4 ms·

I really want to see how that fares against Nvidia for ML development, because it's hard to compete with that kind of flexibility and price.