McKinsey points out the quandary facing companies like CoreWeave

So-called neocloud companies are facing a dilemma: They need to move up the AI stack to avoid being commoditized, but they risk competing against their big hyperscale customers if they do.

Neocloud operators, or GPU-as-a-service providers, sprang up to take advantage of the huge demand for compute using GPU accelerators for AI. Many customers were forced to rent the chips because they were unable to secure their own supply.

But this business model is fragile, according to consultants McKinsey & Company, because it is inherently commoditized. There is limited differentiation in renting out access to specific hardware - typically Nvidia products - and price-driven competition can be fierce.

Investors backing neoclouds such as CoreWeave, Nebius (both publicly traded), and Crusoe (privately held), are betting that the companies will be able to transition to offering services built on their AI infrastructure instead. Such platforms would comprise training tools, inferencing services, and domain-specific stacks ...