⚠ The Customers Are Becoming the CompetitorsHigh threat

The deepest structural problem with CoreWeave's performance-and-scale moat is that the companies it out-executes are the same companies it depends on for revenue — and they are building to replace it. CoreWeave's edge is over the generalist hyperscalers, but those hyperscalers, and the AI labs CoreWeave serves, are among its largest customers, and every one of them has powerful reasons to bring AI infrastructure in-house over time. Microsoft, which accounted for roughly two-thirds of CoreWeave's 2025 revenue, is simultaneously building out its own vast AI data-center capacity; the other hyperscalers are doing the same; and the largest AI labs are increasingly securing their own compute. CoreWeave's biggest customers are its most capable future competitors.

This dynamic caps the durability of the whole moat. CoreWeave is valuable to these customers today primarily as a fast, flexible supplement to their own capacity — a way to get compute now, while their own build-outs catch up. As those build-outs mature, the strategic logic points toward insourcing: the hyperscalers would rather own their AI infrastructure than rent it and enrich a competitor, and the economics of owning at their scale are compelling. So CoreWeave occupies an inherently unstable position — essential in the current scramble for compute, but serving customers whose long-term interest is to need it less. It can prosper enormously in the build-out phase, and the demand for compute may grow fast enough that even a shrinking share of a exploding market is large. But an investor must weigh the reality that CoreWeave's performance edge is over its own customers, that those customers are the richest companies in the world, and that they are spending hundreds of billions specifically to build what they currently rent — the clearest reason the moat is thin, and the sharpest long-term threat to the entire business — the largest customer alone was ~two-thirds of revenue¹.