Anthropic invests $50bn in American AI data centres with Fluidstack

Anthropic announced a $50 billion investment in purpose-built US data centres, developed with the cloud provider Fluidstack at sites in Texas and New York, with further locations to follow. The company said it chose Fluidstack for its ability to deliver gigawatt-scale power capacity quickly, and that the sites were expected to come online progressively through 2026. Anthropic said the buildout would create roughly 800 permanent jobs and 2,400 construction jobs, and that the new capacity was intended to support both continued frontier research and the growing demand for Claude from what it described as hundreds of thousands of business customers.

The announcement marked Anthropic’s first large move into directly financing and shaping physical infrastructure, rather than relying solely on compute bought from cloud partners such as Amazon and Google, both already investors in the company under existing cloud-credit and compute-supply arrangements. Anthropic framed the investment as supporting the Trump administration’s AI Action Plan and domestic technological leadership, language that echoed similar infrastructure commitments other frontier labs, including OpenAI, made around the same period as competition for data-centre capacity intensified.

The move reflected a broader shift among the largest AI developers toward treating compute capacity as a strategic asset to be secured directly rather than purchased on the open market, at a time when demand for training and inference compute was widely reported to be outpacing the buildout of new data-centre capacity across the industry, and when the scale of such commitments — measured now in tens of billions of dollars per company — had itself become a subject of scrutiny.