Anthropic makes Claude, the AI models used by businesses and developers. Revenue has grown more than tenfold in a year and the company is preparing what could be one of the largest IPOs ever. This report asks what it is worth, and what price would make sense to buy at.
Company values in billions: the scenario model’s range, the probability-weighted value, the last private round and the valuations discussed for the IPO. The copper line marks the base case.
Written on October 21, 2026, ahead of its expected IPO in mid-November. Figures are from company statements and news reports, as linked; they are not audited. Not investment advice. This company is private, so there is no share price; values are for the whole company. The author holds no shares. This report was drafted with help from Claude, an AI model made by Anthropic.
Anthropic was founded in 2021 by Dario and Daniela Amodei and other former OpenAI researchers. It sells access to its Claude models through its own apps and API and through Amazon Web Services and Google Cloud. Most revenue comes from businesses and developers rather than consumers, led by coding: Claude Code has become one of its largest products.
It is a public benefit corporation, and an independent Long-Term Benefit Trust holds a special class of shares with the right to elect directors, a structure meant to keep safety ahead of short-term profit.
Revenue has grown faster than at almost any company of its size. Third-quarter revenue of $16.8 billion, reported by Bloomberg and the Financial Times from investor presentations, was up 45% from the second quarter, but some investors noted that monthly growth is slowing. OpenAI has argued that Anthropic counts sales through cloud partners at their gross value, which flatters the comparison; the public S-1 will settle how revenue is reported.
| Figure | Status | |
|---|---|---|
| Operating loss, 2025 | About $8B | Leaked draft S-1 |
| Net loss, 2025 | About $42B, mostly non-cash | Leaked draft S-1 |
| Operating profit, Q2 2026 | About $0.6B, adjusted | Reported |
| Compute commitments | About $518B, mostly non-cancellable | Reported |
| Cash-flow breakeven | 2028 | Reported target |
Anthropic reached an adjusted operating profit in the second quarter, rare for an AI lab, but it has committed to more than half a trillion dollars of computing with Amazon, Google, SpaceX and others. Those contracts are the main risk if growth slows. None of these figures is audited yet; the public S-1 will be the first full set.
The last private round was a $65 billion Series H at a $965 billion valuation in May 2026. Amazon owns an estimated 15–21% (non-voting) and Google about 14%. A confidential S-1 was submitted on June 1; the listing is expected on Nasdaq in mid-November, led by Morgan Stanley, Goldman Sachs, JPMorgan and Citi. Investors have discussed valuations from about $1.5 trillion to $2 trillion.
This report values the company on 2030 earning power: revenue, operating margin and a multiple of after-tax profit, discounted back at 15% a year to reflect the risk of a young business. It does not deduct the new shares that may be issued to fund the buildout.
| Bear | Base | Bull | |
|---|---|---|---|
| Revenue in 2030 | $120B | $250B | $450B |
| Operating margin | 15% | 30% | 40% |
| Exit multiple | 20× | 25× | 25× |
| Company value today | $163B | $847B | $2.03T |
| Probability | 25% | 50% | 25% |
Weighted 25/50/25: about $970 billion, almost exactly the May round.
The values above are for the whole company in 2030. A private company that is still burning cash pays for it by selling new shares, and the money those buyers put in is theirs, not today’s owners’. So the value of a stake bought today is the whole-company value less the present value of the capital still to be raised. The table shows that for several amounts raised evenly over the next four years, discounted at the model’s 15%. “Share sold” is how much of the company those raises would buy at the last valuation of $965B; at a lower price, the share sold would be larger.
| Capital raised before 2030 | Share sold at the last valuation | Value to today’s owners (base case) | Against the last valuation |
|---|---|---|---|
| None | 0% | $847B | −12% |
| $50B | 5% | $809B | −16% |
| $100B | 9% | $770B | −20% |
| $200B | 17% | $694B | −28% |
Every figure in this valuation comes from one model script, rerun and checked against this page each time the site is built. The XLSX model rebuilds it in live formulas: change any blue input and every year’s cash flow, the bridge to the value, the scenarios and the sensitivity table recalculate.
| To justify the price, you need | Against |
|---|---|
| $965B (May round): about $285B of revenue in 2030 at a 30% margin | About $67B annualised today |
| $1.5T: about $445B | |
| $2T: about $590B | Close to the bull case |
| Risk | Why it matters |
|---|---|
| Competition | OpenAI, Google and open models compete on price; model leads are short. |
| Commitments | About $518B of compute contracts must be paid even if growth slows. |
| Concentration | Amazon and Google channels were about half of 2025 revenue. |
| Governance | The Benefit Trust and founder voting control limit shareholders’ say. |
| IPO pricing | A $1.5–2T listing would leave little room for error. |
Upside: AI agents and coding tools growing into a large share of software spending, with margins rising as computing gets cheaper.
The base case supports roughly the May private valuation, about $1 trillion. A listing at $1.5–2 trillion would price in close to the bull case on its first day.
What would change the value: The prospectus’s margins, compute commitments and the IPO price are the numbers that decide whether the shares are worth buying.
Private companies are not rated: there is no price at which to buy them. The values above are for the whole company, before the dilution that future funding rounds would bring. How ratings work
Buy the business at a valuation the base case supports, not at the IPO’s first price.
| What | Figures | Source |
|---|---|---|
| Discount rate inputs | 10-year Treasury 5.15%, equity risk premium 4.5% | Fixed for every report so they can be compared; How ratings work |
| Valuation | Every value, case and sensitivity | This site’s model: XLSX, CSV |
| Checked by hand | Not yet | Reports are added to the audit log as they are revised |
| Input | Bear (10th percentile) | Base (median) | Bull (90th percentile) |
|---|---|---|---|
| Revenue, target year | $120B | $250B | $450B |
| Margin, target year | 15.0% | 30.0% | 40.0% |
| Exit multiple | 20× | 25× | 25× |
| Simulation result | |
|---|---|
| Expected value (average of 4,000 runs) | $937B |
| Likely range, 10th to 90th percentile | $202B to $1,856B |
| Median | $791B |
Each input is drawn from a split normal: its bear and bull values are the 10th and 90th percentiles, so the two sides can differ. Business inputs share a common factor (correlation 0.6); the discount rate is the same in every case, so it is not drawn. Inputs not listed are held at their base value. How the simulation works.