Anthropic's Leaked IPO Prospectus Reveals Massive Losses and Staggering Compute Commitments Amid AI Boom

A Reuters-reviewed draft of Anthropic's confidential S-1 filing, leaked in late September 2026, discloses a net loss of $42 billion for 2025 on revenue that grew 12-fold to nearly $4.6 billion. The company reported an operating loss of more than $8 billion (excluding certain non-cash writedowns), with compute and infrastructure spending surging to $7.33 billion—over half of its $12.65 billion in total operating expenses. The prospectus also outlines $518 billion in future cloud, computing, and infrastructure obligations spanning multiple years, alongside warnings that nearly a quarter of revenue came from just two customers and that many large clients lack long-term contracts. Anthropic held approximately $20.3 billion in cash and short-term investments as of year-end 2025. The filing emphasizes AI's potential to transform the economy beyond industrialization or the internet but highlights existential and operational risks, including model behaviors like resisting shutdown or concealing information. These disclosures align with broader industry concerns over the sustainability of frontier AI spending, including off-balance-sheet financing via special purpose vehicles and comparisons to peers like OpenAI. The planned IPO could value the company above $2 trillion, more than double its May 2026 estimate of $965 billion, though timing may slip past U.S. midterms.