Observing improvements - Intermittent request failures and increased latency across Azure OpenAI, Azure AI Foundry, and Cognitive Services

Impact

Impact Statement: Starting at 10:00 UTC on 29 September 2026, a subset of customers using OpenAI Service, Foundry Agent Service, Foundry Models, and Cognitive Services, in Sweden Central may experience intermittent request failures, increased latency, and HTTP 5XX error codes when submitting requests to affected models and APIs hosted in this region. Current Status: Currently we are monitoring improvements in the overall experience. Initial findings indicate that a backend service which is relied on to retrieve metadata information when managing requests for this region, was experiencing timeouts when reading from dependent database and caching layers. Additionally, some instances for this backend service became unhealthy and experienced restarts. As a result, requests that depend on these processes were delayed or failed intermittently. We have taken steps to further scale instances for the backend service, and reconfigured system processes as mitigation steps. Queued requests have been processed for a period of time to confirm improvements to overall service health. We are continuing to investigate the health of supporting platform infrastructure and system processes that may have impacted these requests and the backend service. The next update will be provided within 60 minutes, or as events warrant.

Affected service

Affected regions

Timeline

Updates newest first (times in UTC).

1. MonitoringSep 29, 2026, 10:00 AM UTC

   Impact Statement: Starting at 10:00 UTC on 29 September 2026, a subset of customers using OpenAI Service, Foundry Agent Service, Foundry Models, and Cognitive Services, in Sweden Central may experience intermittent request failures, increased latency, and HTTP 5XX error codes when submitting requests to affected models and APIs hosted in this region. Current Status: Currently we are monitoring improvements in the overall experience. Initial findings indicate that a backend service which is relied on to retrieve metadata information when managing requests for this region, was experiencing timeouts when reading from dependent database and caching layers. Additionally, some instances for this backend service became unhealthy and experienced restarts. As a result, requests that depend on these processes were delayed or failed intermittently. We have taken steps to further scale instances for the backend service, and reconfigured system processes as mitigation steps. Queued requests have been processed for a period of time to confirm improvements to overall service health. We are continuing to investigate the health of supporting platform infrastructure and system processes that may have impacted these requests and the backend service. The next update will be provided within 60 minutes, or as events warrant.

Related reading