For which customers would accessing GPU compute through a hyperscaler make more sense?

Primarily, larger enterprises like Intel, Nvidia, and Cadence would prefer hyperscalers. These companies are deploying ChatGPT equivalent generative AI for internal use, which is not available externally. They can do coding and validation among other internal uses. These enterprises prefer hyperscalers due to existing relationships and a desire to avoid disruption. They also appreciate the common interface. So, larger companies, in terms of size and revenue, still go with hyperscalers. Smaller companies, which are making partial increments to the models and don't have as much funding, tend to choose CoreWeave.