Anthropic reportedly signs $517 billion in compute deals after Dario Amodei warned rivals about reckless risk
Anthropic has signed compute contracts worth up to $517 billion over eleven months, locking in at least 14.8 gigawatts of computing power since October 2025 The company is planning its own data centers on top of its existing 1-2 gigawatts of capacity Anthropic's annualized revenue has topped $65 billion, while OpenAI's stands above $40 billion as of July Despite the massive commitments, Anthropic's planned capacity still falls short of OpenAI's 30-gigawatt target for 2030 CEO Dario Amodei previo
Analysis
TL;DR
- Anthropic has signed compute contracts worth up to $517 billion over eleven months, locking in at least 14.8 gigawatts of computing power since October 2025
- The company is planning its own data centers on top of its existing 1-2 gigawatts of capacity
- Anthropic's annualized revenue has topped $65 billion, while OpenAI's stands above $40 billion as of July
- Despite the massive commitments, Anthropic's planned capacity still falls short of OpenAI's 30-gigawatt target for 2030
- CEO Dario Amodei previously warned rivals about reckless risk-taking, but Anthropic is now racing to catch up on compute infrastructure
Why It Matters
The article highlights the intensifying compute arms race between Anthropic and OpenAI, two of the leading AI labs, and underscores the financial scale required to remain competitive in frontier AI development. It also reveals a striking reversal of positions: Amodei cautioned competitors about reckless investment, yet Anthropic is now making equally aggressive compute commitments, while Sam Altman is urging caution about "unsustainable silliness" from neo-cloud providers.
Technical Details
- Anthropic locked in at least 14.8 gigawatts of computing power since October 2025, building on an existing base of 1-2 gigawatts
- Many of Anthropic's compute contracts extend well past 2030, making direct capacity comparisons with OpenAI's 30-gigawatt 2030 target difficult
- Anthropic is planning its own data centers to support its compute needs
- Neither company can currently cover its compute commitments from revenue alone, signaling heavy reliance on future monetization or external funding
- OpenAI's 30-gigawatt target for 2030 remains the benchmark for compute scale in the industry
Industry Insight
- The compute arms race is escalating beyond what current revenue models can sustain, suggesting that AI labs will need to either dramatically increase pricing, secure massive external capital, or achieve breakthroughs that reduce compute requirements per unit of progress
- The role reversal between Amodei and Altman on investment caution highlights the tension between safety rhetoric and competitive pressure, which may influence how the industry approaches risk and regulation
- Neo-cloud providers and compute infrastructure companies are becoming strategic gatekeepers in the AI race, giving them outsized leverage over the pace and direction of frontier model development
Disclaimer: The above content is generated by AI and is for reference only.