⚠ Power Is a New Bottleneck and a New Capital SinkModerate threat

CoreWeave's push to secure power addresses a real and growing constraint, but it deepens the company's capital intensity and operational complexity in ways that add risk to an already-stretched model. Power has become a binding bottleneck on AI infrastructure, and controlling it is genuinely valuable — but acquiring and developing electricity and sites requires large additional capital, on top of the tens of billions already committed to GPUs and data centers, deepening CoreWeave's reliance on continued financing. It also drags the company into power development, permitting, and long-lead-time construction — a slow, difficult, capital-heavy world far from its core skill of operating GPU clusters, bringing execution and integration risks it has less experience managing.

The deeper concern is that securing power layers more fixed, long-dated, capital-intensive commitments onto a business whose demand rests on a possibly-cyclical boom. Power assets and the debt to fund them are long-term obligations; if AI demand cools, CoreWeave would be left servicing not only its GPU debt but also its power and site commitments, against reduced revenue — compounding the leveraged build-ahead-of-demand risk that already defines the model. The move is strategically sound and reflects real foresight about where the constraints are heading, and controlling power could be a meaningful advantage in a power-scarce future. But an investor should weigh that it also enlarges the capital sink, extends the company into an unfamiliar and difficult domain, and adds another layer of fixed, cyclical exposure to an enterprise already defined by its debt and its dependence on the AI boom continuing — a sensible bet that nonetheless increases the capital at risk and the ways in which a cooling of demand could strain the whole structure — on top of $31-35B of committed GPU capex¹.