Anthropic's IPO filing and its risk disclosures
Anthropic is heading toward an IPO at a $2 trillion valuation while losing far more money than it brings in — a bad setup for anyone who buys in, and possibly for the wider economy.
- The filing leaves out some costs, including what it spends training AI models.
- Revenue grew sharply last year to under $5 billion, but the company lost about $8 billion running the business.
- Roughly a quarter of that revenue came from just two customers, and most big clients have no long-term contracts.
- The filing devotes more space to warnings that the technology could be dangerous than to the financials — a way to make the product sound world-changing enough to justify $518 billion in promised cloud and infrastructure spending.
- Google, Meta and Nvidia all hold stakes, and banks have lent heavily to AI firms, so a collapse would spread fast.
Outlook: The IPO may slip past the midterms, and if AI valuations break, the damage runs through the big tech holders and into the banks.