Google is introducing two versions of its AI chips, available next year, as it expands its tensor processing unit (TPU) lineup to compete with Nvidia for cloud and enterprise workloads.

The TPU 8tx is for confidential computing for governments and regulated industries, and pairs with a central processing unit (CPU) from Advanced Micro Devices. Confidential computing involves protecting data while it is processed in hardware-based isolated environments.

The TPU 8i Max is designed for AI applications that need long context windows, where models can consider more information at once. This reduces the need to split large inputs and supports workloads such as coding, drug discovery, and maths.

The launch comes as cloud providers and chipmakers invest in AI accelerators beyond Nvidia.

🔗 Source: Bloomberg