assessments - 

1. whether AI is a productivity tool for employees or employee replacement, with the latter creating a larger market for its products and services.
2. whether that employee displacement will be primarily at the top end (top quintile) of the wage scale or across the board, with the latter pushing up the total market.
3. whether AI's impact will be concentrated in a few industry groups or across many or most industries, with the market expanding as AI's reach expands
4. whether the AI disruption will be primarily in the United States and a few other geographies or global, with a global reach leading to a bigger market.

I used the picture below to illustrate the range of total market outcomes:

I believe that AI will have the power to disrupt some industries, and that its most powerful (and highest priced) models will displace white collar (and highly paid) employees, yielding an end market size of $5 trillion or greater for its products. In my view, and that is based on a very early reading of the AI business today, Anthropic is aiming for the premium portion of the market and looks like a leader in that portion of the market. The one emerging add on is Anthropic's ambition (made explicit in the prospectus) to be a competitor in a subset of the businesses that its agents are disrupting, with the pharmaceutical business emerging as a potential play. In effect, Anthropic Pharma, or whatever they choose to call it, will then use its agents to do drug research and testing, displacing the companies that currently provide this service, and reap at least a portion of the revenues from the drugs they develop. With that story in place, I believe that it is plausible that Anthropic's revenues could approach a trillion dollars in a decade. For perspective, there is only no company in the world with more than a trillion in revenues today, with Amazon and Walmart sitting at around $700-$750 billion (albeit in current dollars and time).

    Clearly, these are my views, and yours may be very different. If you are willing to play along, I would like you to make your own assessment and make your choice below:

My choice is shaded black, and as you make your choice, pay more attention to the target revenues that you believe Anthropic will have in 2036 than to the compounded annual growth rate between years 2 and 5, which is just an input to get to your target revenues.

2. The Profitability Story

    There is clearly movement towards profitability in the prospectus, but where will the profit margins end up after Anthropic has gone through its growth pains? To estimate where margins will converge, it may help to start with the recognition that the AI product and service market, once established, is likely to have a mass market segment where AI agents and models, many of which may be open-sourced and less powerful, will compete for customers primarily on pricing, and a premium segment, where frontier models, many of which will remain closed models, powered by client data, dominate. The operating margins for the former are more likely to resemble those of business services (roughly 10%), whereas the operating margins for the latter will converge on the operating margins of enterprise software (close to 40% for the top quartile), with compute power and R&D becoming the reinvestment needed to deliver these products. 

    Anthropic is targeting the premium market, as reflected not only in its current revenue breakdown, with almost 80% from enterprises paying based on usage, but also in its focus on using its newly acquired compute power to build next generation models rather than service current clients. I will assume an end-mix of 75% of revenues from premium products, and 25% from the mass market, giving me an unadjusted pre-tax operating margin of 32.5%; that would put them at the ninth decile of US companies in 2026. 

    Here again, you may find room for disagreement, either on the market mix (mass market versus premium), the margins for each market segment or business economics, as companies scale up. You can make your choice below:

Again, my estimate of the target margin is highlighted in black, but with plenty of room for a different choice.

3. The Reinvestment Story

    The reinvestment assumptions you make for Anthropic will revolve around understanding compute power, in terms of both how it is created as well as how much frontier models will need that power, as they become more powerful and scale up. For the moment, Anthropic, with its compute commitments in excess of $600 billion and greater, has sent signals that it views compute power as a necessity to power its growth ambitions as well as to maintain an edge over competitors (current and prospective).

    The nature of compute power, and especially with the chips that represent a large percentage of costs, is that investments will age quickly and will need replacement. This is an indication that if Anthropic's growth ambitions come to fruition, there will be more reinvestment needed in future years, to both replenish and grow compute power. It is true that technology may make chips cheaper, but that may be more than offset by new models demanding proportionately more power, and if you add in the real estate and power needs of data centers, there is room for diverge on how much additional investment will be needed in future years. I do believe that the compute power crunch will stay in place for the near term, but will ease over time, and that belief translates into my assumption that Anthropic will need to invest about $1 for every $3 in additional revenues, going forward. That would be in line with what technology companies are generating as revenues per dollar of invested capital today, but with the leap in revenues that Anthropic is forecasting, it will translate into large dollar reinvestments each year.

    As with revenue growth and profitability, I will leave the door open for disagreement, and you can make your choice below:

While I have highlighted my base case assumption in black, note that lower sales to capital ratios will make the business more capital-intensive, and increase the cash drain from reinvestment.

4. The Risk Story

    As you start listing out the risks that you will be exposed to as an Anthropic investor, you may be overwhelmed both by the number of risks as well as their diversity. To gain some control in this process, I will divide the risks into three groups:

a. Operating risk: The AI business is still evolving, and while Anthropic's growth and customer stickiness so far are noteworthy, the business will face challenges as it grows, and lives through market and economic turmoil. Thus, for Anthropic's customers spending on AI as a productivity tool, we do not know whether they will cut back their spending, and if so how much, in bad times, and for those who plan to replace employees with AI agents, whether the pace and extent of usage will change with the economy. The number that is designed to convey this risk into the valuation is the cost of capital, and for Anthropic, assuming that it carries the risk of large technology companies with an added burden of the compute commitments, I will estimate a cost of capital of 9.74%. Again, for perspective, note that the median cost of capital for US firms in September 2026 was 9.16% and that the third quartile for US firms is 9.88%. If you disagree, here are your choices:

My advice, and feel free to ignore it, is that you should not spend too much time finessing the cost of capital, since of the many assumptions that drive this valuation, it is among the least critical. While I highlight my base case cost of capital, remember that when investing in a publicly traded company, what you would like to use as a cost of capital has to be constrained by what the market is offering. 

b. Financing risk: The main talking point for many Anthropic bears is the compute commitment that runs into hundreds of billions, creating fixed costs that will be a drag on profitability, especially if revenue growth falls short of expectations, as well as the risk of distress. While bondholders are right to be trepidatious and will point to the current bond rating of BB+ for the company as backing for their concerns, there are two reasons why failure risk will be suppressed. The first is that the company has the cash balance to sustain itself through the near terms and meet its contractual obligations. The second is that, as a high-profile trillion-dollar plus publicly traded company, it will have access to capital. I do believe, though, that if revenues grow at anemic rates, both these benefits will dissipate, as cash gets burned through and the market cap collapses, and I will estimate a probability of failure of 10% for the company, with a loss of half its value if that happen. If you want to disagree, go into the input sheet worksheet of the spreadsheet, and if you scroll down to cells B52 and B53, you can put in your own views:

c. Regulatory risk: With AI companies, the question with regulatory risk is not whether it will show up, but when it does, what form it will take. Anthropic and OpenAI have both invited the government to restrict how this business will grow, and while the cynical rationale for these pleas is that it may keep out competition (especially from China), the regulations and restrictions are coming. As with the large social media companies, the effect of these regulations will not be a cessation of business but an increase in operating expenses and loweer profit margins. I estimate that the regulatory effects are likely to be mild, at least on a relative basis, and that it will shave my initial operating margin estimate of 32.5%, for Anthropic, down to 27.5% (a reduction of 5%). You can make your own call on the regulatory effects on margins below:

You may disagree with me both on the likelihood of regulation and its effects, and while my base case is highlighted in black, you can diverge in your choice.

