Anthropic Secures $3.4 Billion Deal for 1.7 Gigawatts of Gas Generators for AI Data Centers
The AI compute build-out accelerates with major infrastructure investments in power generation and new chip launches, as companies race to secure dedicated energy and specialized processing capabilities.
The story
The race for AI compute capacity is increasingly focused on securing dedicated power infrastructure, as evidenced by Anthropic's recent $3.4 billion deal. The AI research company has committed to purchasing 1.7 gigawatts of heavy-duty gas-fired generators from Caterpillar, originally ordered by OnePWR Solutions.
This move, finalized in late September but reported on November 8-9, 2026, represents the largest known single acquisition of power-generation equipment by an AI lab. The agreement highlights a growing trend among leading AI companies to bypass intermediaries and directly invest in their own electricity infrastructure, ensuring reliable power supply for their expanding data centers amidst rising demand and equipment shortages. Anthropic's broader infrastructure commitments are projected to exceed $895 billion through 2045, predominantly for computing power and infrastructure.
Silicon
TPU 8tx
Maker: Google
What: Designed for confidential computing, combining Google's chip technology with AMD's fifth-generation EPYC processors to keep sensitive data encrypted during processing, offering up to twice the speed of its predecessor.
For Whom: Government agencies, financial institutions, and organizations with strict data protection requirements using Google Cloud's TPU 8tx system.
TPU 8i Max
Maker: Google
What: Combines two TPU chips into a single package, doubling computing power and memory bandwidth for large-scale data processing and longer context windows.
For Whom: Applications handling extensive datasets and technical information, such as coding assistants, mathematical reasoning, and drug discovery.
Fifth-generation EPYC processors
Maker: AMD
What: Server processors integrated into Google Cloud's TPU 8tx system, bolstering its AI infrastructure capabilities.
For Whom: Google Cloud's AI infrastructure, specifically for powering the TPU 8tx system.
The build-out
| Project | Who | Scale | Where |
|---|---|---|---|
| Gas-fired generator procurement | Anthropic (from Caterpillar via OnePWR Solutions) | 1.7 gigawatts, valued at $3.4 billion | For Anthropic's expanding AI data center infrastructure |
Supply & policy signals
TSMC's October revenue reached NT$586.89 billion ($17.88 billion), up 59.7% year-on-year and 14.7% from September.
Implication: Indicates sustained, strong demand for advanced AI chips, outstripping supply and driving significant revenue growth for the world's largest contract chipmaker.
China revenue share for ASML, Lam Research, KLA, and Tokyo Electron fell in Q3 2026, with ASML seeing the steepest drop (down 19 percentage points to 23%).
Implication: Reflects the impact of US export restrictions on advanced chip equipment to China, alongside increased AI chip investment in South Korea and Taiwan reshaping global semiconductor equipment sales.
What we'll be watching
- Google's TPU 8tx and TPU 8i Max are expected to become available in 2027, expanding specialized AI compute options.
- Microsoft plans production deployments of its Maia 300 AI accelerator chip beginning in 2027.
- NVIDIA is forecasting Q3 fiscal 2027 revenue of approximately $108 billion, plus or minus 2%.
- AMD anticipates Q4 2026 revenue of around $17 billion, plus or minus 3%.
- Broadcom expects Q4 fiscal 2026 revenue of approximately $34.8 billion.
Reporting + analyst voices: grounded via Google Search at publish time.