Anthropic Says Its Market Is $40 Trillion: What Is Claimed and What Is Verified

Anthropic now reportedly says its total addressable market is more than $40 trillion a year, a figure built from the economic value of knowledge work worldwide in 2025. That is the claim, and it is a step up from the roughly $30 trillion the Wall Street Journal reported in August. Where the number appears is less settled: some coverage ties it to the public S-1 filed on October 26, while other summaries say the filing itself contains no numeric TAM.

This post lays out what each source says, what is confirmed and what is not, and how to read a number this large if you build on Claude. For the filing's financials, see our S-1 analysis.

TL;DR: claimed vs verified

What exactly is being claimed?

PYMNTS reported on October 26 that Anthropic estimates its total addressable market at more than $40 trillion. It describes the basis as the economic value generated by knowledge workers globally in 2025, which it calls roughly one-third of the world economy and larger than the US economy, citing IMF data. It also stresses that the number represents economic activity Anthropic hopes to serve, not a forecast of AI spending.

Techmeme's summary of the Wall Street Journal says the S-1 implies a $40 trillion-plus annual TAM based on knowledge-work value in 2025, topping SpaceX's $28.5 trillion estimate without naming it. That wording matters: "implies" and "without explicitly naming it" suggest the filing describes the opportunity qualitatively and the number is derived by reporters or analysts.

Where does the number come from? The $30 trillion precedent

The first version of the story was different. In August, Fortune relayed the WSJ report that Anthropic would tell investors its TAM exceeded $30 trillion, roughly the size of US GDP. The method then was described as the full scope of work AI models could complete, rather than a category like enterprise software. Alex Brunicki of Backed VC told Fortune that "the TAM for those products is essentially the labor market for that work output."

So the logic has been consistent: start from what humans are paid to do with their minds, and treat the AI product as a substitute for that labor. What changed is the headline number, from $30 trillion plus to $40 trillion plus. The sources we could read do not explain the difference, and we do not know whether it reflects a new base year, a wider definition of knowledge work, or different reporting. We are not going to guess.

Is it actually in the filing?

This is the part of the story the coverage disagrees on, and it is worth being careful about. Our search found three different framings:

• PYMNTS attributes the $40 trillion figure to Anthropic and links the S-1 for revenue details, but its TAM paragraph carries no filing citation.
• Techmeme's WSJ summary says the S-1 "implies" the figure.
• Other summaries of the filing say the preliminary S-1 offers no numerical TAM, and that the number was pitched to investors outside the document.

All three can be partly true: a filing can describe knowledge work as the target market, while the dollar figure lives in a roadshow deck or in journalists' arithmetic. We could not retrieve the WSJ article or the filing text in this run, so we cannot settle it. The primary document is the S-1 on the SEC's EDGAR system; searching its "Market opportunity" section for the number is the fastest check.

How a TAM works, and why this one is unusual

A TAM is the annual revenue a company would earn if it captured 100 percent of its market. Normally it is built bottom-up: customers times price. Anthropic's version is top-down and labor-based: total wages and value of knowledge work, treated as spendable.

There is precedent. Fortune notes Uber cited a $6 trillion TAM in 2019, based on the value of all personal-car and public-transport miles worldwide, and some analysts called that aggressive marketing rather than serious math. SpaceX's $28.5 trillion estimate is the other comparison point in this cycle.

Three caveats apply to the Anthropic number:

1. Value is not spend. Knowledge workers create value worth far more than their pay, and no one is going to hand an AI vendor the whole value of a lawyer's or engineer's output. Price per task falls as models get cheaper.
2. Competition caps share. Brunicki warned that cheaper, industry-specific open-source models could pressure Anthropic's share. For the latest revenue claims, see our Q3 revenue claimed-vs-verified post.
3. Substitution is not total. Many knowledge-work tasks need people for accountability, regulation or trust. A TAM that assumes full replacement is a ceiling by construction.

