← Back

Wire · technology

Google releases new AI chip with doubled HBM bandwidth

Published

10 November 2026

Topic

technology

◆ Sectors

AI & ML

◆ Geography

United States

◆ Source

Read at eu.36kr.com →

◆ Verified

Fusion42 · 10 November 2026 · Fusion42 review

Google has released two new 8th-generation TPU models, the TPU 8i Max and TPU 8tx, that double HBM bandwidth and improve latency for AI computation. These chips target confidential inference, large context windows, and memory bandwidth challenges, and Google also upgraded its unified open-source TPU software stack for easier developer access.

This Wire brief sits within Fusion42's coverage of AI & ML, and 3 sources have reported it between 9 Nov 2026 and 10 Nov 2026.

◆ ◆ The Wire takeaway

Google just doubled AI chip performance on memory bandwidth and cut inference latency by half. You building AI tools should test if this new hardware can unlock faster model runs or confidential workflows in your stack.

◆ Coverage

3 sources · first reported 9 Nov 2026 · latest 10 Nov 2026

◆ Related on Wire

◆ Topics

AI & MLgoogletpuai-chipconfidential-computingdeveloper-toolshardware-acceleration