Is Anthropic Worth $1 Trillion—or $1.3 Trillion? Inside AI’s Biggest IPO Test

Anthropic is preparing to sell 825 million shares to the public at between $70 and $78 apiece.

That is the simple part.

The difficult part is deciding what those shares make Anthropic worth.

Count the shares expected to be outstanding after the initial public offering and a concurrent private placement, and the Claude maker would be valued at approximately $916 billion to $1.02 trillion. Include options, restricted stock units and other securities that could eventually become shares, and the fully diluted figure rises to roughly $1.3 trillion at the midpoint of the range and as much as $1.4 trillion at the top.

Both figures describe Anthropic. They do not describe the same thing.

That valuation gap is the first test facing investors in what could become the largest conventional IPO in history. The second is harder: can a company founded in 2021 grow quickly enough to justify a price normally reserved for the most powerful businesses on earth while paying for the most ambitious computing build-out ever disclosed by an artificial-intelligence laboratory?

Anthropic is not arriving with an ordinary technology prospectus. It is offering Wall Street a stake in one of the fastest-growing companies ever recorded, attached to approximately $900.9 billion of computing and infrastructure commitments and a governance system designed to ensure that ordinary shareholders never control it.

The IPO is consequently more than a fundraising event. It is the first full public-market examination of frontier AI economics.

Why are there two Anthropic valuations?

Anthropic’s amended prospectus, filed with the US Securities and Exchange Commission on November 10, says the company expects approximately 13.09 billion Class A, Class N and special-class shares to be outstanding after the offering and private placement.

Multiply that number by the proposed share price and the result is an equity value ranging from about $916 billion at $70 a share to approximately $1.02 trillion at $78.

The larger figures reported elsewhere use a fully diluted share count. That calculation adds securities such as outstanding employee options, unvested restricted stock units and shares that could be issued when Amazon-linked notes convert. At the $74 midpoint, that broader approach produces a value of roughly $1.3 trillion. At $78, it can exceed $1.4 trillion, depending on precisely which potential shares are included.

The distinction is not cosmetic.

The outstanding-share figure describes the company immediately after the transactions. The fully diluted figure attempts to show what existing ownership could look like if options, awards and convertible instruments eventually become stock. Investors use both measures, but a headline that simply says Anthropic is seeking $1.3 trillion without explaining the denominator gives the appearance of certainty where none exists.

There is another complication. Options are not free shares. Employees must generally pay their exercise prices, while some restricted awards may never vest. A fully diluted calculation is useful for understanding possible dilution, but it is not identical to the market capitalisation that will appear on a trading screen when Anthropic lists.

For investors, therefore, the first argument over Anthropic is not whether Claude is worth $1 trillion. It is which Anthropic they are valuing.

How large is the IPO?

Anthropic is selling approximately 801.35 million of the 825 million shares in the public offer. Existing shareholders will sell only 23.65 million.

At $74 a share, the public offering would be worth approximately $61.05 billion before fees. Anthropic itself would receive about $59.3 billion from the shares it sells, while the selling shareholders would collect approximately $1.75 billion.

The company has also arranged a concurrent $26 billion private placement of non-voting Class N shares with strategic investors including Google, Microsoft, Nvidia, AMD, Broadcom, Micron and TSMC. Anthropic estimates that its combined net proceeds from the IPO and private placement would be approximately $84.6 billion at the midpoint, rising to $93.7 billion if underwriters fully exercise their option to buy additional shares.

For perspective, Saudi Aramco’s 2019 flotation raised $29.4 billion after its over-allotment option was exercised. Anthropic’s public offer alone could raise more than twice that amount at the top of its range.

This is also overwhelmingly a company fundraising exercise rather than a mass exit for early investors. Approximately 97 per cent of the shares in the base public offering are being sold by Anthropic. The company is raising historic capital because it has assumed historic obligations.

Is Anthropic growing quickly enough to justify it?

The growth belongs in a different corporate era.

Anthropic generated revenue of $41 million in 2023, $386 million in 2024 and $4.59 billion in 2025. In the first nine months of 2026 alone, revenue reached $33.24 billion, compared with $2.55 billion during the corresponding period a year earlier.

