Anthropic Plans $518 Billion AI Infrastructure Buildout Over a Decade

Anthropic plans to spend $518 billion on AI infrastructure over the next decade, according to a report from ET Datacenters. The figure places the frontier model developer's long-term compute ambitions squarely in the open.

The report does not break the number down further, but spread across ten years the commitment implies annual spending in the tens of billions of dollars. That is capital-expenditure scale rather than the expansion of a single data center.

For a frontier lab, compute is the precondition for training the next generation of models and serving them at scale. Larger models, longer context windows and heavier agent workloads all push demand for GPU clusters and for the electricity that feeds them, and the company that locks in supply first gains room to iterate on its own schedule.

That dynamic helps explain the industry's broader shift from renting compute to building it. Long-term contracts, campus siting and power procurement are becoming competitive dimensions as important as algorithm design, with vendor relationships and energy access now shaping product timelines.

The open questions are funding and pacing. Whether $518 billion is a binding commitment or a long-range aspiration depends on how quickly model demand grows, and on how long capital markets stay patient with heavy AI spending.

At the industry level, a commitment of this size would lift demand across data centers, power, chips and cooling, and could sharpen competition for limited capacity. Every link in that supply chain could be repriced as a result.

What to watch next: whether Anthropic publishes specific sites, partners and construction timelines, and how the outlay is matched against revenue growth. For observers, the real question is not whether to build, but how fast — and who ultimately pays.