← Latest news 
Anthropic’s Compute Bet: $30B run rate and a CFO map for scaling AI without breaking the company
Business
Published on 15 May 2026
The CFO admits he spent a third of his time on compute
In a first podcast appearance, Anthropic CFO Krishna Rao lays out how the AI lab scaled from about $250M in run-rate revenue two years ago to roughly $30B by early 2026—driven by a ruthless “compute” strategy. He describes long-horizon planning, multi-vendor chip bets across Trainium, TPUs, and GPUs, and why interpretability and safety research is now boosting enterprise trust and retention.
- Revenue jumped from ~$9B run rate to over $30B in the first quarter of 2026
- Rao says he spends 30–40% of his working hours on compute procurement and planning
- Anthropic is contract-bound by long lead times, making compute choices today shape strategy 18 months out
- Customer traction is strong: nine Fortune 10 clients and net dollar retention above 500%
Read the full story at YourStory
This summarization was done by Beige for a story published on 
YourStory