Anthropic recently updated its S-1 filing ahead of its IPO, revealing fresh insights into its financials, specifically the economics of compute. The company posted $17 billion in revenue for the quarter ending September 30, 2026, up from $11.5 billion the previous quarter, reflecting robust growth. However, despite this revenue jump, operating losses widened significantly due to soaring compute costs. Critically, the portion of revenue consumed by compute expenses increased from 48% to 58%, marking a deterioration in compute margins when investors might have expected improvement.

The surge in compute costs is largely driven by Anthropic’s need to rapidly scale computing capacity to meet growing demand. This forced the company to rely heavily on short-term high-cost compute agreements with neocloud providers like SpaceX, Nscale, Akamai, and Volta. These suppliers offer flexibility but at higher prices compared to multi-year contracts with traditional hyperscalers. Anthropic spent $10 billion on compute in the quarter, with half attributed to these short-term deals, highlighting a costly scramble to secure enough infrastructure that is unlikely to ease soon.

To address this, Anthropic is investing in its own compute infrastructure project called “Anthropic Direct,” aiming to reduce reliance on expensive external providers over time. Still, the company acknowledges that demand growth will continue to outpace these efforts, keeping compute a dominant and costly expense in the near future. The report emphasizes that compute and infrastructure will remain the largest operating costs as Anthropic expands. While strong revenue growth currently overshadows these challenges, vulnerabilities remain tied to broader data center capacity constraints and market conditions affecting infrastructure investments.

This filing also signals the broader AI industry's dependency on a trillion-dollar compute infrastructure buildout, which faces risks like potential slowdowns in data center expansions due to financial or regulatory hurdles. Anthropic's financials reflect the high stakes and tight economics of running large-scale AI labs, where compute cost management is crucial to long-term profitability. As Anthropic approaches its IPO, investors are closely watching whether it can balance rapid growth with improving cost efficiencies amid volatile infrastructure dynamics.