4 ms·

I expect this trend to begin reversing as we start getting AI models that are intended to run locally.