Azure outages hit gateway services in 18 regions, with Azure OpenAI failing for six hours in Sweden

Microsoft Azure had a rough end to September. According to Shattered.io, two separate incidents between September 29 and October 1 created a streak of about 40 hours in which gateway services were degraded across 18 regions. Gateway services sit in front of customer workloads and route traffic, so failures there can cascade into every application behind them. The most visible AI impact was in Sweden, where Azure OpenAI customers saw nearly six hours of failed API calls.

Six hours of failed model calls is a long outage for production AI. Many enterprises use Azure OpenAI specifically for data-residency reasons: European customers choose Swedish or other EU regions so their data stays in the EU. That choice limits their failover options, because rerouting traffic to a US region could break their compliance commitments. Customers with strict residency requirements may have simply had no fallback for the duration.

The outages contrast sharply with Microsoft's market story. Microsoft stock just had its best quarter since 1998, up 37.7%, driven by Azure growth that hit 43% year over year with 45% guided for the next quarter. Microsoft 365 Copilot has passed 30 million paid seats and Agent 365 has nearly 40 million registered agents. Faster growth means more load and more dependence on Azure's reliability. As agents run continuously instead of on demand, uptime matters even more, because an outage stops work rather than just a chat reply.

The coverage comes from a single secondary source, and Microsoft's own post-incident reviews were not in the source material, so root cause and exact customer impact are unconfirmed. For architects, the lesson is familiar but sharper in the agent era: build multi-region or multi-provider fallback where compliance allows, and keep usable degradation paths when model APIs fail. Watch for Microsoft's official incident reports and whether it offers service credits.