Market · AI in practice
Google builds TPU 8tx to run AI on sensitive data with hardware protection
On November 9, 2026, Google announced TPU 8tx, an eighth generation TPU system for secure and confidential inference that keeps data and model weights protected in hardware during processing too. Availability dates and pricing were not disclosed.
What the sources report
On November 9, 2026, Google expanded its eighth generation of TPUs with two systems, TPU 8tx and TPU 8i Max, reaching four TPU systems launched in a single year. TPU 8tx is optimized for secure and confidential inference and also works for small scale training. It uses the same silicon as TPU 8t, with 5th Generation AMD EPYC CPU hosts, and moves security into the hardware layer: sensitive data and model weights stay protected and isolated even while they are being processed. Google says latency is up to 2x lower than on TPU 8t, with a performance penalty close to zero.
The system lets teams fine tune models on private data, distill large models into smaller ones and train agents through reinforcement learning without migrating sensitive data between separate clusters. Google says TPU 8tx was built for regulated enterprises and governments running frontier or proprietary models, and autonomous agents, alongside their most sensitive data. According to TechTarget, Google did not provide availability dates or pricing. TPU 8tx is a new system configuration built on existing eighth generation silicon.
My take
Training and serving on the same confidential system is the part of the announcement with the most practical effect. Fine tuning on private data, distillation and reinforcement learning without taking sensitive information to another cluster cut down the copies and transfers a security team would otherwise have to control. Combined with protection of data during processing, that answers one of the most frequent compliance objections to running frontier or proprietary models on customer data. Still without a date or a price, TPU 8tx already goes on the list of requirements I would bring to the next conversation with a cloud provider.
Sources
- https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/tpu-8i-max-and-8tx/ (opens in a new tab) Google
- https://www.techtarget.com/news/google-specializes-tpus-for-agentic-confidential-computing/874084/ (opens in a new tab) TechTarget
- https://www.datacenterknowledge.com/data-center-chips/google-adds-tpu-8tx-8i-max-for-enterprise-ai (opens in a new tab) Data Center Knowledge