Anthropic reported that aggressive price cuts on its Claude models erased about $600 million of third‑quarter revenue growth, while compute and infrastructure expenses surged 77%, turning a $974 million operating profit from the prior quarter into a $567 million loss, according to the company’s amended SEC filing.

Price Reductions Undermine Revenue Gains

The AI startup said the decline reflects discounts tied to new model releases and contractual incentives. Claude Opus 5.5 lowered token prices relative to Opus 5, and the newly launched Haiku 5.5 targeted high‑volume users with a cheaper rate. Anthropic estimates that Opus 5.5 cut per‑task costs by roughly 40% and Fable 5.1 reduced typical workloads by about 25%, combining lower pricing with efficiency gains.

Compute Costs Outpace Revenue

Compute and infrastructure spending climbed from $5.56 billion to $9.86 billion, a 77% increase, meaning the firm now spends roughly 58 cents on compute for every dollar of revenue, up from 48 cents a quarter earlier. The rise includes research, training, and internal operations, not just serving paying Claude customers. Short‑term commitments alone accounted for $5 billion of compute that Anthropic says costs significantly more than multiyear contracts.

Implications for the Upcoming IPO

Anthropic’s filing for a public offering reveals aggregate lease and purchase commitments of about $895.4 billion, virtually all linked to compute capacity. The scale of these obligations has prompted questions about whether infrastructure partners, chiefly Google and Nvidia, can secure financing to meet the demand. Analysts note that while cheaper, more capable models may drive higher usage, the company must reduce the cost of building and running those models faster than price cuts erode margins.

What to Watch Next

Investors will be monitoring Anthropic’s ability to align pricing with declining compute costs, any revisions to its long‑term capacity contracts, and competitive pressure from OpenAI’s upcoming IPO, which offers a different pricing and quota structure.