Google Adds TPU 8tx and TPU 8i Max to Eighth-Generation TPU Lineup
Google announced two additions to its eighth-generation TPU lineup on November 9, 2026: TPU 8tx and TPU 8i Max.
1 source 1h ago
Reporting from 1 source: GIGAZINE.
Why it matters
Google is answering the spread of AI agents by building systems around existing TPU chips rather than new silicon, splitting the eighth generation between confidential-data processing and large-context inference.
Key Facts
- Google announced TPU 8tx and TPU 8i Max on November 9, 2026.
- TPU 8tx uses the same chip as TPU 8t and combines it with AMD's 5th-generation EPYC processor as host CPU.
- TPU 8tx reduces processing latency by up to half compared with TPU 8t through the Boardfly networking technology.
- TPU 8i Max combines two TPU 8i chips into a single package.
- The eighth-generation TPU lineup now has four types: TPU 8t, TPU 8i, TPU 8tx, and TPU 8i Max.
TPU 8tx pairs the TPU 8t chip with AMD's 5th-generation EPYC processor as host CPU and protects confidential data at the hardware level, with Boardfly networking cutting latency by up to half. TPU 8i Max combines two TPU 8i chips in one package to handle large context windows.
Google is not designing new silicon for either system. TPU 8tx uses the same chip as TPU 8t and adds AMD's 5th-generation EPYC processor as host CPU, plus hardware-level protection and isolation of data and model weights while AI is processing. The Boardfly low-latency networking technology cuts processing latency by up to half versus TPU 8t, with almost no performance impact, according to the company. TPU 8tx also handles small-scale training, distillation, and reinforcement learning, so companies can tune models without moving confidential data elsewhere.
TPU 8i Max is aimed at AI that reads widely while it works. Memory transfer speed limits processing when an AI must pull huge amounts of data, as with entire codebases or several hours of video, so the system pairs two TPU 8i chips in a single package. The eighth-generation lineup now holds four entries: TPU 8t, TPU 8i, TPU 8tx, and TPU 8i Max.
- TPU 8tx: Uses the TPU 8t chip with AMD 5th-generation EPYC host CPU; protects confidential data and model weights at hardware level; Boardfly networking cuts latency by up to half versus TPU 8t.
- TPU 8i Max: Combines two TPU 8i chips in a single package for AI that references large amounts of information.
Written with AI by Yomimono from the 1 cited source below, including Japanese-language reporting where cited. See our editorial policy.
Add Yomimono as a preferred source on Google