The Valuation

    With my storyline and inputs for Anthropic in place, I valued the firm, and you can check out the spreadsheet at this link. Rather than drown you in cash flow line items, I will present my Anthropic valuation, with a trillion in revenues by 2036, a target operating margin of 27.5% (with the target margin of 32.5% adjusted for regulatory drag), reinvestment that delivers three dollars of revenues for every dollar of capital invested over time and a cost of capital of 9.74%:

Download spreadsheet

With my upbeat story for Anthropic, I estimate a value for the equity of $1.475 trillion, at the lower end of the offer pricing range ($1.5 to $2 trillion). Note that my debt for the company is close to half a billion and includes my estimate of the present value of their compute commitments, and that my cash balance includes an estimated $100 billion in proceeds from the offering. If you are bullish, you will undoubtedly find ways to tweak my numbers, and find that a slightly higher operating margin and a lower cost of capital or a longer growth period, will get you to whatever number you want. If you are bearish, you will do a similar exercise in reverse, lowering growth and raising discount rates, to capture your priors.  It was precisely to avoid this gaming that I tried to get you to make your own choices on revenue growth, margins, reinvestment and risk along the way, and if you had the time and patience to play along, here is what the range of outcomes would look like. 

• The worst-case Anthropic story, where its target revenues limp (at least relative to  expectations) to $200 billion in 2036, margins converge on the mass market (10%), compute power stays scarce and costly (sales to capital of one), cost of capital rises to the third quartile for US companies (9.88%) and the regulatory effect on margins is drastic yields worthless equity, In fact, the value of equity is -$373 billion, and the company will not survive as a going concern.
• The best-case Anthropic story, where target revenues surge to $1.5 trillion in 2036, margins converge on the premium market (40%), compute power becomes cheap (sales to capital of four), cost of capital falls to the first quartile for US companies (7.79%) and the regulatory effect on margins is non-existent yields a value of $4.04 trillion for equity.

While I believe that the best case and worse case scenarios are both possible, but unlikely, you can assess your own story for the company, with the resulting value, applying what I call the 3P test: Is it possible? Is it plausible? Is it probable? Investing is a personal decision, and whether you choose to buy Anthropic, sell it short or sit on the sidelines, you should base it on your own assessment. In fact, just to get a sense of what the crowd judgment on the Anthropic IPO looks like right now, I have created a shared spreadsheet on Google, where you can enter your numbers and see how they compare to others valuing the company. I have entered my name, because I don't see a point in being anonymous, but if you want to use a pseudonym for anonymity, I don't have a problem with that.

Corporate Governance and Lock-up Provisions

    With publicly traded companies, corporate governance obviously matters, but in the last few decades, that recognition unfortunately has led us away from a sensible rationale of why it matters. In fact, much of academic research on corporate governance as well as practitioner measures of it have become focused on the minutia and details of governance, with number of board members, length of their tenure and board composition (insiders vs outsiders) all vying for front place. The truth is that corporate governance scores, which claim to measure how strong or weak governance at a company is, have been gamed by companies, making them poor indicators of shareholder power.

    So, what exactly is corporate governance? In my view, it is a measure of the power that shareholder (in public) and outside investors (in private businesses) have to change the way businesses are run, if they disagree with the way they are run by incumbent. It is therefore true that this power will be greater at companies where boards are more willing to push back against CEOs and activist investors are involved, and will diminish when you change the rules of engagement to make it more difficult for shareholders to push for change. I have long argued that the institutional investors who sat on their hands, when Google decided to break a precedent, at least in US markets, at its initial public offering by issuing shares with different voting rights. Their success has opened the door for a host of tech companies that have used the same structure to protect founder power at companies and effectively reduced the power that shareholders at these companies have to push for change to selling their shares and moving on.

    With Anthropic, these concerns about entrenching founders at companies is magnified by two choices that the company has made:

1. Voting rights: This company has decided to exhaust the alphabet in terms of share classes, with five classes of shares, with different voting rights.
   
   
   
   Note that the class A shares come with standard voting rights, but the voting structure, with a benefit trust picking the board and the founders controlling 50.1% of voting, no matter how many more shares may be issued in the futur, effectively emasculates public investors. 
2. Public benefit corporation: The company has structured itself as a public benefit corporation, and in its prospectus, it states that "we may make decisions that we believe appropriately balance the pecuniary interests of our stockholders, the best interests of those materially affected by our conduct, and the Company Purpose, but that negatively impact the short-, medium-, or long-term performance of our business or the market price". Given the worries that people have about AI being a force for destruction, a concern that Dario Amodei has actively cultivated, this protection against shareholder power may seem to be a good thing, but we need to be open eyed about what this structure does. It does not protect society from AI's evils, but puts power in the hands of the few (the founders and members of the benefit trust that they pic) at the top of Anthropic, hoping that they use that power well. If you, as an investor, have had bad experiences with top management at companies using corporate responsibility and ESG as cover for self-serving activity and hypocrisy, the Anthropic structure is that structure on steroids. 

As an investor who holds shares in other companies that have dual classes, with different voting rights, including Alphabet and Meta, I will not draw a line in the sand and assert that I will not invest in companies that have "fixed" the voting process in favor of founders. That said, I think Anthropic has managed to put those companies to shame with its combination of a benefit corporation structure (which gives the company the legal license to do pretty much whatever it wants with your money) and a voting share design that is so deviously self-serving that it would put Kim Jong Un to shame.

    In its nineteen IPO rounds, Anthropic has already accumulated an investor base that includes some of the biggest names in venture capital, as well as other companies in the AI game. I know that for some looking at the Anthropic IPO, there is the concern that these investors will cash out at the expense of public market investors, and that concern will probably be magnified by the fact that the company will, unlike in many other IPOs, allow some investors to cash out at the offering, with more shares being unlocked over time as the company trades. On the first trading day, insiders (but not founders and certain major figures) will be permitted to sell up to 10% of their shares. The remaining shares will be unlocked in tranches, across the first 180 days of trading, and founders and key named figures will be allowed to trade after 365 days. To be honest, I think that too much is made of share unlocking, perhaps because it gives investors and journalists something to talk about before unlock dates. With companies like Anthropic, where you have large institutional investors in the game, and plenty of liquidity, the effects of these unlocks is likely to be muted. My view, though, also reflects a belief that the smart money (which presumably includes all of those selling at the unlock) is not that smart and that their selling has far less informational value than you might believe.

    As you look at both the corporate governance and lock-up provisions that Anthropic has put into its prospectus, it seems to me that they are daring institutional investors to make a choice of speaking up against these provisions (by not partaking in the IPO) or staying passive, and have already calculated that they will do the latter. 

