The Microsoft 365 outage exposed a shared weak point for AI platforms that rely on Microsoft’s cloud services.
The disruption spread from workplace tools to AI assistants that retrieve business data. Users could lose access to email and calendars through those assistants even while the underlying models remained available. Microsoft continued to report problems early Friday as engineers worked to restore service.
Microsoft 365 connectors fail in ChatGPT and Claude
Anthropic said most email, calendar and search requests through Claude’s Microsoft 365 connector failed. Claude for Outlook also encountered problems.
OpenAI reported errors when ChatGPT users tried to create new Microsoft connections or use existing connector tools. Neither company reported an outage of its core AI models.
Outlook users began reporting connection problems shortly before 1 p.m. ET on Thursday, Nov. 12. Microsoft traced much of the disruption to services that rely on Exchange Online.
Teams chats and online meetings continued, but users had trouble scheduling meetings. Users could still open files directly from SharePoint.
Microsoft said Copilot Chat could fail to load or respond, while Microsoft Graph requests could time out.
The company also warned of delayed email and Safe Links verification failures. It listed disruptions across Defender, Intune, Purview, Planner, To Do and administrative tools. Some Power BI exports, Education features, Universal Print functions and Windows 365 Cloud PCs also experienced problems.
Failed health checks cut Exchange Online capacity
Microsoft said some servers failed automated health checks and dropped out of service, leaving fewer machines to handle requests.
Microsoft said there had been no overall surge in traffic across its network at the time. Engineers found that a process for establishing secure connections became overloaded and stopped servers from responding.
They suspected the component that handled health checks and were still investigating the trigger.
Microsoft paused deployments and configuration changes while engineers redirected Outlook connections away from affected U.S. infrastructure.
They redistributed traffic across the global network, restarted unresponsive servers and tested another health-check method. Some servers stopped responding again when traffic returned.
At 2:14 a.m. UTC Friday, Nov. 13, Microsoft’s dashboard still showed disruptions for businesses and consumers.
Microsoft was adding capacity as regions recovered and monitoring performance. At that point, it had yet to confirm full restoration or announce a permanent fix.
Cloud concentration and the cost of outages
Synergy Research Group put third-quarter cloud infrastructure spending at $168 billion, up 55% from a year earlier.
AWS held 26% of the market, Microsoft 19% and Google 16%, giving them a combined 61%. Gartner expects AI infrastructure spending to reach $2.4 trillion in 2027, or 59% of global AI spending.
Uptime Institute found that 57% of surveyed operators said their latest major outage cost more than $100,000.
One in five reported costs above $1 million. Third-party providers accounted for roughly two-thirds of publicly reported outages over nine years.
The UK’s competition watchdog flagged switching barriers and recommended strategic market status investigations into Microsoft and AWS. Cryptopolitan previously reported on a service that links AWS and Google Cloud over private networks. Such links can improve resilience across providers, but an AI connector still needs Microsoft 365 to retrieve Outlook data.
