Google Launches Two Custom AI Chips to Expand Confidential Computing and Memory Capacity

On November 10, TechCrunch reported that Google has launched two eighth-generation Tensor Processing Unit (TPU) variants: the TPU 8tx and TPU 8i Max, expanding the application scenarios for its self-developed AI chips.

Co-designed with AMD and based on the TPU 8t, the TPU 8tx is tailored for confidential model training and inference, safeguarding sensitive data and model weights in use at the hardware level. Its custom inference network can run inference during training, enabling real-time distillation and the training of smaller models.

The TPU 8i Max integrates two inference-focused TPU 8i chips into a single package, doubling memory capacity, processing power, and bandwidth, which enables models to process a wider range of information during queries.