Trading the IPO

    I would draw your attention to my last post on Anthropic, where I argued that the company will be priced, and that almost everyone who trades the stock right after the offering will be pricing, not valuing the stock. That is not meant as a slight, but as a reality check for investors, and especially those investors who seem to believe that the market owes them a payoff, just because they care about fundamentals. Markets owe nothing to any of their participants, and as an investor, I have to not only recognize that reality, but act accordingly. Consequently, I will keep an open mind and wait for Anthropic to be priced for its IPO, and if it is priced well above my estimate of $1.48 trillion, I will sit out the frenzy. To the question of why I would not sell short, if it is priced well above my value, my response is that while I have faith in my valuation, albeit with a range around my estimate, I have no reason to believe that the market will or should agree with me, and even if it has made an error,  correct within my time horizon. 

    Once Anthropic starts trading, its story will continue to evolve and change, and given AI's presence in the economy and markets, I can guarantee you that it will be a roller coaster. It is possible that we have underestimated the possible payoff from AI, and that Anthropic continues to rise in market cap, but it is just as possible that major parts of the storyline, from revenue growth to margins to risk, will come under stress, leading the market to reassess what it thinks Anthropic should be priced at. Obviously, those stresses will also cause my estimate of the value to change, but it is not just a possibility but almost a guarantee that there will be a day that Anthropic will be undervalued, and it may be sooner than you think. I valued Facebook at the time of its IPO in June 2012, and while I found it over valued then, at a stock price of $36, it became a bargain just three months later, when the price hit $18, making me a buyer of the stock. The structure of Anthropic, with the public benefit facade and the voting rights abuse, will require a much larger discount on my value estimate, but the market may accommodate that requirement as well. 

YouTube Video

Anthropic Prospectus

1. Prospectus (S1) filing on October 26, 2026

Anthropic Valuation

1. My Anthropic IPO valuation (and you can make it yours)
2. Google shared spreadsheet for your valuations

Blog Posts on AI in 2026

1. To a Trillion Dollars and Beyond: A SpaceX Odyssey!
2. Revisiting SpaceX: A post-Prospectus valuation
3. OpenAI, Anthropic and SpaceX: The S&P 500 Inclusion Question and Investment Consequences
4. AI Scenarios: From Doomsday to Do-nothing!
5. AI's Bar Mitzvah: From hope & hype to business questions!
6. Stock Prices, Earnings and Cashflows: The AI Effect!
7. From Eyeballs to ARR: Buzzy Pricing Metrics and the Anthropic Pricing

From Eyeballs to Annual Run Rates (ARR): Buzzy Pricing Metrics and the Anthropic Trade!

    I was in San Francisco last week to talk to NYU Stern Alumni on October 15, and it was the day that Anthropic had a pre-IPO meeting with prospective investors. While I was not at the meeting, the focus  of teh gathering was laying a foundation for its upcoming IPO, with the company reporting third-quarter revenue of $16.8 billion, which news reports described as surging 45% over the previous quarter's revenues, and positive profits, albeit on an adjusted basis. During the meeting, Anthropic also reported an annual revenue run rate (ARR) of $67 billion, a metric that those in the AI space have adopted with gusto, with an estimate that the ARR would exceed $100 billion by the end of the year. The nature of ARR, where choices on how revenue is calculated (gross versus net) and on annualization periods (daily, weekly or monthly) can make a significant difference, has led investors to be skeptical about comparing these numbers across time and companies.  While I share some of that skepticism, it is worth remembering that young businesses are often priced on inventive metrics, some with more basis than others. In the late nineties, dot.com companies were priced not just on the number of visitors to their websites, but on eyeballs, where Mary Meeker argued that ad spending at these companies was correlated with time that these visitors spent on websites (eyeballs), rather than with total visitors. In the last decade, you saw social media companies priced based upon the number of users and subscription-based companies on subscriber numbers. In this post, I will try to lay out the rationale for why pricing supplants valuation in young companies and why that pricing almost always has to be built around messy metrics like eyeballs, users and ARR, rather than more conventional metrics like earnings, book value or cash flows.

Young Companies in a Corporate Life Cycle

    All businesses, no matter how large and established they might be today, started as young businesses and matured. With that perspective in mind, I will start this discussion by placing young businesses in a corporate life cycle in the figure below:

One of the first steps in analyzing a company is placing it in the right stage of the life cycle, and there are three choices that you can use in making this placement:

1. Chronological age: The first and simplest choice is to look at the age of a business, putting companies that have been in operation for short periods into the young company grouping and classifying those that have been in operation for longer as mature. While this approach is easy to use, and for the most part, effective, it can be misleading in businesses like infrastructure, where it can take decades to get established, and in technology, where the lack of capital intensity can allow a company can go from start-up to mature in a few years.
2. Industry grouping: Another choice is categorizing industry groups into life cycle stages and assuming that any company that belongs to a specific industry group shares its life cycle characteristics. For decades, investors have viewed technology as a shorthand for growth and put technology companies into the earlier stages of the life cycle and utilities as mature or even declining and put all utility companies into the mature to declining company grouping. The limitation of this approach is that there will be divergence within each group, where some companies separate themselves from their industry group; there are mature and even declining companies that are in the technology grouping, and there are power companies that have taken advantage of the data center boom to become growth companies. With technology in particular, where companies age in dog years, I talked about how companies can get mis-categorized and sometimes mismanaged as growth companies, even though their best days are behind them. 
3. Operating metrics: Ultimately, life cycle effects show up in the operating metrics of companies, and there are common characteristics that young companies share, and they include the following:

• High revenue growth: Young companies usually have high revenue growth, and while this is usually off a small revenue base, there are exceptions, where companies with scaled up revenues are still able to show high growth.
• Operating losses: As young companies scale up, they usually struggle with unformed business models, and at low scale, their fixed costs overwhelm their revenues, resulting in negative or very low operating profit margins.
• Negative free cash flows: Free cash flows are after reinvestment, and as young companies invest to deliver on their potential growth, that reinvestment will generate negative free cash flows.
• Failure risk: The confluence of small revenues, losses and negative free cashflows, which must be funded with external capital (equity or debt) exposes young firms to significant risk of failure.

The figure below captures the shifts in operating metrics, as companies move through the life cycle:

Put simply, if a company has revenue growth in single digits, stable margins and positive free cash flows to the firm, it is a mature company, no matter how it views itself. Conversely, a firm with high revenue growth, negative or very low profit margins and large negative free cashflows to the firms is a young growth firm, no matter what its scale. Recognizing those characteristics is critical in understanding why so many give up on valuing young companies, and why pricing will often be based on unfamiliar metrics.

    How do Anthropic and OpenAI measure up in the life cycle? Using a patchwork of public data and leaked information, here are the numbers on key operating metrics that I have as of October 2026 for both companies:

The sky-high revenue growth and operating losses are a clear indication that these are young, growth companies, but note that the revenues (albeit on an ARR basis) at both companies are already large, suggesting that they are growing at scale. That should come as no surprise, since they are both rumored to be looking at IPOs in excess of a trillion dollars, in terms of total market capitalization. The business models are still in flux in both companies, as they wrestle with how best to structure their business models (usage versus subscription) and struggle with market targeting questions (premium versus mass market). I know that Anthropic claimed to be profitable in the second quarter of 2026, but I don't think that even Anthropic believes this to be true, since these are adjusted profits and the adjustments represent a patchwork of add-backs of legitimate expenses. As for cash flows, both Anthropic and OpenAI are deep in the hole, with negative free cash flows, at the moment, and whether they are able to generate positive cash flows in the future is entirely a function of how the AI business evolves over time.

