Anthropic details $265.6B compute commitments in S-1, cites OpenAI and Google as rivals
The filing outlines rapid revenue growth alongside massive long-term compute obligations, framing a high-stakes bet on recursive self-improvement and enterprise AI.
1 source · cross-referenced
- Anthropic’s S-1 filing discloses $265.6B in non-cancelable compute purchase commitments through 2045 and $321B in lease agreements primarily for compute and infrastructure.
- Revenue grew fourteenfold to $16.3B in the first half of 2026 compared to the first half of 2025, with adjusted income of $1.62B after removing accounting items.
- Anthropic names only OpenAI and Google as rivals and reports 900% net revenue retention, with about three-quarters of customers spending at least $100,000 annually.
- The company faces intense competitive pressure from frontier labs and open-source models, requiring continuous model training to maintain performance.
Anthropic’s S-1 filing outlines a business strategy centered on aggressive compute investment, with $265.6 billion in non-cancelable purchase commitments for compute and $321 billion in lease agreements primarily for compute and infrastructure capacity, both with terms extending through 2045. The company frames these commitments as necessary to sustain its model development and competitive positioning.
The filing reports fourteenfold revenue growth to $16.3 billion in the first half of 2026 compared to the same period in 2025. After adjusting for accounting items such as an $80 billion loss and employee compensation, Anthropic reports $1.62 billion in adjusted income for the first half of 2026, signaling that revenue growth has outpaced costs in reported periods.
Anthropic identifies OpenAI and Google as its only named rivals, emphasizing a focused strategy that avoids consumer-facing products like image generators or social media clones. Instead, the company targets enterprise use cases such as coding and cybersecurity, where it reports 900% net revenue retention. Approximately three-quarters of customers in the first half of 2026 spent at least $100,000 annually.
The company acknowledges intense competitive pressure from both frontier labs and open-source models, noting that performance gains require continuous model training. Anthropic’s reliance on new model releases to drive revenue growth creates a cycle of escalating compute demands, as it must train more efficiently than larger competitors to maintain its position.
Anthropic’s S-1 includes placeholders for specific spending figures through December 31, 2030, but confirms that purchase and lease commitments extend through 2045. The filing’s ‘Commitments and contingencies’ section details the $265.6 billion in non-cancelable purchase commitments with terms between one and 20 years, alongside $321 billion in lease agreements for compute and infrastructure capacity that have not yet commenced.
- Nov 11, 2026 · TechCrunch — AI
Data center projects face high uncertainty despite billions in AI infrastructure investment
Trust74 - Nov 11, 2026 · TechCrunch — AI
Nvidia board launches internal probe into CEO’s alleged affair and national security concerns
Trust74 - Nov 10, 2026 · TechCrunch — AI
OpenAI acquires Odessia, an AI travel agent startup founded by Sonder’s creator
Trust79