CoreWeave, a cloud provider specialising in Nvidia GPUs, raised approximately $1.5 billion in an initial public offering that valued the company near $19 billion, according to Reuters. The listing came after the company secured billions of dollars in debt financing backed by its GPU fleet.
The IPO is the cleanest public-market test of a thesis that defined the last cycle: that renting compute to AI labs is a business large enough to be worth the capital intensity it requires.
The balance-sheet model
CoreWeave's distinctive move was to borrow against its GPUs, using the hardware itself as collateral. That converts a capital-expenditure problem into a financing problem, and it works as long as the GPUs generate enough rental income to service the debt.
The structure concentrates risk. If GPU prices fall, utilisation drops, or a faster generation makes the collateral obsolete faster than the amortisation schedule assumes, the model compresses quickly.
Why the listing matters
It gives public investors their first direct exposure to the economics of AI compute, separate from the model labs and the chipmaker. The reported financials will let the market price the layer between them for the first time.