Pricing versus Valuation: Across the life cycle

   I will start with a distinction that I believe needs to be made between two approaches to attaching numbers to assets and businesses. The first is to value that asset or business, based on its capacity to generate and grow cash flows, and the risk in those cashflows, i.e., intrinsic valuation. The second is to price that asset or business by looking at what others are paying for "similar' assets, thus reducing pricing to demand and supply. The picture below captures that difference:

It is true that if you believed in efficient markets, price would converge on value, but if you did, you would save yourself the time and aggravation associated with trying to price or value assets by investing in index funds, and moving on to more productive pursuits. If you don't believe in efficient markets, you can take one of two paths. You can take the investor route,  trying to value assets and businesses, with the intent of buying undervalued and/or selling overvalued companies, and hoping that price converges on value. Alternatively, you can take the trading path, where you throw up your hands on intrinsic value, arguing that it is too subjective and difficult to estimate, and make your money by buying investments at low prices and selling them at higher prices,  using whatever tools (including charts and technical indicators) you want to use, to make your judgments. 

    While you can value and price any business, there are reasons why the usage of each can vary depending upon where a company is in the life cycle. To value a business, you have to estimate expected future cash flows, and to make these estimates of cash flows, growth and risk, and you generally draw on three sources of information:

1. Financial statements: The current financial statements of a company provide information on unit economics, profitability and risk, and historical financial statements are useful in assessing growth in the business and changes in business profitability. For mature businesses that have been in operating for long period, much or even all of what you need to value the company can be in its financial statements, but for younger companies, the current financials may not yield much (as growth and unformed business models create noise) and there is little history,
2. Market size and growth: Knowing the size of the total market that a company is aiming for, and how quickly it is growing can help put reasonability constraints on the growth rate and market you estimate for a company. For companies in established markets, this should be relatively easy to do, but for companies in new and evolving markets, this can become difficult.
3. Industry statistics: As you forecast profit margins and the risk profile for a company, having that information for the crosssection of companies in the industry can establish bounds and help in making estimates.

If you frame an intrinsic valuation in terms of the big estimation questions that need to be addressed to arrive at value, starting with the cash flows from existing investments, moving on to estimating the value that will be added by future growth and assessing risk, you can see that valuing young companies will always be more complicated than valuing more mature businesses. In fact, looking at Anthropic and OpenAI, you can see why valuation will be daunting:

While I am a glutton for punishment, and I will try, albeit with error, to make my best judgments on Anthropic's future cash flows, risk and value (as I did with SpaceX), when the prospectus is made public, I understand why most will turn away in the face of uncertainty and choose not to value the company. Some of these investors will also look at how those that they believe know more about that companies are assessing its value, and peruse prior venture capital rounds at both Anthropic and OpenAI for clues on value.  Both Anthropic and OpenAI have been active in raising equity as private companies, and the most recent rounds yield estimates of what these private investors thought they were worth, at least at the time the capital was raised. The table below summarizes the number of VC rounds, with capital raised, at both companies, as well as the pricing in their most recent rounds:

Anthropic's VC round in May 2026, from a cross section of big-name venture capitalists, raised $65 billion for the company, at a post-money pricing of $965 billion.  OpenAI's most recent capital raise was in March 2026, when it raised $122 billion, at a post-money pricing for the company at $852 billion. OpenAI also has a pending round of capital raise ahead of an IPO, where they expect to raise $30 billion with a pricing for the company of roughly $1.4 trillion. 

    While you may be tempted to assume that the VC pricing in these rounds are rooted in value, venture capitalists are in the pricing game, as I have argued in prior posts, In fact, they, stretch pricing metrics and definitions of peer groups, and VC valuation is almost always forward pricing:

In the case of Anthropic and OpenAI, this forward pricing process is both more compressed and even less informative than at most young companies, with investors viewing the companies as imminent IPOs, and pricing them based upon their expectations of the pricing of the public offering, and discounting that expected price for negative surprises about the company's prospects that may lower the pricing or delay/cancel the IPO:

The expected pricing will be a function of pricing in prior VC rounds, management wishlists and even rumors, and the discount applied will increase as the time to the IPO lengthens and with the uncertainties about the company and its potential market. For Anthropic, for instance, at the time of the VC round, the expected pricing for the IPO was $1.15 trillion, well below what the expectations are today, and reflective of how much can change over short periods with young companies, and the $965 billion pricing in the VC round would have represented a $150 billion discount on the estimate.

Pricing Process

    While the reluctance to value young companies, and fall back to pricing is understandable, I would push back on the notion that doing so dispenses with the need to make the assumptions about cashflows, growth and risk that are required in intrinsic valuation. In this section, I will break down the pricing process into steps, and look at the issues that you faced when valuing these companies also create impediments in pricing, and in the next section, why using a multiple of annualized revenue run rates (ARR) or even eyeballs makes implicit assumptions about future profitability, growth and reinvestment. 

The Steps in the Process

    If, in pricing, you attach a number to an asset or business, based upon what others are paying for similar assets or business, how exactly does this process play out when valuing a company or business? No matter who does the pricing, and how sophisticated they claim to be, there are three steps in pricing companies:

1. Standardize the price: The nature of pricing shares in public markets is that the share price is a function of share count, which is arbitrary; a stock split for instance halves the price while doubling the count and total market capitalization or value is a function of company size. To allow for comparisons across companies, you standardize the market price by dividing by a common variable that you can obtain for each of the companies in your comparison, with the choices ranging from revenues (or revenue drivers) to earnings to cashflow variants to book value. Since the market price can be for just equity (market capitalization) or for the entire business (enterprise value, which includes net debt), the denominator has to be defined in the same terms, with equity (operating) income and book value of equity (invested capital) going with market capitalization (enterprise value).
2. Find peer group companies: For pricing to be functional, you have to find other priced companies just like the one that you are pricing. The simplistic solution is to start with other companies in the same industry or business grouping, with added constraints on geography and company size thrown in. Needless to say, even if done with care, the choice of companies in the peer group is subjective and can vary across analysts and investors, even when looking at the same company.
3. Control for differences: No mater what pricing metric (PE, EV to Sales etc.) you pick, there will be fundamental reasons why companies will trade at different values, and for a pricing to have substance, you have to control for differences. Thus, if PE ratio is your metric, a company that expects higher growth in earnings or lower risk than its peer group, should trade at a higher PE ratio than the peer group does, and that difference has to enter your discussion.

The picture below captures these steps, with the variants on pricing metrics and the considerations in peer group construction included:

While this process remains the same for every pricing, there are reasons why pricing metrics will vary across companies and why peer group construction may be easier in some cases than other, and I will take Anthropic and OpenAI through the process.

