CoreWeave trims IPO to $1.5b as OpenAI GPU deal shapes AI cloud race

CoreWeave’s downsized IPO is a reminder that even in an AI boom, public markets are becoming choosier about how they price pure‑play GPU clouds. Cutting the raise from $2.7 billion to $1.5 billion and accepting a lower valuation doesn’t change the strategic picture much: this is still a capital‑intensive bet that hyperscale customers will keep pouring workloads into specialist providers rather than only the big three clouds. What really matters is the structure of CoreWeave’s revenue — heavily concentrated in a handful of clients, with OpenAI and Microsoft reportedly accounting for the bulk of sales. That concentration makes the company both central to and fragile within the current AGI supply chain.

For the race to AGI, the more salient detail is the multi‑year, nearly $12 billion infrastructure deal with OpenAI and the planned acquisition of Weights & Biases. Those moves effectively turn CoreWeave into a vertically integrated training platform: GPUs on the bottom, MLOps tooling on top, and a flagship frontier lab in the middle. If the IPO succeeds even at the reduced size, it will validate public‑equity appetite for financing the massive capex required to train successive frontier models and may encourage similar listings from other specialist clouds.