2 ms·

How much are you burning to keep your services running? For most of these gpus are required for inference and I know gpu instances aren't cheap