⚠ Scale Built on Depreciating, Financed AssetsHigh threat

CoreWeave's scale is impressive but rests on the least durable kind of asset: GPUs, which depreciate rapidly and which CoreWeave has financed with debt secured against the chips themselves. Unlike the fiber, buildings, and power infrastructure that make a traditional data-center moat durable, GPUs lose value quickly as Nvidia releases each new, far more capable generation — a two-year-old chip can be worth a fraction of its cost as the frontier moves on. CoreWeave's enormous fleet is therefore a wasting asset, earning revenue today but eroding in value continuously, and requiring constant, capital-hungry refreshment to stay current. The scale that looks like a fortress is really a treadmill.

The debt financing sharpens the risk into something acute. CoreWeave borrows against its GPUs to buy more GPUs, so its balance sheet carries heavy debt secured by collateral that is depreciating — a structure that works while the chips retain value and generate revenue, but that is dangerously exposed if GPU values fall faster than expected, if utilization drops, or if the AI-compute demand that fills the fleet softens. The interest on that debt runs into the hundreds of millions of dollars a quarter, and the depreciation into the billions, together swamping the company's explosive revenue and keeping it deeply unprofitable. So CoreWeave's scale is a genuine competitive asset that is simultaneously its greatest financial vulnerability: an enormous fleet of fast-depreciating chips, bought with debt, that must be perpetually refreshed and kept fully utilized to justify the borrowing — a leveraged bet on the continued value and demand for GPUs — carried at >$500M of interest a quarter¹ — not the durable, compounding scale advantage of true infrastructure.