Google unveils new AI chip systems TPU 8tx and 8i Max

What happened. In April Google showed its eighth TPU generation: 8t for training models and 8i for fast answers. On November 9 two more systems joined them. It is the first time Google has released four TPU systems in one year.

TPU 8tx. Data is usually encrypted when stored or sent, but it has to be unlocked in memory while a computer works on it. 8tx keeps data and model weights protected even while in use. It is built on the same silicon as 8t, with 5th generation AMD EPYC CPU hosts, and Google says it delivers up to 2x lower latency than 8t. It targets banks, governments and other regulated industries.

TPU 8i Max. When dozens of AI agents work across a whole codebase or hours of video, memory becomes the bottleneck. Google put two 8i chips in one package: double the compute, SRAM and HBM bandwidth of 8i, in the same rack space.

Software. Google open-sourced TorchTPU so PyTorch code runs on TPUs without rewrites, and released a tool suite that plugs into Claude Code, Cursor and Google Antigravity. The goal is simple: make it easier to move off Nvidia GPUs.

已确认的事实

• On November 9, 2026 Google unveiled TPU 8tx and TPU 8i Max, expanding its eighth TPU generation; that makes four new TPU systems in one year.^[1][2]
• TPU 8tx is built for confidential inference: data and model weights stay protected while in use; it uses the same silicon as 8t with 5th generation AMD EPYC hosts and delivers up to 2x lower latency than 8t.^[1]
• TPU 8i Max combines two 8i chips in one package, with double the compute, SRAM and HBM bandwidth of 8i.^[1][2]
• The TPU business has nearly 100 customers, including frontier labs, trading firms, retailers and universities.^[1][2]
• According to Bloomberg, cited by Gizmodo, both systems are expected to be available in 2027.^[2]
• Google open-sourced TorchTPU, so PyTorch code and Hugging Face models run natively on TPUs.^[1]

影响

• 已确认Large customers now have more AI hardware options beyond Nvidia.
• 我们的解读Protecting data while in use could win Google orders from banks and governments that could not send sensitive data to the cloud. This is our inference.
• 我们的解读Stronger TPUs add price pressure in the AI chip market. This is our inference.

仍未知

• The exact availability date and pricing of the new systems
• Which customers will move to 8tx and 8i Max first
• How Nvidia will respond

影响到谁

United States, Global. People who follow Alphabet and Nvidia stock, the AI market and AI tokens

来源

1. Google: Our newest TPUs: 8tx and 8i Max, built for the agentic era 2026年11月9日 · 官方
2. Gizmodo: Google Has Two New Weapons in Its Fight Against Nvidia 2026年11月9日 · 媒体