1. Pricing metrics

    With some companies, your choice of pricing metrics will span the spectrum, with revenues, earnings, cash flows and book value all competing as scalars. However, your choices will narrow at others, and one reason that is relevant to the Anthropic/OpenAI discussion is where a company falls in the life cycle:

As you can see, early in the life cycle, when companies are losing money, have negative free cashflows and little tangible capital invested in assets in place, you will see revenue often as the only viable metric, leading to revenue multiples dominating the discourse. Put simply, desperation drives you up the income statement until you find a positive number. It should be no surprise, therefore, that the metric that companies and investors have focused on with Anthropic and OpenAI is revenues. Those revenues are growing so quickly at these companies, that their argument is that trailing numbers are dated by the time hey are released, and that has led to the shift to annualized revenue run rates (where using revenues from the last week or month allows for more updated vales). As we noted earlier, the perils with annualized revenue rates (ARR) is that they are not standardized and can be easily gamed. Just in the last few weeks, there have been news stories on the divergence between gross and net revenues with OpenAI shifting towards using the latter, lowering ARR from $70 billion to $50 billion, while Anthropic continues to report gross revenues and the cloud provider payments as expenses. In addition, you can compute ARR using the last day's, the last week's and the last month's revenues, and while Anthropic and OpenAI have both converged on monthly annualization, there is nothing that binds them to continue to do so.

    There is a second choice that will come into play as the debate about Anthropic's pricing becomes more heated, and that is whether to use revenues in a future period, i.e.. forward revenue. By forecasting the annualized revenue run rate of $100 billion plus by the end of 2026, Anthropic has already opened the door, but do not be surprised if investors and analysts shift to looking at revenues in 2030 or 2036, since scaling prices today to forward revenues five or ten years out will make the pricing multiple look lower, and thus more defensible. Once you decide to go forward in time, it is plausible that you can even shift to expected earnings in 2031 or 2036, built around expected operating or net profit margins; arguing that Anthropic is cheap because it trades (today) at 25 times 2036 earnings (based on your estimated revenues and net margins) brings the multiple back to earth, but at what cost? Investors use frames of what comprises normal or high, based  upon their acquaintance with a multiple, and many will (mistakenly) compare these forwards earnings multiples to current earnings multiples in the market to make their investment judgment on cheap or expensive.

    There is a third choice, and that is to relate the pricing of these companies to a variable that you believe will be correlated with future revenues and earnings. With that choice, you can see the rationale for estimating enterprise value per user or subscriber for user and subscriber based companies, with the connector being the logic that more users and subscribers today is a leading indicator of higher revenues and earnings in the future. It was that logic that led Mary Meeker and other internet analysts to use website visitors and eyeballs in pricing dot com companies, and while you may disagree, the logic holds. With OpenAI and Anthropic, where the power, reach and revenues of AI agents is dependent on having access to compute power, you could scale the market cap and enterprise value at these companies to the compute power at each company, based upon either measures of compute power (in gigawatts) or money spent to acquire the power (compute commitments). The downside of using metrics that are specific to a business (users, eyeballs or compute power) is that you can compute that multiple only for the companies in this business, restricting comparisons, and you have no perspective on what comprises high, low or typical for that multiple. 

    The table below captures the pricing metrics that emerge with each of these choices for Anthropic and OpenAI, at $1.5 trillion and $2 trillion dollar market capitalizations:

As you can see, you can make the case that Anthropic is cheaper than OpenAI, based upon the most recent ARR or the ARR in 2026, but OpenAI is cheaper relative to expected revenues in 2030, compute costs/capacity and based on user count. Each of these pricing multiples comes with caveats, with obvious explanations for differences, some in how the metric is computed (gross revenues for Anthropic versus net revenues for OpenAI) and some in fundamentals (Anthropic is priced more highly more user, but given that its usage base is more enterprises than individuals, and that they pay based on usage, each Anthropic customer should be priced more highly than an OpenAI customer).  No matter which of these pricing metrics you use, though,  the truth is that it is entirely possible that both Anthropic and OpenAI may be overpriced, and by comparing the two,  you are just finding the less overpriced entity.

2. The Peer Group challenge

    Pricing loses much of its power, without access to a peer group of similar companies that are priced in the market place. With a software, chemical or retail company, pricing becomes easier because you have dozens, or even hundreds, of companies that you can define as being in your peer group. It is true that the company you may be pricing may be an outlier in terms of growth and risk in your peer group, but that is a problem for the next section, where you have to control for differences. Pricing becomes much more difficult, when you have unique companies or companies in cutting edge businesses, where there are a handful or even no other companies that are comparable, with market pricing. Using both of these components (number of companies in the same space and whether they are publicly traded), there are four scenarios that play out, ranging from least to most difficult.

• If you are pricing a company in a business, where there are lots of publicly traded companies, you will not only have a choice of pricing multiples, but also more robust pricing, since each of the companies has an updated market price and the financial statements are standardized (GAAP or IFRS). 
• If you are pricing a company in a business, where there are only a few publicly traded players, you may have a choice of multiples, but small sample size will reduce the power of your pricing, though having a long historical pricing record for all of the companies may sometimes compensate. For much of the last century, in the United States, GM and Ford were priced against each other, with traders developing ranges of what each company should trade at relative to the other.
• If you are pricing a company in a business, where there lots of companies that are privately held, as is often the case when you have a nascent business with lots of potential, you are confronting a challenge that many venture capitalists face, and you can try to get the most recent pricing for each company, though that pricing will often be dated (depending on when the last capital raise was) and sometimes untrustworthy.
• Your most difficult challenge will be when you are pricing a company in a business, where you have both very few companies in the space, and almost none are publicly traded. 

If you define peer group for Anthropic and OpenAI as other companies in the LLM space, you are in effect in the last category, since neither of the companies is traded yet, and their competitors are either parts of bigger businesses - Gemini is a small part of Alphabet, xAI is a bigger part of SpaceX which has other functioning and profitable businesses and Muse is obscured by Meta's advertising business. When Anthropic's IPO show hits the road, the lack of a peer group will be an impediment, but investors and traders are creative and come up with solutions:

• In the narrowest version of the pricing game, Anthropic and OpenAI will be twinned with each other, since they are not just competitors but also playing the same business game. Initially, or at least until OpenAI files its prospectus and goes public, the comparison will be between a public pricing for Anthropic and private pricing for OpenAI, with SpaceX thrown into the mix as a noisy peer.
• in a broader version of the pricing game, and especially if Anthropic is even close to its $1.5 trillion market cap, the comparisons will be to the Mag Seven companies, all of which command market capitalizations close to or well in excess of a trillion. The argument will be that investors who seek out large market cap companies will compare them to each other, even though Tesla, Apple and Anthropic are very different companies in terms of business model and where they are in the corporate life cycle. You could also bring in SpaceX, not part of the Mag Seven, but also a trillion dollar company perhaps at about the same stage of the life cycle: The table below compares Anthropic's pricing (using a target pricing of $1.5 and $ 2 trillion) to the pricing of the Mag Seven companies, as well as SpaceX:

As with the pricing comparison with OpenAI, there is good news and bad news for Anthropic bulls and bears. The bulls can argue that Anthropic is cheaper than SpaceX, and that if it looks expensive relative to the Mag Seven, it is because it has higher expected growth. The bears will push back, noting that the AI buzz that is carrying Anthropic is also pushing SpaceX to highs, and that the yawning gap between the Mag Seven companies and Anthropic on revenues and earnings (which are not reported in this table, but make the case for the Mag Seven stronger) cannot be explained away with hope and potential. 

