Anthropic postponed its IPO from October to November after reporting third-quarter 2026 revenue of $16.8 billion, up 45% quarter on quarter, for an annualized run rate of about $67 billion. The company seeks a valuation around $2 trillion despite warnings in its own prospectus about autonomous systems and existential risks. The delay adds one more quarter of data, but it leaves the central question unchanged.
Why the IPO delay drew attention
New York City Comptroller Mark Levine, supervisor of city pension funds with roughly $7 million in Anthropic exposure, sent a letter to CEO Dario Amodei urging a pause of the offering. He cited risks also flagged in the prospectus: increasingly autonomous systems that can behave unexpectedly and create security problems. The funds could hold far more if Anthropic enters the Russell 3000 after listing, which raised the stakes of the request.
OpenAI added a second signal when CEO Sam Altman called an IPO at this moment ill-advised and said the company will not go public in 2026. The framing positions OpenAI as the more cautious actor on AI safety, while also showing that the second-most valuable AI company prefers to wait. Separately, technology shares slid last week, with a $2 trillion chip-stock selloff linked in part to concern over AI buildout costs. Together, the events turned a one-month delay into a broader test of timing.
The prospectus itself contains figures the roadshow is unlikely to emphasize. A quarter of Anthropic 2025 revenue came from two unnamed customers, creating concentration around a small number of accounts. The company also carries $518 billion in locked infrastructure commitments, a fixed obligation that sits alongside rapid growth. At the proposed valuation, investors would pay 435x 2025 revenue, according to the source analysis.
What this means for enterprise AI buyers
For companies that buy or deploy AI, the dispute centers on the pricing premium for closed models over open-weight alternatives. Enterprise adoption of open-weight AI rose from 23% to 67% in 2026, with reported cost savings of 70%-90%. DeepSeek R1 is cited at 7 cents per million input tokens, while independent analysis puts comparable Anthropic input and output pricing roughly 18x and 23x higher. DeepSeek weights can be downloaded and self-hosted without per-token billing.
The structural risk is that the capability gap closes within six months of a frontier release, which would make a durable premium harder to defend. Gartner forecasts at most $3.6 trillion in revenue attributable to AI data centers through 2030, against $8.5 trillion needed to cover investment, financing costs and a 12% return. New Constructs calculated that a $2 trillion valuation would require an immediate 10% after-tax operating margin for a decade and $358 billion in annual after-tax operating profit by 2035.
The marker to watch is whether November brings an explanation of that math rather than another quarter of growth. Worldwide AI spending is forecast at $2.59 trillion in 2026, up 47% year over year, yet a July 2025 MIT NANDA report found 95% of examined organizations getting no return from generative AI, and KPMG reported only 7% of leaders with established ROI. If deployment, governance and measurable outcomes do not improve, strong model revenue alone will not settle the valuation debate.
