How the Microsoft 365 Outage Happened: A Hidden Network Fault, a Health-Check Cascade and a 38-Hour Recovery (Video)

Microsoft 365 was fully restored at 10:22 Mauritius time (06:22 UTC) on 14 November, about 38 hours after a latent fault from a network capacity upgrade triggered a health-check cascade; our video shows how it unfolded.
Watch the Video
The Microsoft 365 outage that began on Thursday, 12 November 2026 is now over. At 10:22 Mauritius time (06:22 UTC) on Saturday, 14 November, Microsoft said it had addressed the underlying issue and restored service to all remaining customers and regions. That is roughly 38 hours after the first Exchange Online failures.
The video above walks through the whole incident in about three and a half minutes: how Exchange Online keeps itself healthy, the hidden fault that started it, the cascade that followed, how far the damage spread, and how Microsoft fought its way back. The server counts in the animation are illustrative, not real figures.
What Happened, in Short
- 20:17 (16:17 UTC), Thursday 12 November: the Exchange Online incident EX1550543 begins, with mailbox connection failures for users in the Americas.
- 22:26 (18:26 UTC): Microsoft opens the wider Microsoft 365 advisory MO1550821 as the impact spreads to other regions and services.
- Overnight into Friday 13 November: Microsoft pauses deployments, moves traffic, restarts servers, and reverts a recent network change that coincided with the start of the problem.
- Friday: service recovers in phases across Asia Pacific, Europe and North America. Microsoft extends the email delivery timeout to 48 hours so delayed messages are not lost, and by 21:06 (17:06 UTC) says email queues are significantly reduced.
- 10:22 (06:22 UTC), Saturday 14 November: Microsoft says the underlying issue is fixed and service is restored for everyone.
Why It Happened
Microsoft's updates describe a failure that started small and fed on itself.
- A hidden fault. Microsoft said the incident was "introduced by an issue during an earlier network capacity upgrade". The problem stayed hidden until a more recent network change brought it to the surface.
- Health checks pulled servers out. Exchange Online servers began failing their automated health checks. The system did what it was designed to do and removed them from service, so every server that dropped out pushed its share of the work onto the servers that were left.
- Not a traffic surge. Microsoft stressed that the load was not driven by an increase in traffic. Demand stayed normal while healthy capacity shrank.
- The survivors were overwhelmed. On the remaining servers, a core process, which Microsoft described as the one responsible for establishing secure connections, became overwhelmed and processor use rose sharply. Those servers stopped responding too, and the failure cascaded.
- Not an attack. Microsoft said: "This was a technical issue not related to cyber or AI / agents."
Microsoft has promised a full post-incident report. Until it is published, the details above are Microsoft's preliminary explanation.
What Was Affected
Because so much of Microsoft 365 depends on Exchange Online, the outage spread well beyond email:
- Email and collaboration: Exchange Online, Outlook, Teams, SharePoint Online, OneDrive
- AI: Microsoft 365 Copilot
- Security and management: Defender XDR, Defender for Office 365, Intune and Purview
Features that did not depend on Exchange Online generally kept working.
How Microsoft Fixed It
- Froze deployments and configuration changes so nothing new could make things worse
- Moved traffic away from struggling systems and restarted affected servers
- Tried an alternative way of processing health checks, which brought some servers back. Some of them became unresponsive again as soon as traffic returned
- Rolled back the recent network change
- Added an automated process to find and restart servers that had stopped responding
- Applied a patch across its infrastructure, then restored service gradually and checked stability at each stage
Attackers Moved Fast
Microsoft found no sign that the outage was malicious, but criminals used it within hours. Microsoft Threat Intelligence warned of a phishing campaign that impersonates Microsoft 365 service health notices and references the real incident EX1550543.
- The lure: emails claiming that messages were "held during the outage" and would be deleted within 24 hours unless the recipient clicked to restore them
- Subject lines: "Outage - " followed by a six-character code
- Sender names: an organisation-specific prefix followed by "Outage Notification" or "Service Health"
- Sender domains: moments-notice[.]com, planetmaison[.]com and newbooker[.]com
Microsoft said the activity is consistent with credential phishing but has not confirmed that. Real outage updates appear in the Microsoft 365 admin center under Health > Service health. Microsoft does not ask users to click an email link to "recover" mail.
What Admins Should Do Now
- Hunt for the phishing emails. Search for the sender domains and the "Outage - " subject pattern in Defender for Office 365 Threat Explorer, and check the EmailEvents, EmailUrlInfo and UrlClickEvents tables in Defender XDR advanced hunting for anyone who clicked.
- Warn users. No outage email needs a click to restore messages. If someone clicked and entered a password, reset it and revoke their sessions.
- Check delayed mail. Messages queued during the outage should now be delivered. Confirm that important mail flows, such as invoices, alerts and approvals, arrived.
- Review delayed security signals. Look for Defender alerts and Intune policy syncs that may have been held up while the service was degraded.
- Claim SLA credits if you qualify. Business customers can request a service credit, quoting EX1550543 for Exchange Online or MO1550821 for the wider incident.
Lessons
- Safety mechanisms can amplify a failure. Health checks that act on bad signals can remove capacity exactly when it is needed most.
- Latent faults hide. A problem from one change can stay invisible until a later change exposes it.
- Headroom beats raw scale. Normal demand on shrinking capacity is enough to start a cascade.
- Every big outage becomes a phishing lure. Prepare user warnings in advance so they can go out within the hour.
Resilience isn't about never failing. It's about failing small.