3. The Connection to Value

    You can price a company based upon any metric that you choose to use, but if pressed on which metric is the best one, I will deflect (and perhaps deliver a non-response), because I don't believe that there is a "best" metric for pricing. The key, though, is that no matter what metric you pick, you should think through how that metric will play out in the company's value and what you need to control for in your pricing. Let's start simple with PE ratios and look at the chain that connects current earnings to overall value:

The first link is earnings growth, to generate future earnings, with higher growth leading to higher value. The second is the efficiency with which you deliver that growth, which results in reinvestment, and converts earnings to growth; a company that can earn a higher return on equity on its reinvestment should be able to generate more in cash flows, for the same earnings, than a company that earns a lower return on equity. Finally, risk comes into play when you estimate the present value of future cash flows; less risky cash flows will allow you to use a lower discount rate and attach a higher value for equity. Put simply, when comparing PE ratios across companies, you have to control for differences in earnings growth, return on equity and equity risk.

    Since PE ratios are a non-starter with Anthropic and OpenAI, I will turn to enterprise value to revenues (using either the ARR or the trailing 12 month revenues):

As you can see, for Anthropic to be highly priced relative to revenue, its revenues will have to not only grow a lot, but do so quickly, making the revenue growth rate and the speed at which it ramps up the first determinant of pricing. The second component is profitability, since the company is reporting sizable losses now, and can expect to do so in the near term, the "adjusted" profit kabuki dance notwithstanding. Here, there are two key determinants that will drive value, with the first being unit economics, since having higher gross margins will ease the pathway from losses to profits, and the second being economies of scale, with compute costs in particular growing slower than revenues, pushing up operating margins. The third is risk, with operating risk (from evolving business models), financial risk (from the compute commitments) and regulatory risk (from restrictions on AI) all coming into play; companies that are less exposed to these risks should be worth more. To make a case for Anthropic at a $1.5 or $2 trillion pricing, you will have to make defensible arguments for sky-high revenue growth in the early year, a surge in operating margins to 30% or higher, with favorable unit economics and economies of scale providing tailwinds and a risk profile that quickly converges towards the average, as regulatory risks recede and the business model finds it feet. While you may find this story to be a reach, it is a plausible one, and I plan to give those who tell the story a chance to make an argument, rather than dismissing it outright.

Conclusion

    As Anthropic heads towards its initial public offering, my suggestion is that you step back from the frenzy, since like the parable of the six blind men who visit an elephant, and having touched different parts of it, come away with wildly different descriptions of what an elephant looks like, there will be Anthropic bulls and bears who will latch on to a part of the company's lore and spin tales about its future. Thus, an Anthropic bull whose entire argument for why you should buy the company is that it has had high revenue growth in the last year or that gross margins right now are 80% is telling part of a story, and not a very persuasive one, whereas a bear whose entire thesis is built around the company losing money and having negative cash flows is missing the narrative train just as badly.

    A full (bullish) story for Anthropic will be anchored in a belief that AI's reach in terms of market will give it a massive addressable market, that Anthropic will be one of the big winners in that market and that it will find a way to convert its dominance into market share and high profits; the high pricing will be justified with high revenue growth and great unit economics (with high gross margins) arguments. For bears, a fully thought through story will be premised on an AI market that struggles to get off the ground, partly because of flaws in AI agents, partly because of competition from cheaper open-source models and partly because of regulatory constraints, and Anthropic, as a company, is constrained by weak corporate governance and an untrustworthy management. The overpricing argument will point to the revenue growth that will be lower than anticipated, which combined with their compute commitments will lead to profit margin compression, will leave you at a much lower justified pricing. I will not be taking sides here, but not because I am copping out. It is because I am investor at heart and feel more comfortable making my decision based upon an intrinsic valuation of Anthropic, which will have to wait until the full prospectus gets filed (hopefully next week). 

YouTube Video

Stock Prices, Earnings and Cashflows: The AI Effect plays through!

In a post at the end of August 2026, I talked about interest rates in 2026 and marveled at the capacity of equities to keep rising in the face of rising rates. I argued that the resilience of stocks during the year could be traced to higher-than-expected earnings being reported by companies in 2026, and a concurrent increase in expected earnings in 2027 and 2028. Now that September 2026 is one for the record books, it is time to take stock and dig a little deeper, especially since the month brought about one of the largest increase in treasury yield rates in recent memory, and stock prices still held their own. In particular, I want to examine the earnings at US companies, in the aggregate and by sector, trying to trace out where the earnings increase is coming from, and how that increased earnings is playing out on corporate balance sheets and cash flows.

Stock Prices and Rates

In my earlier post on interest rates, I had looked at treasury rates by day through the end of August, and I will begin this section by updating that chart to include a tumultuous September:

I had described the rise in rates between January and August as gradual, but the rise in rates in September was anything but, as the ten-year rate rose from 4.75% at the start of the month to 5.29% at the end. In fact, to put the 54 basis point rise in the ten-year rate in context, take a look at the distribution of monthly rate changes in the ten-year treasury in the chart below:

The September rise in rates would put in the top ten percent of 770 monthly rate changes that we have seen between 1962 and 2026. In the face of the mark up in rates, stocks held their own in September, at least in the aggregate, and you can see that in the chart below, where I look at aggregate market values, by month, and by sector:

=

The market added $2.5 trillion in market capitalization across all stocks in September 2026, but almost $1.5 trillion of that came from technology As you look across the entire year, broken down by quarters, here is what you see in the aggregate market caps:

Only companies listed at the start of 2026 included

I excluded firms that were not listed at the start of 2026 from the list since including them will give a misleading sense of returns to investors; adding SpaceX, for instance, which was listed in June 2026, will increase the value of communication service companies by more than $2 trillion, but it was listed at roughly that value. There are many who have pointed to the fact that US equities, while up for the year, have seen divergence in performance, and you can see that phenomenon play out in two statistics. The first is that three sectors - technology, energy and materials have carried the market, with technology being the biggest contributor to market gains. The second is that the percent of companies within each sector that are up for the year is about 50% across the market, and in the third quarter, about 65% of all listed stocks saw dropping stock prices.

    It is often difficult to make sense of what is moving stock prices because there are so many factors from growth to interest rates to cash payout that are pulling in different directions. It is to counter this confusion that I have resorted to estimating an implied equity risk premium, where I estimate the internal rate of return you can earn by buying equities, given how they are priced, and their expected earnings growth and cash flows, as well as  interest rates. That calculation, which I have done at the start of each month since September 2008, yields an expected return of 8.99% for equities and an equity risk premium of 3.70% (3.92%) over the ten-year treasury rate (dollar riskfree rate) of 5.29% (5.07%) at the start of October 2026: 

Note that this estimation is model-agnostic and is an internal rate of return for investing in stocks, given expectations of the cash flows from investing in equities. If you are a market timer, and I am not one, you could use this implied equity risk premium as a barometer of market priciness, with a lower number indicating overpricing and a higher number indicating underpricing. 

