Analysis
Anthropic's revenue surged to roughly $17 billion in the third quarter of 2026, and the AI lab swung back to an operating loss, The Information reported Monday, in a disclosure tied to the amended S-1/A registration statement the company filed with the Securities and Exchange Commission on November 2. The Information's report is paywalled beyond its headline, so Pulse is citing the figures as reported by that outlet pending the full public prospectus.
What changed since Anthropic's Japan filing
Pulse previously covered Anthropic's filing to sell up to $3.5 billion of shares to Japanese investors as part of its planned US listing -- a filing that disclosed a share-sale ceiling but not the IPO's total size or valuation. The new S-1/A amendment adds the first disclosed revenue and profitability figures since that filing, giving investors a second data point just as the roadshow conversation around a reported $1.5 trillion valuation intensifies.
What Anthropic does
Anthropic, founded in January 2021 by seven former OpenAI employees including siblings Daniela and Dario Amodei, is the San Francisco-based maker of the Claude family of large language models, sold through a chatbot, an API and the company's own agent harnesses, and has grown into one of the two AI labs, alongside OpenAI, that investors most often cite when pricing the broader AI infrastructure buildout. Its last disclosed valuation was $965 billion, set in a May 2026 Series H round led by Altimeter, Dragoneer, Greenoaks and Sequoia Capital -- a mark that already topped OpenAI's $852 billion valuation from March 2026. The company's product lineup includes the Claude Code developer tool and enterprise API access sold directly to large corporate customers, a commercial base The Information's reporting suggests is now large enough to produce a double-digit-billion-dollar swing in quarterly results.
The open questions in the amended filing
The Register, reviewing the amended prospectus, reported that Anthropic has disclosed multiyear compute commitments running for roughly the next two decades, and flagged risk-factor language warning that missed payments on those commitments could cascade into reduced compute capacity and further revenue pressure. The filing also discloses that some of Anthropic's largest customers are simultaneously its compute suppliers and competitors -- the cloud providers that host its training runs -- a structural risk The Register said the prospectus does not fully resolve. Pulse has not independently verified the specific dollar figures in the amended filing beyond what The Information and The Register have reported, and will update once the public prospectus is available in full.
What to watch next
Anthropic has not priced an IPO, and announced is not completed: the Japan filing disclosed only a share-sale ceiling, and the revenue and profitability figures reported here come from an amended registration statement, not a completed offering. The real test for the swing to an operating loss is whether it reflects one-time compute accounting or an ongoing cash dynamic -- a distinction that will shape whether the $1.5 trillion figure some investors have discussed holds up once the roadshow actually begins.