RunPod has its issues, but the way it handles payment is basically my ideal. Nothing brings peace of mind like knowing you won't be billed for more than you've already paid into your wallet. As long as you aren't obliged to fulfil some SLA, I've found that this on-demand scaling compute is really all I need in conjunction with a traditional VPS.

It's great for ML research too, as you can just SSH into a pod with VScode and drag in your notebooks and whatnot as if it were your own computer, but with a 5090 available to speed up training.