Anthropic's amended IPO filing: $17 billion in Q3 revenue, $895 billion in compute commitments and a warning on self-improving AI
Anthropic filed an amended IPO prospectus with the US Securities and Exchange Commission on Monday, adding third-quarter results: about $17 billion of revenue, $33.24 billion over nine months, and a net loss of $104 billion driven mostly by accounting changes. The updated risk factors also say that using AI to speed up AI research could outpace safety work, and that the company may soon meet its own internal threshold for "dramatically accelerated progress".
In an amended Form S-1 filed on Monday 2 November, Anthropic added figures for the third quarter to the prospectus it unveiled last week, which had covered only the first half of the year. The company reports revenue of $33.24 billion for the nine months to 30 September, up roughly 1,200% on the same period a year earlier, and about $17 billion for the third quarter alone. Quarterly growth is slowing, however: revenue rose about 47% over the second quarter, against 144% growth in the second quarter and 132% in the first.
Nine-month net loss was $104 billion, compared with $25.1 billion a year earlier; Anthropic says most of it comes from changes in the fair value of convertible notes and derivative liabilities. The third-quarter net loss was $23.29 billion. Compute and infrastructure expenses rose by $14.4 billion over the nine months, an increase of 329%, and the company has purchase commitments and leases with an aggregate committed spend of about $895.4 billion, almost all of it for compute and infrastructure.
The filing's risk factors were also updated. Anthropic says that using AI systems to accelerate its own research "may cause the capabilities of our models to advance more rapidly than our ability to develop, validate, and implement corresponding safety, alignment, and security measures", and that it may in the future fail to identify a significant acceleration in its rate of progress. TechCrunch and other outlets read the amended text as a sign that the company may soon meet its own internal definition of recursive self-improvement; reports on the filing note that no current model is designated above that threshold, which is covered by the company's Responsible Scaling Policy.
The amended prospectus lands as investors grow wary of the cost of the AI build-out. OpenAI confidentially filed its own prospectus in June, but its executives have signalled they are aiming for a listing in 2027.