The Enterprise Business CoreWeave Doesn't HaveThin moat

Amazon, Microsoft and Google each serve hundreds of thousands of ordinary enterprise customers running unglamorous workloads — databases, storage, internal applications — that generate steady, diversified, price-insensitive revenue regardless of what happens to AI. CoreWeave has essentially none of that — it holds an estimated 15-20% of the dedicated AI IaaS market and a trivial share of cloud overall¹.

The absence is deliberate and was the right decision. Building a general-purpose cloud would have meant competing with the hyperscalers on their own ground with a fraction of their capital, and CoreWeave's whole advantage is that its infrastructure is purpose-built for one workload rather than compromised across many. Specialisation is why it won.

It also means there is no floor. When a general-purpose cloud's AI business slows, the rest of the company continues; when CoreWeave's does, there is nothing underneath. The moat pages describe genuine switching costs for customers running training workloads on CoreWeave's infrastructure — those are real and they operate within a contract term rather than across a downturn.

Watch for any move into inference or enterprise services. Inference workloads are steadier, more numerous and less concentrated than training, and a meaningful inference business would be the first thing resembling a floor under this company's revenue.