That is growth of 1,206 per cent.

Consumption-based services accounted for $26.3 billion, or 79 per cent, of the nine-month total. Subscriptions contributed another $6.94 billion. By October 31, Anthropic said its annualised run-rate revenue had reached approximately $81 billion.

Run-rate revenue should be treated carefully. Anthropic calculates it by annualising a recent 28-day period of consumption fees and one month of subscription fees. It is neither audited annual revenue nor a forecast. A usage surge during the measurement period can make the number look stronger than the revenue ultimately recognised over a full year.

The company itself warns that early-adoption growth may not continue and that consumption can fall if customers reduce usage, shift to cheaper models or adopt competing products.

Even so, the underlying expansion is extraordinary. Anthropic’s remaining performance obligations, representing deferred revenue and contracted commitments expected to become revenue, stood at $13.8 billion at September-end. It expects to recognise about three-quarters of that amount within 12 months.

The valuation is not being placed on a laboratory waiting for a business model. Claude has become a large business at remarkable speed. The debate is whether remarkable can remain normal.

Did Anthropic really lose $104 billion?

Yes, according to the accounting statement. But the number does not mean Anthropic spent $104 billion more cash than it earned while running Claude.

The company reported a net loss of $104.04 billion for the first nine months of 2026. Approximately $94.3 billion of that arose from changes in the fair value of convertible notes and derivative liabilities as Anthropic’s enterprise valuation increased.

Paradoxically, making Anthropic more valuable made its accounting loss larger because instruments that could convert into shares became more expensive liabilities on paper.

The operating numbers provide a clearer view of the business. Anthropic’s GAAP operating loss narrowed from $5.22 billion in the first nine months of 2025 to $1.26 billion in the corresponding 2026 period. After excluding stock-based compensation, non-cash charitable contributions and a litigation settlement, the company reported adjusted operating income of $4.1 billion.

Anthropic also generated $6.57 billion in operating cash flow during the first nine months of 2026, compared with cash use of $2.02 billion a year earlier.

The $104 billion loss should therefore neither be sensationalised nor erased. It is primarily a valuation-driven accounting charge, but the instruments behind it still matter because their conversion can dilute shareholders. A non-cash loss may not empty the bank account. It can still change who owns the bank.

Why does the $900.9 billion compute bill matter?

Because Anthropic has contracted today for an enormous portion of the computing power it expects to need tomorrow.

The company says purchase commitments, leases, short-term arrangements and leases that have not yet begun represent an aggregate committed spend of approximately $900.9 billion. Substantially all of it relates to compute and infrastructure.

Some agreements contain minimum-purchase or take-or-pay clauses. Anthropic can, therefore, be required to pay for capacity even if customer demand does not grow as expected or the company cannot use all the infrastructure efficiently.

Approximately $313.8 billion of non-cancellable purchase commitments with terms longer than one year appeared in Anthropic’s contractual-obligations table at September-end. The wider $900.9 billion figure also includes newer agreements, short-term commitments and leases outside that table.

The number converts Anthropic’s growth forecast into a financial obligation. If demand continues exploding, committed capacity can become a competitive advantage. Anthropic will possess the chips and cloud access required to serve customers while slower rivals wait for infrastructure.

If demand weakens, model prices collapse or a competitor makes AI dramatically more efficient, the same capacity becomes a burden. Compute is Anthropic’s factory. It has ordered much of the factory before knowing exactly how large the market will be.

The IPO proceeds provide a large cushion, but not enough to pay the commitments by themselves. The business must continue generating revenue and cash, while its infrastructure partners must finance and deliver the capacity on schedule.

That is the economic wager inside the listing: Claude must grow into the bill already written for it.

Are Anthropic’s investors also its suppliers?

Several of them are, and that is both strategic strength and concentration risk.

Amazon and Google have provided capital, cloud infrastructure and distribution. In 2025, 47 per cent of Anthropic’s revenue reached customers through the cloud marketplaces operated by Amazon and Google, according to a Reuters analysis of the company’s prospectus. During the first nine months of 2026, the proportion was approximately 44 per cent.

The concurrent placement deepens the network. Google, Microsoft and leading chipmakers are committing another $26 billion while participating in an ecosystem that could benefit from Anthropic’s infrastructure spending.