Download spreadsheet

The jump in US treasury rates in September 2026 has had an effect, with the equity risk premium dropping below 4% for the first time this year.  In fact, even as the ten-year treasury rate has climbed this year from 4.18% to 5.29%, the expected return on stocks has also gone up from 8.41% at the start of 2026 to 8.99% on September 30, 2026.

The AI Cap Ex Boom: Accounting and Corporate Finance Consequences

    The focus on stock prices and interest rates can sometimes distract us from paying attention to corporate investing, financing and cash return policies that drive value. It should not surprise you, given the times we live in, that AI is at the heart of the business story that is driving corporate behavior, and in the process, providing the fodder for market resistance to higher rates. I will begin with an assessment of how the trillions of dollars in AI cap ex will show up in financial statements:

As you can see, the AI cap ex story is a complicated one, if you are looking at market aggregates, because the market includes both the companies that are spending the money building the AI architecture, which includes data centers and other infrastructure, as well as the companies that are supplying the ingredients for that infrastructure. 

• The builders of the architecture are the hyperscalers (Meta, Alphabet, Amazon, Microsoft et al.), and the money they spend on cap ex will cause lower earnings, at least until the cap ex starts paying off, in the form of amortization of the AI cap ex, as well as a hit to their free cash flows, which are after cap ex. 
• The money spent on AI cap ex though becomes revenues to the chip makers (Nvidia and TSMC leading the way), network equipment manufacturers (Broadcom, Micron and Marvel, to name just three), power plant builders (Constellation Energy et al.) and even real estate developers focused on data centers (Equinix, Digital Realty), and ultimately net profits (with net margins driving the bottom line). 

It is true that there are gray zones here, with some AI builders also  benefiting from being suppliers (Amazon is spending money on AI cap ex but is also benefiting from the usage of its cloud space for data storage, and Nvidia, while selling the chips that go into the architecture, is also investing directly or indirectly into data centers). 

    As we trace through the aggregated effects of the AI cap ex boom on accounting statements, there are two caveats that need to be stated up front. The first is that the data that is accessible to the public, and which I will be using, will be data from publicly traded companies. To the extent that some of AI's big players (builders and suppliers) are private, I will be missing the revenues, earnings, cash flows and invested capital at these large private players (which include at least two companies in Anthropic and OpenAI that are expected to command trillion dollar plus market caps. The second is that some of the AI cap ex is taking the form of joint ventures and off-balance sheet entities, and the accounting for these (especially on the debt side) may not fully reflect the consequences for firms. As a result, the numbers you see in the public company financials will understate the full effect across all businesses.

    With those caveats in place, the business story for US equities starts with massive capital expenditures in AI, with trillions being invested into data centers and AI architecture. It is true that this cap ex is top heavy, with the top ten hyper-scalers accounting for more than $2 trillion of the AI cap ex, but in the table below, I look at the aggregate cap ex reported in corporate financial statements at all publicly traded companies in the United States:

Even with the caveats about understatement, but you can still see that cap ex in the second quarter of 2026, which is our last completed quarter of reported financials, was up $133.4 billion from the cap ex in the second quarter of 2025, an increase of almost 36%. Again, the surge in cap ex is concentrated, with technology, communication services and consumer discretionary all registering growth of more than 50% in the quarter-to-quarter comparison. 

    Accounting incorporates capital expenditures into the balance sheet as assets, and reflects how this cap ex is funded (debt or equity) by increasing the book values of the funding used in the investment. A surge in cap ex, such as the one that we have seen in 2026, will show up as higher book values for equity, debt and invested capital, and we capture this effect, by sector, in the table below:

Across all US stocks, the book equity has increased almost 13%, between the second quarter of 2025 and the second quarter of 2026, and total debt is up almost 8%. In dollar terms, the book equity at US companies increased by $1.8 trillion between the second quarter of 2025 and the second quarter of 2026, and book debt by $1.9 trillion, over the same  period.  Technology, being the most active player in AI cap ex, has seen much bigger increases in both numbers, with book equity rising almost 30% and total debt up about 18.8%. 

    From an earnings perspective, the focus on cap ex and book values may seem misplaced, since the former can only decrease cash flows and the latter impacts accounting returns. In 2026, though, the increased capital expenditures on AI are affecting earnings for a simple reason. The money spent on cap ex by a company building AI architecture will become revenues (and earnings) for other companies that supply the building blocks for the architecture. There is a reason why Nvidia has been the biggest beneficiary from the AI cap ex boom so far, since its chips, marked up massively, power the data centers, and there others, from electrical equipment makers to power companies to real estate developers who have also reaped the benefits. It is true that there should be increased amortization expenses at the AI builders, but the longer amortization schedules being used by many of them is reducing the current hit to earnings at these companies. The earnings effect of the AI story can be seen in the table below, where I look at aggregate net income at US companies, broken down by sector:

Note that in both the first and second quarters of 2026, US companies have seen earnings surge over the corresponding quarters in 2025, with aggregate earnings increasing from $511 billion to $692 billion (translating into an earnings growth rate of 35%, quarter-to-quarter) in the first quarter of 2026 and from $576 billion to $904 billion (translating into an earnings growth rate of 57%, quarter-to-quarter) in the second quarter of 2026. As with stock prices, the earnings benefits are not broad-based, with more than half of all companies in the market reporting declines in net income, and there are wide differences in earnings growth across sectors. Technology, financials and communication services have seen the biggest increases in earnings, and health care, utilities and real estate have lagged.

    A cap-ex driven surge in aggregate earnings comes with an asterisk, since the higher earnings across firms will be partially or even fully offset by capital expenditures across firms, leading to free cashflows to firms often growing at much lower rates than earnings. Since these free cash flows are what fund dividend payments and stock buybacks, I looked at cash returned to shareholders in both forms in 2026:

In the aggregate, dividends in the last twelve months are up about $43.3 billion (about 5%) from dividends in the 2025 calendar year, and stock buybacks are up about $106.7 billion (about 9%). In fact, if you net out stock issuances, which spiked in the second quarter of 2026, from buybacks, net buybacks have grown bout 7% between the last calendar year and now. Those numbers represent reasonable step ups from the last year's numbers, but they clearly have not kept up with the earnings growth in 2026. One way to see the disconnect that is occurring between earnings and cash flows is to look at the cash returned as a percent of earnings for the S&P 500 companies over time:

For much of the last two decades, US companies have returned 80% or more, and sometimes more than 100% of their earnings, to shareholders in dividends and buybacks. Starting in about 2024, you can see a divergence with earnings rising much faster than cash returns, and in the last twelve months leading into 2026, the companies in the S&P 500 returned 63% of their earnings to shareholders, a low not seen since 2004. Many of those who were criticizing US companies for buying back too much stock and not investing enough back into businesses are now finding fault with those same companies scaling back buybacks and investing more into AI cap ex, leading to the conclusion that these critics will find fault no matter what companies do. 

