Anthropic's amended IPO filing says its AI may soon sharply speed up its own research
Anthropic's amended IPO prospectus adds a warning that future models may soon meet its own threshold for AI-accelerated research, alongside Q3 revenue of $16.96 billion and an operating loss.
Anthropic told prospective shareholders on Monday that its own AI may soon cross an internal threshold for speeding up the company's research, a warning added to the amended S-1 registration statement it filed with the SEC on 2 November. The same filing brought the Claude maker's financials up to 30 September, showing $16.96 billion in third-quarter revenue and a quarterly operating loss after a profitable second quarter.
The amendment comes a week after Anthropic made its IPO prospectus public with numbers only through June. The company has applied to list its Class A stock on the Nasdaq Global Select Market under the symbol ANT.
A new risk factor about the pace of AI research
Anthropic's Responsible Scaling Policy sets capability thresholds that trigger extra safety work. One of them covers AI systems that substantially accelerate AI research and development, the scenario often called recursive self-improvement. In the amended filing, the company says its internal estimates and expectations about future models, "rather than our current models," indicate it "may imminently meet" the policy's definition of "dramatically accelerated progress attributable to AI."
The filing adds that no current model has been designated as exceeding that threshold and that the company is still assessing the data. It also concedes a measurement problem: much of its evidence on AI-driven acceleration comes from lagging indicators, so Anthropic says it has in the past failed to spot a significant jump in its rate of development until after it happened, and may again. TechCrunch flagged the new language.
The business section explains why this matters inside the company. Anthropic says Claude now helps run experiments, generate and process training data, automate evaluations and safety audits, and write and review code for the next generation of models, a process it calls "closing the loop." It says it does not expect to be able to safely let Claude design and build successive frontier models on its own without meaningful human supervision, and that it is considering changes to its scaling policy to keep humans in control of frontier model development. The company says it will publish further reports on the threshold as its policy requires.
What the third quarter added
Revenue for the nine months ended 30 September reached $33.24 billion, compared with $2.55 billion a year earlier, according to the filing. Third-quarter revenue of $16.96 billion was up 1,127% from the same quarter of 2025. Growth is slowing in sequential terms, though: the quarter was about 47% above the second quarter, against roughly 144% in the second quarter and 132% in the first, CNBC calculated.
Costs rose faster in the quarter. Compute and infrastructure expense was $9.86 billion, up from $5.56 billion in the second quarter, and the company swung to an operating loss of $567 million after operating income of $974 million in the prior quarter. Over nine months, the operating loss narrowed to $1.26 billion from $5.22 billion a year earlier.
The headline net loss is far larger. Anthropic reported a nine-month net loss of $104 billion, mostly from changes in the fair value of convertible notes, derivative liabilities and SAFEs, which accounted for $94.3 billion of it. The third-quarter net loss was $23.29 billion.
Customers, cash and compute commitments
The filing lists more than 1,900 customers paying $1 million or more over the trailing twelve months and about 250 paying $10 million or more, and puts net revenue retention at roughly 900% as of 30 September. About 44% of nine-month revenue came through cloud provider marketplaces, which bill customers on Anthropic's behalf.
Anthropic says it had raised $141.1 billion in total and held $95.6 billion in cash and cash equivalents plus $36.3 billion in short-term investments at quarter end. Against that, it has purchase commitments and leases with total committed spend of about $895.4 billion, substantially all for compute and infrastructure. Roughly $447.5 billion of that is expected before the end of 2030, with most of the rest falling between 2031 and 2045.
Those obligations are the backdrop for the roadshow. Anthropic sources much of its capacity from Amazon, Google and SpaceX, and some from neoclouds, at a time when investors are questioning how fast new data center capacity can come online, CNBC reported. The filing also lands as other labs keep shipping frontier models, including Mistral's Large 4 open weights last week. OpenAI confidentially filed its own prospectus in June and has signaled a 2027 debut.
The price range and share count are still blank in the amended prospectus; Anthropic will need to fill them in before it can price the offering.
