⚠ Switching Costs Are Weaker Than They LookModerate threat

CoreWeave's switching costs are real but modest, and an investor should not overweight them, because AI cloud is a business the industry is deliberately building to be portable. Wary of repeating the lock-in that made the traditional clouds so sticky, the AI ecosystem has standardized around portable software layers — common frameworks, containers, orchestration tools — expressly designed to let workloads move between providers with less friction. This erodes the switching costs that might otherwise protect CoreWeave: a customer that has built on portable tooling can move to a cheaper or better alternative, or to its own infrastructure, with far less pain than migrating a deeply-integrated enterprise application.

The switching costs are also weakest exactly where they would matter most — against the largest customers. The hyperscalers and best-funded AI labs that dominate CoreWeave's revenue and backlog are the ones with the engineering resources to move workloads at will and the strongest motive to avoid dependence on a single outside provider, especially one they compete with. For these customers, the switching costs are a modest inconvenience, not a real barrier, and their contracts hold them only until renewal, when their own build-outs or alternative providers may make leaving attractive. CoreWeave does enjoy genuine stickiness within a contract term, and for smaller customers the switching friction is more meaningful. But an investor should recognize that the lock-in protecting CoreWeave is moderate at best, actively undermined by an industry designing for portability, and thinnest against the very customers whose departure would hurt most — a real but limited moat element that cannot be relied upon to hold the concentrated, powerful customer base — top-two customers were ~65% of revenue¹ — once the contracts that currently bind it expire.