Google Cloud on Wednesday announced that its eighth generation of custom-built AI chips, or tensor processing units (TPUs), will be split in two. One chip, named the TPU 8t, will be geared for model training and another, the TPU 8i, is aimed at inference.

Inference is the ongoing usage of models, aka what happens after users submit prompts.

As you might expect, the company touts some impressive performance specs for these new TPUs compared to the previous generations: up to 3x faster AI model training, 80% better performance per dollar, and the ability to get 1 million+ TPUs to work together in a single cluster. The upshot should be a lot more compute for a lot less energy — and cost to customers — than previous versions. It calls these chips TPUs, not GPUs, because its custom low-power chips were originally named Tensor.

But Google’s chips are not a full frontal assault on Nvidia’s future, at least not yet.…