这篇文章还没有中文版,下面显示的是英文原文。
Microsoft 365 Outage Explained: Why Outlook, Teams and Copilot Went Down
On November 12 and 13, 2026, Outlook, Teams and Copilot failed in 74 countries. What broke, why the fix kept failing, and lessons for your systems.
On November 12 and 13, 2026, Outlook, Teams and Copilot went down for millions of people in 74 countries. Then Microsoft posted the words nobody wants to read from the people fixing it: the changes were "not taking as expected".
Microsoft had not published a root cause when I wrote this. Figures come from Microsoft's status posts, Downdetector and the news reports linked below.
▶ WATCH ON YOUTUBEWatch the video
▶ WATCH ON YOUTUBEFull video
Timeline
| When (US Eastern) | What happened |
|---|---|
| Thursday morning | Teams messages in the Americas fail for about an hour. Microsoft calls it resolved |
| Around midday | Outlook stops opening. Microsoft investigates problems reaching Exchange Online mailboxes in the Americas |
| 2:50 PM | Microsoft says servers are failing automated health checks and dropping out of service |
| About 5 PM | A core process on those servers becomes overwhelmed and they stop responding |
| 10:13 PM | "The changes were not taking as expected" |
| Overnight | Microsoft extends the email timeout window to 48 hours, so delayed messages are not thrown away |
| Friday | Microsoft rolls back a network change, without saying if it was to blame. Recovery begins, North America first |
Exchange Online is the engine behind work email. Outlook is the window you look through.
How far it spread
- Outlook reports in the US reached around 130,000 on Downdetector.
- Reports came from 74 countries and territories in about an hour.
- Microsoft's list: Exchange Online, Teams, SharePoint, Microsoft 365 Copilot and Defender XDR.
- Microsoft says its 365 apps have more than 430 million users.
Who got hit
- Japan: government staff made paper copies because ministries could not send email for most of Friday.
- Singapore: government email was down.
- Britain: police forces asked people to call 101 and not use online forms.
- Germany: Deutsche Bahn reported big delays sending and receiving email.
- Hong Kong: bankers told CNN they could not open email or Teams.
- AI companies: OpenAI said some Microsoft connectors were not working.
What broke
Microsoft's first answer was elevated load. Then it corrected itself: the load was not from more traffic. It came from a large number of servers going unexpectedly unavailable.
- Servers fail automated health checks and are pulled out of service.
- Their work lands on the servers still standing.
- Those start to buckle too.
Think of a smoke alarm that keeps going off with no fire. Every false alarm pulls another server off the line. Microsoft later said the failures may come from one component that processes server health, and that it found no recent change linked to the start.
Why the fix kept failing
Microsoft paused all deployments and restarted every affected server. Servers came back and fell over again as soon as traffic hit them.
One analyst explained it to Al Jazeera: a server that comes back is at once swamped by all the other servers trying to talk to it. Another compared it to a four-lane highway that suddenly has two lanes.
Microsoft executive Charles Lamanna said there was no indication of malicious or cyber activity.
Lessons for anyone who runs a service
These are my takeaways from Microsoft's own description. The root cause is not published, so treat them as general lessons, not a diagnosis.
- Health checks can cause an outage. A check that wrongly fails healthy servers removes capacity exactly when you need it. Put a limit on how much of a fleet automation may remove at once.
- Plan for the restart, not only the failure. A cold server that receives full traffic at once falls over again. Bring capacity back gradually and shed load while you do it.
- Retries multiply load. Clients and peers that all reconnect at the same moment create a second wave. Back off, and add randomness to the delay.
- Know your minimum viable service. What must keep running when your cloud provider goes dark? Email, chat and identity are often the same vendor.
- Have a second channel. If your incident plan lives in the tool that is down, you have no plan.
What you should do as a user
- Watch for phishing. "Reset your password" messages are what scammers send after an outage.
- Be patient. Delayed email should still arrive. The 48-hour window exists for this.
- If you run a business, write down what has to keep working without the cloud.
This was not the first time. In January 2026, Microsoft 365 went down across North America for almost 10 hours.
Sources
- Full video: https://youtu.be/I-asafFasPc
- Microsoft 365 Status on X: https://x.com/MSFT365Status
- Al Jazeera: https://www.aljazeera.com/news/2026/11/13/microsoft-hit-by-global-outages-what-happened-and-why
- BleepingComputer: https://www.bleepingcomputer.com/news/microsoft/massive-microsoft-365-outage-disrupts-outlook-teams-copilot-worldwide/
- CNN: https://www.cnn.com/2026/11/12/business/live-news/microsoft-global-outage-intl-hnk
Would you switch away from Microsoft after this?
Not affiliated with Microsoft.


