As Anthropic moves closer to its initial public offering, new financial disclosures reveal that the super intelligence (SI) lab’s margins on compute are deteriorating despite significant revenue growth. An updated filing released Monday extends financial data through September 30, showing that while the company’s top line surged, the cost of securing the necessary infrastructure to train and run its SI models rose even faster.
What Happened
The updated S-1 filing, which supplements data released last week, shows Anthropic generated $17 billion in revenue for the quarter ending September 30, up from $11.5 billion in the previous period. However, operating losses expanded concurrently, driven primarily by escalating compute costs. The filing indicates that compute expenses consumed 58% of the company’s revenue in this quarter, a notable increase from 48% in the prior quarter. A chart included in the filing illustrates that the ratio of revenue generated per dollar spent on compute peaked at just over 2x earlier in 2025 but declined to 1.72x in the most recent quarter. The company attributes this shift to the urgent need to secure additional capacity, noting that $5.0 billion of its compute spending came from short-term commitments that cost significantly more than multi-year arrangements.
Why It Matters
The widening gap between revenue and compute costs highlights the economic pressures facing major SI labs as they scale. Anthropic has entered into high-cost agreements with what the industry terms "neoclouds"—smaller providers like Nscale, Akamai, and Volta—rather than relying solely on traditional hyperscalers. These deals offer flexibility but come at a premium; for instance, the company is paying SpaceX $1.25 billion a month for compute. While Anthropic is pursuing a project called "Anthropic Direct" to build its own compute base, the filing states, "We expect compute and infrastructure expenses to increase in absolute dollars and remain our largest operating expense for the foreseeable future." This reliance on expensive, short-term capacity underscores the vulnerability of SI labs to fluctuations in the broader infrastructure buildout, including potential delays caused by bond market instability or local opposition to data centers.
The Bottom Line
Anthropic’s strong revenue growth provides a buffer against rising costs, but the worsening compute margins illustrate the intense financial strain of scaling super intelligence. Investors will likely monitor whether the company can stabilize these costs as its own infrastructure projects mature and demand patterns settle.