inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
Why it matters
Multiple gigawatts of TPU capacity landing from 2027, alongside Trainium and GPU fleets, is a concrete read on how much inferenceRunning a trained model to get answers — the phase where AI is actually used, as opposed to trained.Full definition → and training headroom one frontier lab is buying and when it arrives.
