Anthropic Signs Deal for All of SpaceX's Colossus 1 Compute Capacity
On May 6, 2026, Anthropic agreed to use all compute at SpaceX's Colossus 1 data center, over 300 MW and 220,000 NVIDIA GPUs, and raised Claude usage limits.
AI infrastructure news: GPUs and custom AI accelerators, data centers, cloud capacity and energy, from NVIDIA, AMD, Google, Microsoft, OpenAI and others.
8 stories
Compute is the foundation of modern AI. This category follows new AI accelerators and GPUs, custom silicon from cloud providers, large data center projects and the energy needed to run them. It also explains how these investments affect model availability, pricing and performance for developers.
On May 6, 2026, Anthropic agreed to use all compute at SpaceX's Colossus 1 data center, over 300 MW and 220,000 NVIDIA GPUs, and raised Claude usage limits.
Google announced TPU 8t and TPU 8i at Cloud Next on April 22, 2026, splitting its eighth-generation TPU into separate chips for training and inference.
On April 20, 2026, Anthropic committed over $100 billion over ten years to AWS for up to 5 GW of compute, and Amazon agreed to invest up to $25 billion more.
Meta and NVIDIA announced a multiyear deal on February 17, 2026 covering millions of Blackwell and Rubin GPUs, standalone Grace and Vera CPUs and networking.
OpenAI signed a multiyear deal on January 14, 2026 for 750 MW of Cerebras computing power, reportedly worth over $10 billion, to speed up AI inference.
On November 18, 2025, Anthropic committed to buy $30 billion of Azure compute and up to 1 GW of NVIDIA systems, while NVIDIA and Microsoft agreed to invest.
Anthropic said on November 12, 2025 it will invest $50 billion in custom-built U.S. data centers with Fluidstack in Texas and New York, coming online in 2026.
Anthropic's Build AI in America report, released July 21, 2025, estimates U.S. AI needs at least 50 GW of electric capacity by 2028 and proposes energy reforms.
No matches. Try the site search.