A $65bn Series H lifts the Claude maker to a $965bn valuation on a $47bn revenue run-rate — leapfrogging a $852bn OpenAI as both labs file to go public within a week of each other.
Anthropic is, for the first time, the most valuable private company in the world. The maker of the Claude family of models confirmed a $65bn Series H — led by Altimeter, Dragoneer, Greenoaks and Sequoia — at a $965bn post-money valuation, nearly tripling its February mark and edging past a recently re-rated OpenAI.
The round lands alongside Claude Opus 4.8, which took the top seat on the Artificial Analysis Intelligence Index, and a reported $47bn revenue run-rate driven largely by Claude Code. On the numbers that implies roughly a 21× multiple — rich by software's historic standards, restrained by the standards of this cycle.
OpenAI, valued near $852bn after its own record $122bn raise from Amazon, Nvidia and SoftBank, filed its draft prospectus a week after its rival. Between them the two labs are seeking, at listing, something close to two trillion dollars of public capital.
“The market is no longer pricing software. It is pricing a share of the world's future compute.”
The sceptics are not silent. Circular financing — hyperscalers funding the very labs that buy their chips and cloud — draws comparison with earlier manias, and run-rate revenue is not the same as profit. Yet the capital keeps arriving: foundational-AI funding in the first quarter alone exceeded all of 2025.
For readers building on these models, the strategic point is quieter than the headline. Capability is converging and price is falling; advantage is migrating from the model to the system around it — the data, the retrieval, the agents and the judgement.
Elon Musk's two ventures merge to build orbital data centres and route public-market exposure to Grok through a SpaceX listing said to target as much as $1.5tn. Hyperscaler AI capex now runs past $725bn a year.
With MMLU long exhausted, evaluation has moved to GPQA Diamond, SWE-bench Verified and Humanity's Last Exam — where no single model wins every column.