Anthropic IPO Filing Puts AI Compute Bill on Trial

The amended S-1 forces traders to price a different kind of AI risk

Anthropic gave IPO investors on Monday what Wall Street always says it wants: more recent numbers. What the artificial intelligence company supplied with its amended Form S-1 filed with the SEC was a sharper, more complicated picture. Revenue is still exploding. But the growth curve is bending. The loss line is dominated by accounting swings, not just operations. And the single most market-moving figure is the company’s $895.4 billion in aggregate purchase commitments and leases, a number that turns the AI race into a balance-sheet stress test for anyone who plans to buy stock. The original filing had only given investors Anthropic’s numbers through the first half of the year.

Revenue is huge, but deceleration is the new watch item

For the nine months ended Sept. 30, Anthropic reported $33.24 billion in revenue, up roughly 1,200% from the same period last year. In Q3 alone, revenue was around $17 billion, up about 1,127% year over year. Those are still jaw-dropping numbers. If a mature listed company told you it had nearly quadrupled year-over-year growth in a quarter, traders would not be asking whether it is interesting; they would be asking whether the rest of the AI trade is underpriced.

The twist is quarter-to-quarter growth. Q3 revenue rose about 47% from the previous quarter, down from 144% growth in Q2 and 132% in Q1. A 47% sequential gain is not weak. But compared with the previous two quarters, it is a slowdown. In IPO valuation, the second derivative matters. Investors may tolerate less explosive acceleration if the company can show durable monetization, predictable capacity and credible paths to profitability. Anthropic’s filing now gives them a chance to test those assumptions in real time.

The loss figure is noisy, but it still needs explanation

The amended document showed a $104 billion net loss for the nine-month period. Anthropic said most of that came from the change in fair value of convertible notes and derivative liabilities. The company had a loss of $25.1 billion over the same stretch, and it recorded a $23.29 billion net loss in Q3.

Traders should separate accounting from cash. Fair-value movements on convertible instruments and derivatives can create spectacular headline losses that do not map one-to-one to current operating burn. That does not mean the number is meaningless. It means investors will ask harder questions: how much of this is paper volatility, how much is dilution pressure, and what does the company’s cash position look like if private AI valuations keep resetting?

The $895.4 billion commitment is the real IPO puzzle

Anthropic said it has entered into purchase commitments and lease agreements with an aggregate committed spend of about $895.4 billion, substantially all related to compute and infrastructure. That is not a rounding error; it is a capital structure, supply-chain and execution forecast all in one line. For the nine months ended Sept. 30, compute and infrastructure expenses rose by $14.4 billion, up 329% compared with the same period last year. “As our compute and infrastructure capacity grows, the associated compute and infrastructure expenses have increased,” Anthropic said in the filing.

This is where the IPO gets harder to underwrite. AI revenue may be scaling fast, but the cost of training and serving models is scaling with it. If commitments are long-dated, take-or-pay, or tied to scarce GPU capacity, they can compress margins and create financing needs. The market’s fear is simple: will demand keep rising faster than the money and hardware needed to meet it?

AI infrastructure anxiety is now a roadshow topic

Wall Street’s mood around AI buildout has turned twitchier. Concerns about rising interest rates making debt more expensive for data center developers and neoclouds are real. Recent reports from MSFT, GOOGL and AMZN added fuel because aggressive AI spending is eating into cash flows at the hyperscalers. Some investors worry that computing capacity may not come online quickly enough to satisfy AI labs such as Anthropic.

That matters because Anthropic gets much of its capacity from Amazon, Google and SpaceX, while also leaning on neoclouds for compute. A revenue forecast that assumes relentless scaling needs a supply chain that does not choke. A $895.4 billion commitment shows Anthropic is preparing for that problem. It also shows the cost of preparing is enormous. If capital markets tighten, even companies with strong top lines can find themselves squeezed between contractual spend and slower customer payment.

Safety positioning does not replace the cash question

Anthropic CEO Dario Amodei co-founded the company with researchers who left OpenAI, now its chief rival. The company has worked to position itself as more safety-conscious than competitors, and its filing includes more than 90 pages of risk factors. Among them: potential political headwinds, complex infrastructure needs and the difficulty of balancing safety and commercial objectives. “We consider it our personal responsibility to help the world understand these risks, to prevent or mitigate them when they arise from our AI models, and to work with governments and the rest of the industry to address them at a societal level,” Anthropic’s co-founders said in the filing.

That framing may help in a market that has learned to ask uncomfortable questions about AI governance, misuse and geopolitical scrutiny. It will not, by itself, answer the pricing question for public investors. OpenAI confidentially filed its prospectus in June, but its executives have signaled a 2027 debut. Anthropic’s roadshow is happening sooner, and the amended filing puts the company on display while the AI narrative is under pressure.