Comment by dartos

The hardware required is the same, just in different amounts.

It’s less (gross) expensive for inference, since it takes less time, but the cost of that time (per second) is the same as training.