# Google adds TPU 8tx for secure inference and TPU 8i Max for long‑context agents

> The new chips pack two 8i units together or pair a TPU with an AMD EPYC CPU for regulated workloads.

Oossa · 2026-11-12 · https://oossa.com/en/google-adds-tpu-8tx-for-secure-inference-and-tpu-8i-max-for-long-context-agents

Google has expanded its eighth‑generation Tensor Processing Unit (TPU) line. The TPU 8tx uses the same silicon as the earlier TPU 8t but swaps the Arm‑based host for a 5th‑gen AMD EPYC CPU and a Boardfly network, targeting hardware‑isolated inference for regulated enterprises and governments. The TPU 8i Max puts two 8i chips in one package, doubling compute, SRAM and HBM bandwidth for long‑context agent work.

## The facts

- TPU 8tx replaces Arm hosts with a 5th‑gen AMD EPYC CPU.
- TPU 8i Max combines two 8i chips, doubling performance.

## Why it matters

Enterprises that need secure, low‑latency AI can deploy the new TPU 8tx without custom hardware design, while developers of large‑context agents get double the processing power in the same footprint.

## Sources & references

1. [Google reworks TPU silicon into specialized systems for agents and security](https://www.tomshardware.com/tech-industry/artificial-intelligence/google-adds-two-new-specialized-chips-for-ai-agents-and-security-to-8th-gen-tpu-lineup-tpu-8i-max-puts-two-chips-on-one-package-for-long-context-work-tpu-8tx-swaps-arm-hosts-for-amd-epyc) – Tom's Hardware, 2026-11-12

Last updated: 2026-11-12