> Once a model is built, the biggest cost is inference

Something i cant find any reliable data for, but would help for a sense of scale: How much use before its equal to training?

I.e. assuming you have the training data and setup, and we only care for compute - How hours of using eg Kimi K3 / Fable, for it to equal the compute required to train it?

if you assume that training requires about 3x the compute of inference (one forward pass, one backward pass, parameter updates), and we take DeepSeek-V3 since their numbers are public.

they used ~14.8 trillion tokens with about 2.66 million GPU hours. 14.8 * 3 = 44.4 t inference tokens.

obviously, this is back of the envelope math, but at 100t/s you would need like ~14k years. scale this to >100k GPUs and your in the hours to a couple days range.