Temporal Serverless Workers on Google Cloud Run entered preview, supporting queue-based autoscaling to zero instances, including workers with attached GPUs.

Google Cloud announced a preview integration in which Temporal Cloud, the managed version of the open-source Temporal execution platform, sizes Cloud Run worker pools according to task-queue workloads. Unlike request-serving applications, Temporal workers poll for work, and the integration does not require changes to existing worker code.

Workers run in the customer’s Google Cloud project, under its service account and on its network. Temporal Cloud monitors task queues and changes instance counts through the Cloud Run Admin application programming interface (API), using customer-granted identity and access management permissions.

Worker pools have no public endpoint, and their instances receive CPU allocation throughout their lifetime. Temporal does not initiate connections to the instances, so the integration does not require opening ports or adding firewall rules for Temporal access.

Each build version runs in a separate worker pool, allowing older versions to remain at zero instances until an associated workflow needs them. Google identifies AI-agent reasoning and tool-call workflows as use cases, with keyless access to Google Cloud AI APIs and optional attached GPUs for inference that retain scale-to-zero support.