Anthropic secures massive compute agreements, escalating AI infrastructure race
The AI firm has locked in up to $517 billion in compute capacity over eleven months, intensifying competition with OpenAI's larger spending roadmap.
Anthropic has reportedly entered into compute procurement contracts valued at up to $517 billion over an eleven-month period, according to recent reporting. The agreements, which span multiple providers, are intended to secure the substantial processing power needed for training and running advanced AI models. The figure, while enormous, remains below OpenAI's publicly stated $750 billion infrastructure commitment through 2030.
The move follows statements made in early 2026 by Anthropic's chief executive, Dario Amodei, who cautioned industry peers about the dangers of accelerating investment too quickly. His earlier warnings have given way to a more aggressive posture as the company seeks to narrow the gap with its chief rival. The shift highlights the intense pressure among leading AI labs to lock in scarce computational resources before competitors do.
OpenAI's Sam Altman has separately voiced concerns about what he described as "unsustainable silliness" in the broader compute buildout, particularly pointing to the rise of less-established cloud providers. Industry observers note that the rapid expansion of data center capacity carries risks of oversupply and financial strain, even as demand for AI services continues to grow. The contrasting strategies of the two firms underscore the uncertainty surrounding the long-term economics of AI infrastructure.