Anthropic’s planned $518 billion AI infrastructure buildout hinges largely on deals that cannot be cancelled, according to a Reuters analysis published on September 29.

The analysis examines the structure of the commitments behind one of the largest infrastructure bets in the technology industry, finding that much of the buildout is tied to agreements with cancellation protections that lock in spending regardless of how demand develops.

The report comes as the AI industry races to secure data centre capacity, chips and power supply, with leading labs signing long-term deals to guarantee compute for training and serving frontier models.

Anthropic has rapidly expanded its compute footprint to train and deploy successive generations of its models, competing with rivals that are also pouring hundreds of billions into infrastructure.

The structure of these commitments matters because it determines how much financial flexibility the company retains if growth slows. Deals that cannot be cancelled transfer risk from the infrastructure providers to the AI lab, making execution on revenue growth central to the bet paying off.