Anthropic's compute costs climb to 58% of revenue in latest quarter
An updated filing from AI company Anthropic ahead of its planned IPO shows quarterly revenue of $17 billion, but a larger share going to compute. The company ties this to fast-growing demand and pricier short-term deals.

AI company Anthropic on Monday filed an update to its S-1, the registration document U.S. companies prepare before an initial public offering (IPO) of shares. The new data covers results through September 30 and shows that compute, the processing power needed to train and run AI models, is consuming a growing share of the company's revenue. That matters for prospective investors, because the company's profitability depends heavily on these costs.
According to the filing, quarterly revenue reached $17 billion, up from $11.5 billion in the previous quarter. The operating loss grew at the same time, largely because of compute spending. Compute took up 58% of revenue, compared with 48% a quarter earlier. The total compute bill for the quarter came to $10 billion.
Costly short-term deals
The filing says $5.0 billion of that amount was for compute under short-term commitments, which currently costs significantly more than comparable capacity bought under multi-year commitments. Anthropic was trying to keep up with fast-rising demand and had to secure capacity quickly. Known arrangements include payments of $1.25 billion a month to SpaceX, as well as multi-billion-dollar deals with Nscale, Akamai and Volta. These firms are so-called neoclouds, smaller providers that offer more flexible terms at higher prices than the large hyperscale cloud companies.
A chart in the filing tracks revenue for every dollar spent on compute. The figure rose steadily from the start of 2025, peaked at just over 2x last quarter, and then fell to 1.72x.
What comes next
Anthropic states that it expects compute and infrastructure expenses to keep rising in absolute terms and to remain its largest operating expense for the foreseeable future. The company is trying to cut costs by building its own compute base under a project called "Anthropic Direct," but fast-growing demand means that effort is unlikely to keep pace. With revenue growing at nearly 50% a quarter, high compute costs may worry investors little for now. The industry still depends on the data center buildout, which could slow if bond markets become nervous or local opposition intensifies.

