Anthropic's Amended IPO Filing: $16.96 Billion in Q3 Revenue, a $567 Million Operating Loss and $5 Billion of Short-Term Compute

On 14 October, Reuters, citing the Financial Times, reported that Anthropic had told investors its third-quarter revenue was $16.8 billion and that it turned a profit in the quarter. The amended prospectus now gives the company's own figures. Revenue was slightly higher than reported, but on the standard accounting basis the quarter ended in an operating loss. The table shows where the money went.

The quarter in numbers

• Growth: revenue rose 47% from Q2 to Q3, after 144% from Q1 to Q2. Year on year it rose 1,127%.
• Price cuts: the filing says price reductions in new model versions and contractual discounts took about $600 million off Q3 revenue compared with Q2.
• Nine months to 30 September 2026: revenue of $33,242 million and an operating loss of $1,260 million.
• Net loss: $23,291 million in Q3, mostly a $21,936 million charge for changes in the fair value of convertible notes, derivatives and SAFEs, which is an accounting item rather than cash spent running the business.

Why the profit disappeared

Compute and infrastructure went from 48% of revenue in Q2 to 58% in Q3. Anthropic says most of its compute is bought 12 to 18 months ahead under multi-year contracts, and when demand runs past that it buys short-term compute that "currently costs significantly more". Q3 included $5.0 billion of it, and the filing says it expects short-term compute use to keep rising in the near term, squeezing margins. Other operating and administrative costs also jumped, from $1,031 million to $3,588 million, which the filing attributes mainly to non-cash charges from its employee donation-matching program. Personnel costs fell slightly, from $1,857 million to $1,711 million, because some option awards finished vesting.

Other details in the filing

• Customers: more than 11,900 customers generated $100,000 or more in trailing-twelve-month revenue at 30 September 2026, up from about 400 in March 2025. Net revenue retention was about 900%.
• Claude Code spending: in Q3, Claude Enterprise users who didn't use Claude Code spent less than $75 a month on average, while the top 5% of Claude Enterprise users who did spent about 20 times that.
• Commitments: about $895.4 billion of purchase commitments and leases, substantially all for compute and infrastructure. About $69.8 billion is due in the 12 months after 30 September 2026 and about $447.5 billion before the end of 2030.
• Listing: Anthropic has applied to list its Class A shares on the Nasdaq Global Select Market under the symbol ANT. The price range, the share count and its run-rate revenue at 31 October are all still blank in this amendment.
• Control: a single Class F share held by a company owned by the seven co-founders will carry 50.1% of the voting power on most matters.
• Safety disclosure: the risk factors say Anthropic's internal estimates indicate it "may imminently meet" its Responsible Scaling Policy's definition of dramatically accelerated progress attributable to AI. It says no current model has been designated as exceeding that threshold.

How the October figures compare

The October report put Q3 revenue at $16.8 billion; the filing's $16,956 million is about $150 million higher. The "profit" in that report doesn't match the filing's operating loss of $567 million. It may refer to Anthropic's Adjusted Operating Margin, which excludes stock-based compensation and some other items, and which the filing puts at 15% for the quarter. Fortune and Crypto Briefing both reported the roughly $17 billion figure, and Crypto Briefing the $567 million operating loss. The nine-month net loss is about $104 billion, so a $104 billion figure is not the quarterly number.

What this means if you rely on Claude

• Capacity is still tight. Anthropic says demand ran past its long-term compute contracts in Q3, and it expects to buy more short-term compute in the near term.
• Prices are moving down per model. The roughly $600 million hit from price cuts in Q3 suggests newer Claude models are cheaper per task than older ones, so re-check your costs when you switch versions.
• Watch the safety threshold. If Anthropic formally designates a model as meeting its AI R&D threshold, its policy calls for stronger safeguards, which could affect how and when new models reach customers.