Meta signed a 1.6GW compute-capacity agreement with data-center startup Crusoe in June 2026. The capacity comes from sites in Childress, Texas and Warrenton, Missouri, and ranks among the largest single AI computing deals by a hyperscaler this year. ASAP frames this deal within the broader scramble for AI infrastructure.

What was signed

The key point is that Meta secured 1.6GW of compute capacity from Crusoe. Per June 2026 reports, Meta signed contracts to buy capacity at Crusoe's Childress, Texas and Warrenton, Missouri data centers. The value of the deal was not disclosed, and neither Meta nor Crusoe issued a separate statement.

Scale and context

The 1.6GW figure is among the largest single AI computing deals struck by a hyperscaler this year. Meta projects AI-related capital expenditure of $60-80 billion annually in 2026, and has pledged to invest $600 billion in U.S. infrastructure and jobs over the next three years. The deal is thus one facet of Meta's aggressive infrastructure expansion.

Why it matters

The Meta-Crusoe deal is a marker of how far Big Tech's 2026 scramble for compute and power has gone. As models scale, the power and compute needed for training and inference become the bottleneck, and companies are locking up capacity from specialists like Crusoe beyond their own data centers. It is also a strategy to diversify supply and reduce single-vendor dependence.

Wrap-up

The Meta-Crusoe 1.6GW deal shows the weight of the 2026 AI race shifting from models to infrastructure. As the $60-80 billion annual capex and $600 billion investment pledge suggest, securing power and compute now translates directly into AI competitiveness.

Source: reporting on the Meta-Crusoe 1.6GW capacity deal (June 2026, Data Center Dynamics, Bloomberg, Tech Startups).