Anthropic IPO Filing Draws "Buy" From Claude, Skepticism From GPT on Lease Risks

Anthropic's S-1 filing drew a cautious "yes" from its own flagship model, Claude Opus 5.5, which recommended investing "at a sensible price" while rival OpenAI's GPT 6.1 Sol declined to endorse the offering. Axios fed the 335-page filing to both models with web search disabled. Claude cited first-half 2026 revenue of $16.3 billion, up from $4.6 billion for all of 2025, a $974 million second-quarter operating profit and $131.7 billion in cash and short-term investments. It flagged blank IPO pricing, Class A holders' limited voting power and hundreds of billions in compute commitments as serious risks. GPT pointed to long-term lease obligations, concentrated executive control and IPO price uncertainty. Claude's risk read centered on non-cancelable data center and computing bills running to 2045, including a roughly $125 billion Broadcom lease, noting most revenue is pay-as-you-go so "if growth slows the bills don't." The filing names OpenAI and Google as primary competitors and discloses an October 2026 FTC civil investigative demand.