A unified API for all the GPU clouds?

I'm thinking something that would let me provision and shut down gpu instances across all the clouds with GPUs through one api interface. I know skypilot exists but that's more from training than just infra.