Google Releases TPU 8tx and TPU 8i Max Compute Systems
Our newest TPUs: 8tx and 8i Max, built for the agentic era
Google announced an expansion of its eighth-generation TPU portfolio, officially launching TPU 8tx for confidential inference and TPU 8i Max for ultra-long context windows. TPU 8tx features hardware-level confidential computing and Google's low-latency Boardfly interconnect, enabling secure fine-tuning and inference on sensitive data within a single system. TPU 8i Max packages two TPU 8i dies together, doubling compute performance and HBM bandwidth to overcome memory walls during multi-turn agent workloads.
The article provides detailed specification comparisons across confidential computing and dual-chip packaging, helping readers evaluate infrastructure choices for ultra-long context and sensitive inference workloads.
Source:Google Infrastructure · blog.google