CoreWeave
GPU cloud provider; hosts OpenAI's compute under an $11.9bn contract.
Last refreshed: 16 May 2026
Is CoreWeave now the de facto compute backbone for OpenAI's scaled-back Stargate?
Timeline for CoreWeave
Mentioned in: OpenAI cuts compute target by $800bn
Data Centres: Boom and BacklashWhat is CoreWeave's contract with OpenAI?
Did CoreWeave go public and what was its IPO price?
Why did OpenAI choose CoreWeave over building its own data centres?
Background
CoreWeave is the largest independent GPU cloud operator in the United States, supplying AI training and inference infrastructure to hyperscalers and frontier AI labs that cannot build fast enough on their own. It signed an $11.9 billion five-year contract with OpenAI in March 2025, and in April 2026 agreed a multi-year Claude inference deal with Anthropic. The company operates 32 data centres across the US and Europe with approximately 250,000 GPUs. Its significance to the data-centre supply race was underlined when OpenAI declined its option on Crusoe's Abilene, Texas flagship lease while simultaneously deepening its CoreWeave commitment.
Founded in 2017 as Atlantic Crypto (a Ethereum mining operation) by Michael Intrator, Brian Venturo, Brannin McBee, and Peter Salanki, CoreWeave pivoted to GPU cloud after the 2018 crypto crash. Nvidia invested $100 million in April 2023, and the company raised $1.5 billion at a $27 billion valuation in a March 2025 IPO — among the largest US tech IPOs in three years. Microsoft accounted for over 60% of 2024 revenues before the OpenAI contract rebalanced that concentration.
CoreWeave's growth crystallises the hyperscaler compute bottleneck: cloud operators are years behind AI labs' appetite, and third-party GPU clouds have become critical swing capacity. The company's rapid scaling contributed to the record $110.75 billion combined Q1 2026 capex the four major hyperscalers reported.