Resolved
Last updated: August 27, 2026 at 12:25 AM UTC
A100 hardware partial outage
Affected Components
Inference and Training A100 Hardware
Updates
Resolved
August 27, 2026 at 12:25 AM UTC
A100 hardware capacity has been restored and all queues are drained.
Monitoring
August 26, 2026 at 11:07 PM UTC
Monitoring started; pods are starting up again and working through the backlog.
Investigating
August 26, 2026 at 09:17 PM UTC
We're investigating a partial outage where most A100 hardware is automatically cordoned due to possible hardware failure.