Google Reportedly Pulls Marvell Into a Two-Chip TPU Plan That Could Reshape AI Inference For ASICs

Google is reportedly discussing two new AI chips with Marvell.

One is a memory processing unit to pair with current TPUs, and the other is a next-generation TPU for model inference.

The effort could boost inference performance and expand Google’s TPU ecosystem.

It also suggests continued demand for custom AI accelerators despite tight semiconductor manufacturing capacity.