Our Take, Up Front: Anthropic's amended IPO filing tells two stories at once - the fastest revenue ramp enterprise software has ever recorded, and a balance sheet loaded with roughly $895 billion of commitments that will take decades to unwind. The revenue is real. The commitments are also real. Investors pressing for a lower IPO price are not being greedy; they are pricing in the risk that today's AI buildout becomes tomorrow's stranded capacity.
The numbers in the filing
In the amended Form S-1 filed with the SEC on November 2, Anthropic, the company behind the Claude AI platform, reported third-quarter revenue of approximately $17 billion, up about 1,127 percent from the same period a year earlier. Nine-month revenue reached $33.24 billion, roughly twelve times the prior-year figure. The pace, however, is decelerating: quarter-over-quarter growth slowed to 47 percent in Q3, after 144 percent in Q2 and 132 percent in Q1. Operating profit swung from a gain of $974 million in Q2 to a loss of $567 million in Q3, and the net loss widened to $23.3 billion for the quarter, reflecting heavy non-cash charges alongside real operating costs.
The company attributed the swing to a mix of factors: roughly $5 billion of compute purchased under more expensive short-term contracts, about $2 billion in non-cash employee donation-matching expenses, and price cuts and discounts on its models. In other words, the operating loss is partly the price of scaling capacity faster than contract terms would allow.
The $895 billion question
The most striking disclosure is the scale of long-term obligations. Anthropic said its purchase commitments and lease agreements total approximately $895.4 billion, substantially all related to compute and infrastructure. About $447.5 billion of that is expected to be paid by the end of 2030, with the remainder stretching into 2045. Roughly $69.8 billion in contractual payments falls due within the next twelve months alone. The company has been explicit that its disclosed model depends on future capital, and that its current liquidity covers only about twelve months of operations.
Anthropic's approach to securing capacity has been unusually aggressive. It has moved toward directly leased chips to separate capacity commitments from accelerator purchases, and it is exploring the conversion of old coal power plants and industrial facilities into data centers as it hunts for grid access and power. In April it committed more than $100 billion to AWS technologies over the next decade, securing up to 5 gigawatts of new capacity to train and run Claude. That hunger for compute is the engine of its growth, but it is also the largest single line item on its path to profitability.
Growth quality under the hood
Beneath the headline growth rate, the filing reveals a business whose revenue model carries its own complications. PitchBook's analysis notes that pay-as-you-go revenue and founder control give outside shareholders less than the top-line numbers suggest. The deceleration from 144 percent to 47 percent sequential growth is partly mechanical - a base effect of the company's own scale - but it also reflects deliberate price cuts and customer discounts as Anthropic fights for enterprise market share against rivals with deep pockets. The company is choosing growth over near-term profitability, and the S-1 makes that trade-off unusually explicit.
The pricing battle
The IPO itself has become a test of whether public markets will underwrite the AI capital cycle at the same scale private investors did. Anthropic filed its initial S-1 on October 26 targeting a listing on Nasdaq under the ticker ANT at a valuation of $1.5 trillion to $2 trillion. After last week's selloff in technology stocks, large institutional investors began demanding a lower price, pushing toward the $1.5 trillion to $1.6 trillion range and arguing for conservative pricing that leaves room for a first-day pop. PitchBook has gone further, valuing the company at $1.42 trillion, about 29 percent below the reported $2 trillion target. The company plans to start its roadshow as soon as next week, aiming for a listing in the week of November 16.
Governance adds another layer
Investors are also weighing a governance structure that concentrates control in special classes of shares that carry no economic rights. That structure gives founders control over a business whose financial trajectory will be defined by decisions made in the 2030s - a long horizon for outside shareholders to accept without economic votes.
Our view is that Anthropic is a genuinely exceptional revenue story, but the IPO price will be determined less by growth than by the credibility of the compute commitments behind it. For the wider AI supply chain - chipmakers, data-center developers, power providers and cloud vendors - the outcome matters enormously. A successful listing at a high valuation validates the buildout; a muted debut would send a warning through the entire AI capital stack. And for those suppliers, the near-term obligations are the clearest part of the picture: nearly $70 billion of contractual payments in the next twelve months is a concrete order book, whatever happens to the stock price.