The AI Business Resolution: Accounting and Market Consequences

    It is true that the massive investments in AI cap ex are being driven by expectations that AI as a business will enjoy not only a large market, but one where the winners can sustain huge profits for the long term. As I noted in my post on AI as a business, this is a plausible path, but there are vast disagreements on whether this is the expected one, given uncertainties about all three layers of the business story - the size of the total addressable market, the unit economics/operating margins of companies in the business and the moats and competitive advantages that will allow for sustainability in profits. In effect, the only point on which you have a consensus on is that companies have collectively invested a huge amount in AI cap ex, and the point at which AI optimists and pessimists is diverge is on what the payoffs from that investment will be, with optimists arguing that AI will deliver on expectations and that the profits that will be generated by AI cap ex investors in the future will be sufficient to earn returns on that investment that exceed the cost of capital, and pessimists positing that this is wishful thinking, and that the profits from AI, even if considerable, will not be large enough to justify the up front investments. It is also worth remembering that determining which side is closer to the truth will not be made by looking at arguments about annualized revenue run rates at Anthropic and OpenAI today, but will require waiting for the AI product and service market to gain substance. Waiting does not come easily to any of us, and there will be premature claims of victory from both sides, often based upon paltry evidence and incomplete information.

    So, what will the accounting and market consequences be, if the AI pathway diverges from expectations? In the table below, I trace out the accounting and market consequences of the AI business working better than expected at delivering growth and profits, as well as if i does much worse than expected:

• In the best case scenarios for AI, the companies that have invested in AI, at least collectively, will be able to deliver not just earnings growth from the cap ex, but enough incremental earnings to generate returns on the AI cap ex that exceed their cost of capital for those investments.  Their lenders will be made whole, with interest and principal payments, and the AI builders will see their cashflows  become more positive,  but the companies, while successful, will emerge as very different businesses than when they entered the space, more capital intensive than they used to be. 
• In the worst case scenarios for AI, there will be both accounting and market carnage, as accountants write off large portions of the AI cap ex, because of its failure to deliver promised profits, and while cash flows may recover, markets will correct the pricing of these companies to reflect a lack of trust in management. For companies that were excessively dependent on debt for their AI cap ex, there will be defaults and increased distress, with lenders feeling the pain as well. 

There are also intermediate scenarios, ranging from AI being a moderate success, where the companies investing AI may be able to extract some earnings from their investment, but not enough to cover the cost of capital, to a moderate failure, where some companies may be able to justify their investments and most will not.

    Given that AI business surprises will have both market and accounting consequences, I will hazard a guess that the market will lead in this process and accounting will follow. In short, if AI is working better (worse) than expected, you should see stock prices at AI-centered businesses rise (fall) before you see accountants respond. Put simply, in the event that AI does not deliver on its promise, waiting to act until accountants write off AI cap ex to sell your AI company implies that you waited too long.

An Investor Perspective: Taking Stock and Taking Action

    In my post on AI as a business, I zeroed in on the debate between AI optimists, pointing to huge (albeit unspecified) markets for AI products and services, and AI skeptics, drawing attention to the outsized capital expenditures. While that debate plays out, financial markets and businesses cannot afford to wait for resolution, and are acting now, with companies making investments in cap ex and markets building in their expectations of what that will mean for future earnings into stock prices. That has made not just the market but also the economy a giant bet on AI, with success vindicating the companies and investors who have bet on it, and failure manifesting in massive write offs at companies (feeding into losses) and stock price markdowns. As investors, there are four choices that you can make, and unfortunately, none of these choices give you the luxury of sitting out the AI debate:

1. Go all in on the AI story winning (and doing it soon): The first betis that AI is an unstoppable force, destined to change the way we live and work, with AI businesses reaping the benefits of the disruption. It is a plausible story, albeit one that raises significant questions about the economic and social costs of disruption, with disagreements about the speed and extent of the disruption. It was the story that Leo Aschenbrenner built Situational Awareness around, and while excessive leverage, driven by hubris and over-conviction, brought him down, it is possible that you could mimic his strategy, of buying the AI disruptors and/or selling the AI disrupted, albeit with far less leverage, and win in the long term.
2. Go with the market consensus: In an age where we worship at the altar of crowd wisdom in almost everything we do, you could examine what the market is pricing in, as its AI story, and go along. At the moment, at least, the market seems to be building in the expectation that AI will be a major disruption that will give rise to large and valuable businesses and it is picking its winners among the AI architecture companies (with Nvidia the biggest so far) and among the LLMs (SpaceX in the public markets and OpenAI and Anthropic in the private markets). For better or worse, you may have already chosen this path implicitly, if your pension funds and savings are invested passively, getting partial exposure to this story with an S&P 500 fund, and more complete exposure if you buy a total market fund for US equities. 
3. Be an AI skeptic: There are many reasons to be skeptical about the AI story, and for some, that skepticism may lead to the belief that AI will not make it as a viable business, or at least one large enough to sustain the pricing and investment you are seeing for it. While that belief may not be strong enough to lead you to act on it, you can steer your new investing away from the AI space, investing in businesses that are least likely to be altered by AI (food processing and leisure) and in geographies where AI is less likely to be a threat, such as the EU (perhaps because of regulation) and parts of Asia (because AI is too expensive to replace human labor in many businesses).
4. Crash out on the AI story: If your belief that AI will fail hardens into a conviction that failure is imminent, you can try to actively cash in on your story. You should, at the minimum, reduce your exposure to equities, especially in the US, by selling your holdings and putting that money into cash,. If you are more risk taking, you can sell short on the companies that have seen their pricing surge on the AI story and perhaps buy the companies that AI was meant to disrupt, flipping Leo Aschenbrenner's story. This has not been a winning strategy for many of the traders and investors who have tried it out for the last two years, and it is worth remembering the adage hat markets can stay irrational longer than you can stay solvent

I am personally going with the "market consensus' choice for the bulk of my portfolio, since I do hold five of the Mag Seven (all except Tesla and Nvidia) and four of my holdings in this group (Amazon, Alphabet, Meta and Microsoft) are heavy investors in AI cap ex, but the new money added to my portfolio in the last year or two has gone mostly into cash (short term treasuries, yielding 4%) for much of the last year, leaving my portfolios more cash-laden than usual. I have left money on the table undoubtedly by doing so, but it has helped me sleep better at night, and my advice to you is that you find a pathway in the AI jungle that helps you pass the sleep test as well. 

   There is one final piece of this puzzle that bears watching, and that is a portion of your portfolio that does not show up (yet) in your holdings. The income you will earn in your occupation, over the rest of your working life, is human capital, and to the extent that you believe that AI disruption is coming for your profession, it behooves you to direct your financial capital away from the businesses most exposed to AI disruption to balance your portfolio. I am old enough not to care much about this component, since I have fewer working years left, but if you are much younger than me, this could change your investment game.

YouTube Video

Blog Posts on AI

1. AI's Bar Mitzvah Moment: From Hope and Hype to Business Questions!

Spreadsheets

1. Implied Equity Risk Premium on September 30, 2026