Anthropic filed an amended public-offering prospectus with the U.S. Securities and Exchange Commission that combines two unusual measures of scale: quarterly revenue approaching $17 billion and approximately $895.4 billion in contractual compute and infrastructure commitments. The S-1/A, filed November 2, also adds a more consequential warning. Internal estimates concerning future models indicate that the company may imminently meet its own threshold for research and development progress that is “dramatically accelerated” by AI.[1]

The infrastructure figure is not money already spent, nor does it describe a single purchase. Anthropic says the total combines purchase commitments and lease agreements, including short-term commitments and leases that have not yet begun. About $447.5 billion would be incurred before December 31, 2030, while substantially all of the remainder is expected between 2031 and 2045. Approximately $69.8 billion is due during the twelve months following September 30, 2026.[1]

Those obligations show how strongly Claude’s expansion depends on data centers, chips, and power capacity contracted years in advance. They also increase Anthropic’s exposure if demand, compute prices, or technical progress develops differently from its forecasts. The prospectus describes the commitments as “substantially all” related to compute and infrastructure. It does not say that the full amount represents capacity already installed or available today.

On the financial side, Anthropic reported $16.96 billion in third-quarter revenue and $33.24 billion for the first nine months of 2026. CNBC reported a quarterly net loss of $23.29 billion and a nine-month net loss of roughly $104 billion, driven largely by changes in the fair value of convertible notes and derivative liabilities.[2] Those accounting losses should not be treated as a direct measure of cash consumed, but they illustrate the financial complexity accompanying the company’s preparation to become publicly traded.

The technical disclosure requires even more care. Anthropic is not claiming that Claude has already entered an autonomous “recursive self-improvement” cycle, or that a current model can independently build its successor. The SEC filing says the company’s internal estimates and expectations for future models — rather than its current models — indicate that it may soon meet the Responsible Scaling Policy definition of “dramatically accelerated progress attributable to AI.” Anthropic adds that it has not designated any current model as exceeding that threshold and continues to assess the underlying data.[1]

The prospectus also acknowledges a measurement problem: Anthropic might fail to recognize a significant acceleration in its development rate until after it has happened. The company says it has previously failed to identify important capability increases immediately. If capabilities were to outpace its safety, alignment, and security measures, Anthropic warns, the likelihood of safety incidents, regulatory intervention, and reputational harm could rise. This is a corporate risk disclosure made under securities-law obligations, not independent evidence that the threshold will be reached on a particular date.

The combination is what makes the filing material. Anthropic is committing resources at a scale that assumes years of compute-intensive growth while telling prospective investors that AI may accelerate the research process itself before evaluation systems can keep pace. The document does not establish that this transition has already occurred. It does turn a possibility debated for years into a formal contingency that Anthropic considers close and consequential enough to include in its public prospectus.

Sources

[1] Read More — Anthropic, PBC — Form S-1/A, filed November 2, 2026
[2] Read More — Anthropic releases amended IPO prospectus with Q3 financials — CNBC