This does not make the demand circular by definition. Enterprise customers are paying to use Claude, and Anthropic is buying the computing capacity required to serve them. But the commercial relationships are unusually intertwined. Companies supplying the chips, cloud services and distribution are also helping finance the customer that will purchase those services.

Public investors must decide whether that structure represents an aligned industrial alliance or a chain in which each participant’s confidence helps support the next participant’s revenue.

How much of Anthropic’s revenue may not return?

The prospectus contains a striking disclosure: Anthropic estimates that customers later found to be violating its usage policies or commercial terms generated approximately 5 to 7 per cent of revenue during the first nine months of 2026.

Applied to $33.24 billion, that range represents approximately $1.66 billion to $2.33 billion.

Anthropic recognises the revenue while providing the service, then restricts or removes customers when violations are detected. The revenue is real in an accounting sense, but may not recur after enforcement.

The disclosure illustrates the strange economics of frontier AI. Rapid adoption can include customers whose activity the company does not want. Safety enforcement can therefore improve the platform while reducing revenue.

What control will IPO investors receive?

Very little.

Anthropic’s publicly sold Class A shares carry one vote each. But a single Class F share is entitled to a fixed 50.1 per cent of the total voting power on matters submitted to shareholders generally.

Its power does not decline when Anthropic issues more Class A or Class N stock. The Class F vote automatically adjusts to preserve majority control.

The structure is linked to the company’s founders and public-benefit governance. Anthropic warns that those exercising control may support decisions consistent with its mission or the interests of people affected by its conduct even when those decisions conflict with the financial interests of ordinary shareholders.

This goes beyond the familiar dual-class arrangement used by technology founders. Public investors cannot simply wait for future share issuance to dilute the controlling block. Anthropic has designed control to resist dilution.

Shareholders are being asked to provide capital, price the risk and absorb market volatility without acquiring a plausible route to control the company they own economically.

That may appeal to investors who believe frontier AI requires insulation from quarterly market pressure. It may alarm those who believe a trillion-dollar public company should ultimately answer to its capital providers.

What does the valuation demand?

At the $74 midpoint, Anthropic’s post-transaction outstanding-share valuation of approximately $968 billion would equal about 29 times its revenue for the first nine months of 2026 and roughly 12 times its stated October run rate.

Those comparisons are imperfect. Nine-month revenue is not annual revenue, and run-rate revenue is not an audited forecast. They nevertheless show how much future success is already embedded in the offer price.

Anthropic must defend its position against OpenAI, Google, Meta and new model developers. It must keep improving Claude without allowing research and inference costs to outrun revenue. It must preserve enterprise trust, deliver sufficient safety and turn contracted capacity into profitable usage.

The bull case is powerful. AI becomes a new computing layer, Claude becomes one of its principal enterprise interfaces, inference demand multiplies and the infrastructure commitments look prescient.

The bear case does not require AI to fail. It merely requires competition to reduce prices, customers to spread workloads across several models, efficiency improvements to reduce compute demand or revenue growth to slow before the obligations do.

At approximately $1 trillion, Anthropic does not need to prove that artificial intelligence is useful. It needs to prove that usefulness can produce durable economics on a previously unseen scale.

Why does this IPO matter beyond Anthropic?

Because the market will be pricing much more than Claude.

A successful listing would give other AI companies a public valuation benchmark, help infrastructure developers justify new capacity and demonstrate that equity markets will fund the next stage of the computing race. It would also strengthen the case for OpenAI and other frontier laboratories eventually seeking public capital.

A weak debut would travel through the same network. It could force investors to re-examine data-centre developers, chipmakers, cloud providers and private AI valuations built on assumptions of sustained demand.

Anthropic’s prospectus captures the entire AI boom in one company. Revenue is expanding at a speed that makes conventional comparisons look obsolete. Capital requirements are expanding almost as quickly. Suppliers are investors. Investors are commercial partners. Public shareholders are invited to participate in the economics without controlling the mission.

Is Anthropic worth $1 trillion or $1.3 trillion? The arithmetic can produce either answer. The IPO must determine whether the business can eventually produce one.