The Numbers
| Item | Figure |
|---|
| 2025 revenue | $4.6 billion |
| 2025 operating loss | $8.06 billion |
| 2024 operating loss | $2.98 billion |
| Revenue growth in 2025 | 1,088% |
| Compute and infrastructure spend, 2025 | $7.33 billion (3x 2024) |
| Cash at end of 2025 | $20.28 billion |
| Future infrastructure commitments | $518 billion |
| Valuation sought | Over $2 trillion |
| Valuation, May 2025 | $965 billion |
The offering could come after November's US midterm elections.
The Existential Risk Language Is Not the Scariest Part
The headline everywhere is that Anthropic told prospective investors increasingly autonomous AI systems could pose existential risks to humanity. It is a genuinely unusual thing to find in a prospectus, and it will be quoted for years.
It also deserves a caveat that most coverage is skipping. Risk factors in an IPO filing are written by lawyers to limit liability, and the incentive is to disclose everything conceivable. A company that has spent years publicly arguing AI poses civilisational risk would be in a strange position omitting it from the one document where failing to disclose a known risk is legally actionable. So the language is both sincere and legally obligatory, and treating it purely as either is a misreading.
The number that should actually give an investor pause is further down.
$518 Billion in Commitments Against $20 Billion in Cash
Anthropic has committed to $518 billion of future infrastructure. It ended 2025 with $20.28 billion of cash and lost $8.06 billion that year.
That commitment is roughly 112 times 2025 revenue. The cash on hand covers about two and a half years at the current loss rate, before any of the $518 billion is paid.
The bet underneath it is straightforward: revenue grew 1,088% in 2025, and if anything close to that continues, the commitments are affordable. If growth flattens even to something most companies would consider spectacular, they are not. The IPO is what funds the gap.
The Risk Factors Nobody Is Quoting
Alongside the existential-risk language, the filing discloses:
- Two customers account for nearly 25% of 2025 revenue
- No long-term contracts with major customers
- Reliance on Google and Amazon for cloud infrastructure
- Unexpected AI system behaviour
- Security vulnerabilities
- Potential for fraud and information manipulation
Read the first two together. A quarter of revenue comes from two customers who are not contractually committed to stay. For a company asking public markets for a $2 trillion valuation on the strength of its growth rate, that is a more immediate structural risk than anything in the existential-risk paragraph.
The cloud dependency is its own knot. Anthropic runs on infrastructure owned by Google and Amazon - both investors, and both operating competing model families.
Spending $2.75 to Earn $1
Revenue of $4.6 billion against an operating loss of $8.06 billion means total costs of roughly $12.7 billion. Anthropic spent about $2.75 for every dollar it earned in 2025.
That is not unusual for infrastructure-heavy companies scaling fast, and the 1,088% growth is the justification. It does mean the entire investment case rests on that ratio improving, and the filing does not promise when.
What It Means If You Build on Claude
Nothing changes about the models this week. What a filing like this gives you is visibility you did not previously have into a supplier you depend on.
Two things worth noting for planning. The customer concentration disclosure means enterprise pricing has room to move if those relationships shift. And a public company reports quarterly, which means from the IPO onward you get regular visibility into Anthropic's compute costs and margins - something no one outside the company has had until now. That is genuinely useful if you are making multi-year bets on a model provider.
FAQ
How much is Anthropic seeking in its IPO?
Over $2 trillion, against a May 2025 valuation of $965 billion.
Is Anthropic profitable?
No. It reported a $8.06 billion operating loss in 2025 on $4.6 billion of revenue, after a $2.98 billion loss in 2024.
What is the $518 billion figure?
Future infrastructure commitments disclosed in the prospectus - roughly 112 times 2025 revenue.
Did Anthropic really warn about existential risk?
Yes. The prospectus references potential existential risks to humanity from increasingly autonomous AI systems, alongside conventional risk factors.
When would the IPO happen?
Reporting suggests it could follow the November US midterm elections. No date has been confirmed.