# Google makes Cloud Pathways generally available for TPU workloads

_Published Monday, November 9, 2026 at 8:38 AM EST · AI, Infrastructure · Latest · Tier 2 — Notable_

![Google makes Cloud Pathways generally available for TPU workloads — Primary](https://storage.googleapis.com/gweb-cloudblog-publish/images/1-header.max-1400x1400.jpg)

Google has made Cloud Pathways generally available on Google Kubernetes Engine, giving cloud customers a runtime that coordinates thousands of TPU chips through a single JAX client process. Google says Pathways users processed more than 130 million Cloud TPU chip-hours over the last 90 days.

The runtime lets training jobs resize across remaining TPU slices after hardware failures, using snapshots stored in host memory. Python containers running alongside workloads handle tasks such as data loading and checkpointing.

Availability covers regions supporting TPU v5e, v5p, Trillium and Ironwood on GKE, with JAX 0.10.1+. Integration with Slurm clusters through Cluster Director remains in Preview.

## Sources

- [Cloud Blog](https://cloud.google.com/blog/topics/ai-infrastructure/cloud-pathways-a-distributed-ml-runtime-is-ga/)

---
Canonical: https://techandbusiness.org/newswire/kRgSsZtPYBRO1kzzRJFu8o
Published: 2026-11-09T13:38:14.416Z
Story chronology: 2026-11-09T00:00:00.000Z
Retrieved: 2026-11-09T16:56:48.257Z
Publisher: Tech & Business (techandbusiness.org)