All figures are FT Lex’s own illustrative calculations, offered as ways of thinking. None is a valuation of Anthropic by any party. Anthropic has not filed, priced, or listed.
An FT Lex column published on September 21, 2026 worked through several defensible approaches to thinking about Anthropic’s potential scale — and the outputs did not cluster. Lex’s own illustrative calculations ranged from roughly $2 trillion on a near-term revenue multiple, through approximately $4.5 trillion and $10 trillion on different estimates of addressable market share, to somewhere near $100 trillion on a scenario anchored to AI’s contribution to national output. Anthropic has not filed, priced, or listed. Every figure in that range is Lex’s illustrative arithmetic, not a forecast, and not ours.
On the revenue side, Lex reports an annualised figure of approximately $65 billion as of August — while explicitly cautioning that unofficial numbers of that kind should be handled carefully. That same column notes that around 18 months earlier, Anthropic itself had expected roughly $12 billion of 2027 revenue. The column also attributes a $320 billion annualised run-rate projection for the end of next year to some investors, while noting in its own voice that those investors are not impartial.
The column is transparent that it is laying out ways of thinking rather than issuing a price target. That transparency is exactly why the spread is analytically useful rather than merely dramatic — it maps which question each method is actually answering.
The key insight: A fifty-fold spread between the lowest and highest illustrative figure is not a statement about how uncertain anyone is regarding one quantity. It is the signature of three different quantities wearing one label — and the word “valuation” is doing the concealing work between them.
A fifty-fold spread is not uncertainty about one number. It is three different questions sharing a word.
The Structural Read
The more useful exercise is not to pick a number from the range — it is to understand why the methods produce different objects in the first place. A near-term revenue multiple prices a claim on an earnings stream that already exists and is at least partially observable. A share-of-market calculation prices a position inside a market that has been hypothesised but not yet demonstrated to exist at that scale. A national-output scenario prices a share of a surplus that would, if it arrived, accrue to an entire economy rather than to any single firm within it.
Those are three structurally different objects. The arithmetic connecting each to any price involves a distinct set of load-bearing assumptions — and those assumptions do not overlap. Calling all three outputs “valuation” is formally coherent but practically misleading, because it implies they are converging on the same answer by different routes when they are instead answering different questions entirely.
Map of AI — Founder Layer
Where a company sits in the AI stack determines which valuation method is load-bearing. A foundation model lab occupies the infrastructure layer — where value is either structural and durable, or commodity and transient. The method you choose is a bet on which of those is true. The range doesn’t tell you which bet to make. It tells you the bet exists.
The bear case has a name. Alex Karp’s term “commodity cognition” — the view that modelvalue slides toward zero and accrues instead to whoever owns the data or builds the application layer — is the mechanism that decides whether the lower or upper end of the illustrative range is more structurally coherent. If cognition commoditises, the revenue multiple is the appropriate anchor and the national-output figure is a category error. If it does not, the share-of-market methods gain traction.
What makes this question unusual is that it is already producing measurable traces in the present rather than remaining purely speculative. Whether commoditisation is or is not happening is an empirical question with observable proxies — routing behaviour, pricing dynamics, task differentiation across models. That does not make it answered. But it makes it a different kind of question from a forecast about the distant future, and a different kind of question rewards different analytical tools.
Which Method Assumes What
Near-Term Revenue Multiple (~$2T, Lex)
OBSERVABLE
Prices an existing, partially verifiable earnings stream. Load-bearing assumption: the revenue continues and expands at a rate justified by the multiple. Most grounded in current data; most sensitive to commoditisation risk.
Share-of-Market (~$4.5T–$10T, Lex)
HYPOTHESISED
Prices a position inside a market whose size is itself an assumption. Two different market-size estimates produce a more-than-twofold output spread before any firm-level variable is introduced.
National-Output Scenario (~$100T, Lex)
MACRO CLAIM
Prices a share of an economy-wide surplus. The load-bearing assumption is that the surplus accrues to a single firm rather than diffusing across the economy. That assumption does most of the work.
Three Implications
IMPLICATION 1 — THE MISS IS SYMMETRICALLY INFORMATIVE
The gap between an internal ~$12 billion 2027 forecast formed roughly 18 months ago and an unofficial ~$65 billion annualised figure reported for August 2026 — with Lex’s own caution about unofficial figures attached — cuts in both directions simultaneously. It is a reason to hold any projection loosely, including the bullish ones that the larger illustrative figures depend on, since those projections are made by people facing exactly the same forecasting difficulty. It is not evidence of unreliability; it is evidence about the upper bound on forward visibility in this market.
IMPLICATION 2 — PRICE FORMATION UNDER METHOD DISAGREEMENT
Where the available methods span orders of magnitude, the price that eventually gets set is determined by which story the marginal buyer finds convincing rather than by arithmetic narrowing the range. That is a structurally different kind of price from one anchored to an observable earnings stream — and it behaves differently, particularly when the story changes. This is not a judgment about any specific price level; it is a description of how price formation works when analytical methods diverge rather than converge.
IMPLICATION 3 — TRANSPARENCY ABOUT METHOD IS THE ANALYTICAL PRODUCT
The Lex column’s explicit framing — “ways of thinking” rather than forecasts — is considerably more useful than a false point estimate produced by selecting one method and not disclosing the others. The general principle scales: in any domain where methods disagree by orders of magnitude, naming that disagreement and mapping its structural source is the analysis. The spread is the signal, not a failure of precision to be overcome by choosing a number.
This is not investment advice, not a valuation, and not an offer or solicitation to buy or sell any security.Anthropic has not filed, has not priced and has not listed, and nothing above says the company is worth, or is valued at, any figure. Every valuation figure above is an illustrative calculation from the FT Lex column, presented there as a way of thinking rather than as a forecast, and none is this publication’s estimate. The $320 billion run-rate is attributed by Lex to investors it notes are not impartial. The annualised revenue figure of roughly $65 billion carries Lex’s own caution that unofficial numbers of that kind should be handled carefully. “Commodity cognition” is Alex Karp’s term. Nothing above says whether model commoditisation is or is not occurring, and the reference to gateway traffic concerns the kind of question it is rather than answering it — one gateway is a small and selected window. No Anthropic profit, margin, cost or headcount figure appears above, nor any other laboratory’s valuation, underwriter target, listing date or share price. Nothing is predicted.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
