Google adds two TPU systems for safe AI and long context
Google says its new TPU 8tx keeps data safe while in use, with a maximum of 2 times lower latency than TPU 8t.
Google announced two new systems in its eighth generation of TPUs, the AI processors that it makes: TPU 8tx and TPU 8i Max. It is the first time that Google has released four TPU systems in one year. Google says almost 100 customers use TPUs. It says the new systems are for firms that must follow the law and for agents with very large context.
How it works
TPU 8tx is for work on private data. Data is usually encrypted in storage and on the network, but not while a computer works on it. Google says TPU 8tx keeps data and model weights safe in use. TPU 8i Max puts two TPU 8i processors on one package.
The numbers
- TPU 8tx: a maximum of 2 times lower latency than TPU 8t, with almost no drop in speed.
- TPU 8i Max: two times the compute capacity, the SRAM and the memory bandwidth of TPU 8i.
- TPU 8i Max: two times the silicon in the same rack space.
What is not known
All numbers are from Google. Google gives no price and no test by a different company here. It does not tell when customers can buy the two systems.
Insight: Google removes what stops buyers from using TPUs
Google and Meta released TorchTPU on 21 October 2026, and PyTorch code can operate on TPUs. Oracle tells that it will offer TPUs from 2027. TPU 8tx is for data that has limits in law. This reads as a move to remove, one by one, the problems that a buyer has with TPUs. Price and capacity are not known. Get the two before you select.
Sources
Posted