We have resolved an issue that caused some customers to experience delayed transcription in US West due to temporary GPU capacity constraints.
The issue affected Deepgram transcription capacity in us-west-2-prod. Service has returned to normal following capacity recovery and regional failover work.
We are continuing to monitor capacity and complete preventive measures, including earlier prescaling and additional regional failover options. No customer action is required.
We apologize for the disruption and appreciate your patience while we worked to restore service.
Posted Sep 24, 2026 - 09:53 PDT
Monitoring
Current monitoring shows that transcription latency has improved, and we are not seeing active customer impact at this time. We have taken mitigation steps, including adding capacity and routing workloads across regions, and are continuing to scale capacity to support stable processing. We will continue monitoring closely and provide another update if the situation changes.
Posted Sep 24, 2026 - 08:38 PDT
Investigating
We are investigating an issue affecting some Cresta customers in the US West 2 production environment where transcriptions may be delayed. Downstream features that depend on transcription may also be delayed or unavailable for some conversations.
Our engineering team is actively investigating and working to restore normal processing.
We will provide the next update within 60 minutes, or sooner if we have material progress.
If your team is experiencing this issue, please contact Cresta Support through the Director Support widget or at support@cresta.ai.
Posted Sep 24, 2026 - 07:24 PDT
This incident affected: Agent Assist - Voice (Live Transcript, Knowledge-base Search, Guided Workflow, Knowledge Assist, Auto Summarization).