Google (opens in new tab) on Nov. 9 announced two new configurations of its eighth-generation Tensor Processing Unit systems, the TPU 8tx and TPU 8i Max, according to the company's blog, SiliconANGLE (opens in new tab) and DatacenterDynamics. They follow the TPU 8t and TPU 8i systems unveiled in April.

Google (opens in new tab) said the TPU 8tx is optimized for "secure and confidential inference" for regulated enterprises and governments, keeping data and model weights protected in hardware even while in use. Amin Vahdat, Google (opens in new tab)'s chief technologist for AI and infrastructure, wrote that the new systems address hardware-level data security and million-token context scale, according to Google (opens in new tab)'s announcement.

The TPU 8i Max is aimed at AI workloads with very large context windows, SiliconANGLE (opens in new tab) reported, summarizing Google (opens in new tab)'s announcement. Google (opens in new tab) described both as full TPU systems — compute, networking, storage and packaging — rather than chips alone. Independent benchmarks of the new systems were not included in the announcement coverage.