Not sure what to put in the sliders? Answer in plain terms: prices, customers, and how much the big risks worry you. Each question starts with a Fact (or Our view for rough benchmarks); your answer is your Assumption. The answer choices are ours, and skipped questions use the middle one.
0 of 9 answered
Moonshot explorer Our view
Markets that don't exist yet but could one day pay Anthropic. Each one is valued like an option: the chance it happens, times the profit if it does, discounted back to today. The market sizes, shares, chances and dates are our guesses: move them, and switch on the ones you believe in to add them to your model.
AI that replaces outsourced software and IT work
Claude writing, testing and running most business software instead of IT service firms.
Fact Companies spend well over $1 trillion a year on IT services worldwide.
What would have to be true:
Agents that work for days without mistakes
Companies trusting AI with their systems
Prices that stay above the cost of compute
If it happens: Anthropic earns $120B a year from 2034, worth $569B then at 20x profit. Weighted by a 25% chance and discounted to today: $51B.
AI scientists
AI running experiments and research for labs and companies.
Fact The world spends about $2.5 trillion a year on research and development.
What would have to be true:
Models that come up with new, correct ideas
Robot labs to test them
A way to get paid a share of the results
If it happens: Anthropic earns $75B a year from 2035, worth $356B then at 20x profit. Weighted by a 15% chance and discounted to today: $17B.
AI that cures diseases and slows ageing
Royalties or partnerships on drugs and treatments found with Claude.
Fact The world spends about $1.6 trillion a year on medicines.
What would have to be true:
AI-found drugs passing clinical trials
Regulators accepting AI-led research
Deals that pay Anthropic, not only its customers
If it happens: Anthropic earns $20B a year from 2040, worth $126B then at 20x profit. Weighted by a 8% chance and discounted to today: $1.7B.
AI doctors, lawyers and accountants
Claude doing professional work directly for people and firms.
Fact Professional licensing rules currently require humans to sign off on most of this work.
What would have to be true:
Licensing rules that allow AI to practise
Insurance and liability solved
Accuracy people trust with their health and money
If it happens: Anthropic earns $50B a year from 2034, worth $237B then at 20x profit. Weighted by a 20% chance and discounted to today: $17B.
Brains for robots
Claude as the thinking layer inside robots made by others.
Fact Robot makers today use a mix of in-house and partner AI models.
What would have to be true:
Fast, cheap models that run on the robot
Robot sales at scale
Robot makers choosing partners over building their own
If it happens: Anthropic earns $20B a year from 2038, worth $95B then at 20x profit. Weighted by a 10% chance and discounted to today: $2.0B.
Switch on the moonshots you believe in to add their value to your model above.
Value today = chance x (market x share) x margin x (1 - tax) x 20x profit, discounted from the start year at your required return and shrunk for dilution. Margins are our view: AI that replaces outsourced software and IT work 30%, AI scientists 30%, AI that cures diseases and slows ageing 40%, AI doctors, lawyers and accountants 30%, Brains for robots 30%.
Maker of Claude. Filed its public IPO prospectus (S-1) on 26 Oct 2026. Below is a valuation lab built on its IPO prospectus: pick a scenario, move the sliders, and the Stonkometer shows whether today's valuation looks like a bubble or a bargain by your numbers. Assumption Our middle case puts Anthropic at $1.09T against a last round of $965B. It is a starting point, not a forecast; yours will differ, and that is the point.
Fact from filings or reports, sourced below · Assumption a Stonkometer default you can change · Our view opinion
Big claims, checked
Fact Each claim is quoted with its source. The maths shows what would have to be true; the slider defaults are assumptions.
Anthropic S-1 (Oct 2026)
“"Powerful AI" matching top human experts across most fields will become widely available within a few years.”
No revenue forecast comes with it. Use the "AI agents" bet below to put your own number on it.
Reported (Bloomberg, Aug 2026)
“Run-rate revenue passed $65 billion at the end of July 2026.”
H1 2026 actual revenue was $16.3B, about $33B a year. Run-rate measures the last 28 days only.
Reported IPO talk (Sep 2026)
“A listing valuation above $2 trillion.”
What profits would $2T need?
What the filings say Fact
Revenue: $4.6B in 2025, $16.3B in the first half of 2026 (13x a year earlier). Consumption-based (API, platform) is about 80%, subscriptions 20%.
H1 2026 operating loss was $0.7B, versus $8.1B for all of 2025. Compute and infrastructure was the largest cost line ($8.9B in H1 2026).
The H1 2026 net loss includes a $72.4B non-cash charge for the change in fair value of convertible notes and SAFEs, which convert at the IPO.
Compute comes from AWS Trainium, Google TPUs, NVIDIA and AMD GPUs, plus a SpaceX/xAI deal at $1.25B a month. Long-term commitments run to 2045.
A public benefit corporation: the Long-Term Benefit Trust and a founders' entity keep voting control after the IPO.
Customers paying over $100k a year produce about three quarters of revenue.
Key risks Fact
Must keep building leading models; a weaker model can lose customers fast.
Needs ever more compute, power and chips, from partners who are also competitors and investors.
Run-rate revenue annualises a 28-day window and may overstate future revenue.
Concentration: a limited number of customers and channels drive much of the revenue.
AI harms, misuse or models acting in unintended ways could bring lawsuits and regulation.
Class A shareholders get little control; directors must balance profit against the public benefit purpose.
The company's own risk factors, summarised in plain English. Read the originals below.
The business lines in the model
Consumption (API, platform, Claude Code usage)$26B h1 2026 x 2, $52B on the latest basis
$13.1B in H1 2026 vs $1.0B in H1 2025.
Subscriptions (Claude plans)$6B h1 2026 x 2, $13B on the latest basis
$3.2B in H1 2026 vs $0.2B in H1 2025.
Bet: AI agents doing whole jobs
The S-1 expects "Powerful AI" with expert-level skills within a few years.
H1 2026 x 2: Six months to 30 Jun 2026 revenue x 2 (S-1). Reported run-rate ($65B, end of July 2026): Bloomberg, Aug 2026. Split 80/20 like H1 2026 (our estimate). Run-rate = last 28 days of usage x 13 + subscriptions x 12, so it swings.. Cash $112.4B + short-term investments $19.3B on 30 Jun 2026 (S-1). Convertible notes convert at the IPO..
Questions people ask Fact
Can I buy Anthropic stock?
Not on the stock market yet. Anthropic filed its IPO prospectus (S-1) on 26 October 2026; shares trade publicly only once the IPO is complete.
When is the Anthropic IPO?
The S-1 was filed publicly on 26 October 2026. The listing date and price have not been announced.
What is Anthropic worth?
Its last private round in May 2026 valued it at $965 billion. IPO talk of more than $2 trillion is unconfirmed. Test both against your own numbers in the lab.
Not investment advice. Stonkometer is an educational calculator, not investment advice. Valuations depend entirely on the assumptions you choose. We are not a broker or adviser, and nothing here is a recommendation to buy or sell any security. Do your own research and consider a licensed adviser. Full disclaimer