Anthropic compute spending grows 4x annually, outpacing OpenAI's 3x rate
·2026.11.18 09:00
Key point
Anthropic spent nearly $10 billion on compute in Q3 2026, up from $433 million in 2023, while OpenAI forecasts $50 billion in total spending for 2026.
Details
- Anthropic's compute and infrastructure spending reached nearly $10 billion in Q3 2026, an annualized rate of approximately $40 billion, representing a 4x annual growth rate since 2023.
- OpenAI's compute spending has grown at roughly 3x annually, with a forecast of $50 billion for 2026, a figure based on testimony from President Greg Brockman rather than disclosed financial statements.
- Anthropic has contractually committed to $447.5 billion in compute spending through 2030, while OpenAI plans to spend around $750 billion over the same period.
- SpaceX accounted for approximately $5 billion of Anthropic's Q3 2026 compute spend, representing roughly half of the company's total compute costs for that quarter.
Data Sources and Methodology
- Anthropic figures are derived from its Form S-1/A filing, which disclosed quarterly compute spending from 2025 through Q3 2026 and annual totals for 2023 and 2024.
- OpenAI figures are compiled from media reports, including projections from The Information and testimony from President Greg Brockman, introducing higher uncertainty than Anthropic's disclosed data.
- Quarterly figures were converted to annualized run rates by multiplying by four, with dates assigned based on assumed steady exponential growth between reporting periods.
Spending Composition
- Most of Anthropic's compute spending supports internal R&D rather than external customer inference; in the first nine months of 2026, customer inference accounted for only 26% of compute spend.
- In contrast, media reports indicate that around 50% of OpenAI's compute spending was allocated to customer inference in 2025, up from roughly one-third in 2024.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.