# AI21 says automated GPU scheduling cuts priority job waits to 12 hours

_Published Friday, October 2, 2026 at 3:35 PM EDT · AI, Infrastructure · Latest · Tier 2 — Notable_

![AI21 says automated GPU scheduling cuts priority job waits to 12 hours — Primary](https://storage.googleapis.com/gweb-cloudblog-publish/images/07_-_Containers__Kubernetes_iY4YTLa.max-2600x2600.jpg)

AI21 says automated scheduling on Google Cloud cut waits for high-priority AI jobs from up to 72 hours to 12 hours, while leaving total compute cost unchanged. In a Google Cloud blog post, AI21 engineers described replacing manual capacity negotiations with Kueue, software that queues and admits workloads on a shared Kubernetes cluster.

The system prioritizes teams that have used less capacity without interrupting running jobs and checks whether workloads fit the physical cluster layout. AI21 says fragmentation fell from 15% to 8% and manual scheduling interventions dropped from about 20 per week to zero.

## Sources

- [Cloud Blog](https://cloud.google.com/blog/products/containers-kubernetes/ai21-trains-its-models-on-ai-hypercomputer/)

---
Canonical: https://techandbusiness.org/newswire/LS9fVbeW8o_bn951qlDK0q
Published: 2026-10-02T19:35:08.864Z
Story chronology: 2026-10-02T19:00:00.000Z
Retrieved: 2026-10-02T21:53:01.111Z
Publisher: Tech & Business (techandbusiness.org)