Google unveils two new specialized Tensor Processing Unit (TPU) variants, the TPU 8tx and TPU 8i Max, as part of its Gen-8 lineup. Coverage describes the 8tx as designed for confidential AI workloads, while the 8i Max focuses on higher-memory, long-context inference for AI systems that handle large amounts of information per request.
Multiple outlets report that the 8tx includes hardware support intended to protect sensitive data at the chip level for confidential computing. Several sources also say the 8tx is paired with AMD EPYC Turin hosts, reflecting a move to help enterprises run regulated or privacy-sensitive workloads. Google’s chief technologist for AI and infrastructure, Amin Vahdat, is quoted in outlets summarizing Bloomberg’s reporting, including context that the chips are reworked versions of existing silicon rather than entirely new designs.
For the TPU 8i Max, outlets emphasize dual-chip packaging to increase processing and memory for long-context tasks, with use cases cited such as drug discovery, math, and coding. Several sources frame the launch as part of Google’s competitive effort in custom AI chips against Nvidia, with at least one outlet noting Google says its current TPU supply is fully committed to customers.