What the filing's real numbers say

A TAM is easier to judge next to actuals. PYMNTS, summarising the S-1, reports first-half 2026 revenue of $16.3 billion against $1.16 billion a year earlier, about 14 times higher. It lists an operating loss of $693 million (against $2.1 billion), cost of revenue of $2.38 billion, consumption-based fees of $13.06 billion (about 80 percent of revenue) and subscription revenue of $3.22 billion. Computing and infrastructure expenses were $8.95 billion, 55 percent of revenue.

It also reports that customers spending $100,000 or more grew from more than 950 to more than 11,900 as of September, and that the five largest customers made up about 25 percent of revenue over the first nine months of 2026, with none above 10 percent. Treat these as PYMNTS's reading of the filing, and confirm any figure you plan to quote against the document. Our earlier coverage compares them with the Wall Street Journal's draft-prospectus numbers and the Q3 revenue claims.

Against $16.3 billion in half-year revenue, even a $1 trillion slice of a $40 trillion market is a growth story of more than 100 times. That is the point of the number: it tells investors there is no visible ceiling. It does not tell them how fast revenue gets there.

How skeptics and supporters read it

Fortune quotes Fred Hickey, editor of The High-Tech Strategist, who pointed out that US GDP is about $32.5 trillion and called such claims "nonsense" that lets Wall Street and Silicon Valley extract money before a bubble collapses. Brunicki took a middle view: professional investors will likely treat the TAM as a mission statement and focus on cash-flow models and near-term targets, while retail investors may take it at face value.

Both reactions are opinions, not findings. The factual points are narrower: the number is large, it is derived from labor value, and it is not a forecast. Our post on Anthropic's IPO governance covers a separate question that matters more to long-term holders: who controls the company after listing. The company's IPO letter on ads shows how monetisation pressure could meet its stated pledges.

What this means for what you build or pay

A TAM has no direct effect on your API bill, but it hints at direction.

• Pricing pressure points both ways. A pitch built on replacing labor favors high-value, agentic and enterprise products over cheap chat. Expect more emphasis on consumption-based enterprise usage, which is already about 80 percent of reported revenue.
• Capacity and limits. With compute at 55 percent of revenue per PYMNTS, usage limits and plan perks remain the lever Anthropic has. Watch plan changes after listing; see the public benefit fund coverage for another commitment the company will be held to.
• Vendor concentration. If you plan around one provider, a pitch this broad is a reason to keep a second model available, not a reason to commit more.

How to check it yourself

1. Open the S-1 on SEC EDGAR and search for "addressable", "TAM" and "knowledge worker". Note whether a dollar figure appears and in which section.
2. Compare the base year and definition with the WSJ's August version. If the definitions differ, the two numbers are not comparable.
3. Read the risk-factor section rather than the market-opportunity section; it is where the company has to be specific about competition, margins and compute commitments.
4. Treat any single-outlet figure as unconfirmed until two readable sources agree, and note which ones are summaries of paywalled reports.

What we do not know

We could not read the WSJ article behind the Techmeme summary, and PYMNTS does not say where in the filing the figure sits. We do not know why the number moved from $30 trillion to $40 trillion, whether the filing includes a methodology footnote, or whether Anthropic will update it before pricing. We will add an update section here if the filing text or the company clarifies.

Bottom line

The verified part is modest: reputable coverage says Anthropic is framing its opportunity as the value of global knowledge work, at a size above $40 trillion, and that this is bigger than SpaceX's $28.5 trillion claim. What remains unverified is the exact wording, its location in the S-1, and any method behind the jump from $30 trillion. As a reader, treat it the way professional investors reportedly do: a mission statement about ambition, to be judged against revenue, margins and compute costs.

Related reading

• Anthropic S-1: what the filing shows, the bubble question and the future of work
• Anthropic Q3 revenue of $16.8B: claimed vs verified
• WSJ draft prospectus: $16.3B H1 revenue and $450B compute
• Anthropic IPO governance: can it resist Wall Street?
• Anthropic's IPO letter and the no-ads pledge
• Anthropic public benefit fund: 5 percent of shares
• SpaceX Q3 2026 earnings: AI revenue, what is confirmed

Figures and reports are accurate as of October 28, 2026, and may change as the filing is amended.