Business

Markets

Technology

Artificial Intelligence

Cryptocurrency

Economics

SUBSCRIBE →

Español · English

Google expands TPU systems for confidential AI and long-context workloads

November 9, 2026. Google announced two additional systems in its eighth generation of tensor processing units: TPU 8tx and TPU 8i Max. The company says 8tx is designed to protect sensitive information at the hardware level, while 8i Max combines two TPU 8i chips for workloads requiring more memory and context.

Google says 8tx uses the same silicon as TPU 8t alongside AMD EPYC processors and can cut latency by up to half compared with 8t under the conditions it describes. It says 8i Max doubles computing and memory bandwidth versus TPU 8i. These are manufacturer claims, not independent benchmarks.

Why it matters

AI infrastructure is becoming more specialized: organizations need both protection for data in use and enough memory to process larger amounts of information.

Rockstar Briefing editorial analysis

For business customers, value will depend on testing actual workloads, total cost and availability. Competition among accelerator providers also turns on software, security and adoption; an announced specification does not guarantee production savings.

Source: Google, November 9, 2026. Illustration generated using artificial intelligence.