Google has added two systems to its eighth-generation Tensor Processing Unit line, TPU 8tx and TPU 8i Max, the company said on 9 November. It is the first time Google has introduced four TPU systems in a single year, joining TPU 8t and TPU 8i, which launched at Google Cloud Next in April.
What is TPU 8tx for?
TPU 8tx is optimised for secure and confidential inference and is also described as suitable for small-scale training. Google says it is built for regulated enterprises and governments that run frontier or proprietary models and autonomous agents alongside sensitive data. The system provides hardware-level confidential compute, so teams can fine-tune models on private data, distil large models into smaller ones and train agents through reinforcement learning without moving that data between separate clusters.
What does TPU 8i Max do differently?
TPU 8i Max is aimed at the "memory wall", the point where memory bandwidth, not raw compute, slows large AI workloads. Google combined two low-latency TPU 8i chips on a single package. The company says this delivers double the compute, double the SRAM and double the high-bandwidth-memory bandwidth of TPU 8i, and fits twice the silicon in the same rack footprint. The target workloads are agents that work across a whole codebase or hours of video, and million-token context windows.
What changes for developers?
Google paired the hardware with software changes. It says last month's open-source TorchTPU lets PyTorch code and Hugging Face models run natively on TPUs. New items include a TPU agent suite in preview, an open-source XLA Serving Engine, and Cloud Pathways, now generally available on Google Kubernetes Engine.
Google Cloud's TPU Sync post describes another open-source piece. It says TPU Sync cuts key-value cache transfer latency by more than 90% and reduces time to first token by an average of 90% for large open-weight mixture-of-experts models of 235 to 480 billion parameters, compared with running without it. Those figures are Google's own measurements.
For readers
These are cloud and data-centre products for businesses, not consumer hardware, so there is no retail price. Google did not give availability dates for the new systems in the post. For more AI news, see Cultvia Pulse.
Frequently asked questions
What is the difference between TPU 8tx and TPU 8i Max?
TPU 8tx is for confidential inference on sensitive data, while TPU 8i Max is for very large context windows and compute density.
Can consumers buy these chips?
No. They are Google Cloud infrastructure for businesses and governments.
Editorial desk
Stories signed by the Cultvia editorial desk follow our editorial policy; every story lists its sources at the end and errors are fixed under our corrections policy.
Last updated:
What changed?
- : First published
Sources
- Our newest TPUs: 8tx and 8i Max, built for the agentic era — Google, 9 November 2026
- Introducing TPU Sync: Fast KV cache and weight transfer for inference and RL on TPU — Google Cloud Blog, 9 November 2026