Anthropic is executing a strategic pivot from its traditional cloud-first model, signing preliminary agreements for more than a dozen US data centers with a combined capacity exceeding one gigawatt. This move marks a decisive shift toward direct infrastructure ownership, aiming to eliminate the markup associated with renting GPUs and TPUs from major cloud providers like Amazon Web Services and Microsoft.
The initiative follows Anthropic’s November 2025 announcement of a $50 billion investment in US data center development. These new lease agreements signal that those capital plans are transitioning into concrete operational assets.
Central to this expansion is a critical support arrangement with Google. The tech giant is currently negotiating a financial guarantee to backstop Anthropic’s lease obligations. This structure provides landlords with assurance, leveraging Google’s $2 trillion market valuation to secure the real estate commitments. Google’s role is multifaceted: it remains an investor, a cloud provider, and now a potential financial guarantor for Anthropic’s physical infrastructure.
This infrastructure buildup mirrors industry-wide trends among competitors such as OpenAI and xAI. All major AI developers face the same constraint: training frontier models requires massive compute power, which in turn demands significant data center space and electricity. By controlling its own servers, Anthropic aims to secure the essential resources needed to scale its artificial intelligence capabilities efficiently.