The hyperscalers aren't going to be on CUDA for long, Nvidia are taking too much of a cut. Google run tensor chips, Amazon and Microsoft both have their own accelerators in the works. They are massively incentivised to do so right now.
Now they do all re-sell Nvidia GPUs in their cloud businesses, but there the cost will be passed very directly on to infrastructure customers, who will see the competing higher level services from those cloud providers (hosted models) likely at lower or competitive prices, and it's going to be harder to justify renting CUDA cores for custom software.
Now they do all re-sell Nvidia GPUs in their cloud businesses, but there the cost will be passed very directly on to infrastructure customers, who will see the competing higher level services from those cloud providers (hosted models) likely at lower or competitive prices, and it's going to be harder to justify renting CUDA cores for custom software.