Nov 9, 2026, 5:00 AM PT · found Nov 9, 2026, 6:30 AM PT
Google adds TPU 8tx and 8i Max to its eighth-generation line
Verdict: Vendor-reported · Model releases · Adoption and cost
Google announced two more TPU systems with latency and bandwidth claims. All figures are Google's, several have no named baseline, and no availability date is given.
Google announced two new TPU systems, bringing its eighth-generation line to four.
- Claim: TPU 8tx, built for confidential inference, has up to 2x lower latency than TPU 8t. TPU 8i Max pairs two 8i chips and doubles compute, SRAM and HBM bandwidth over 8i. Google
- Evidence: Google cites 2.7x training performance for 8t and 80% better inference price-performance for 8i against April’s generation, plus TPU Sync cutting time to first token by over 90%. Google
- Caveats: Several figures name no baseline. Google claims leading price-performance on InferenceX and AgentX with no numbers, and gives no availability or pricing for the new systems. Nobody outside Google has measured them.
- Why it matters: Inference cost, not just model quality, decides what agents are affordable, but this is a spec sheet until it can be rented and measured.
Verdict history
- Nov 9, 2026, 6:30 AM PT: Vendor-reported. Primary post read via fetch summary. No independent measurement found yet, and the post gives no InferenceX or AgentX figures. Posted about 90 minutes before this run.