The cause of the 12–13 November Microsoft 365 outage, per Microsoft's preliminary report, was a defect that had been running on its gateway servers since July 2026 and only broke out under extra load. More than 80% of those servers went down, Outlook was disrupted for almost 24 hours, and a quick rollback wasn't possible.
Key facts
- Per Microsoft's preliminary report of 15 Nov 2026, the Microsoft 365 outage ran from 12 Nov, 16:20 UTC, to 13 Nov, 16:10 UTC (incident EX1550543).
- The cause was a change to the gateway servers rolled out from July 2026, triggered by redirected Teams traffic and nearly 14% missing capacity.
- More than 80% of gateway servers were unresponsive in under two hours; Microsoft says the final report is due by 20 Nov 2026.
Researched and written with AI, to Luigi's brief, positions and editorial rules. Automatically checked against the sources. How I work →
You know the question after every outage: "What did we change last?" Microsoft just showed why that question is too short. The cause of almost 24 hours without Outlook mailboxes wasn't last week's change. It had been running since July. Quietly. Until the load showed up.
What caused the Microsoft 365 outage?
According to Microsoft's preliminary report dated 15 November (incident EX1550543/MO1550821), as covered by TechCentral and Office Watch, the defect sat in a fix built for the networking gateway servers in front of many Microsoft 365 services. Rolled out from July 2026, it ran worldwide for more than three months without impact. Under certain loads, internal Windows tasks got stuck waiting on each other until the servers ran out of processing power and memory.
Two things that were harmless on their own triggered it:
- About 15 hours before the outage, automation moved Teams traffic onto the gateway that also carries Exchange Online.
- Nearly 14% of the gateway's capacity had been offline since the day before because of an unrelated problem.
Then it went fast: in under two hours, more than 80% of the gateway servers were unresponsive. The regional bulkheads meant to contain this didn't hold. Exchange Online, meaning Outlook mail, took the worst of it, plus parts of Copilot that talk to Exchange, per the report. The same link that Claude and ChatGPT connectors depended on, as I wrote in the Microsoft outage and your AI agent's inbox.
Why no test caught it
This is the real story for me. Load tests ran on weaker machines with fewer cores. The defect gets worse on newer hardware with more cores. So the lab tested a system that didn't exist in production.
And rolling back? Not quick. Rollback, flighting and feature flags are built for fresh changes. For a three-month-old change, the fix was a new update restoring the old known-good code, server group by server group. Europe was back to normal around 12:35 CET on Friday, per TechCentral.
Same pattern as the FAA a few days ago: the new AI tool was the suspect, the old network did it. The most dangerous change is the one nobody remembers making.
What this means for your AI automations at Black Friday peak
In my view, this is a wake-up call for every marketing team running AI workflows. The agents, Zapier chains and prompt templates you built over the summer have never seen real peak load. Black Friday and Cyber Monday are exactly the mix of more traffic and less headroom that woke Microsoft's sleeper. AI doesn't fix your processes. It widens the cracks that are already there.
What I'd do now
- List every change to your automations since July, not just since last week. With date and owner.
- Save a known-good version of each critical workflow (export, prompt version, connector settings), so going back doesn't mean rebuilding.
- Run one test at peak volume on the real setup. And agree on a manual plan B for mail and approvals. More on the setup in the guide marketing automation with AI.
For admins: per TechCentral, SLA credits aren't automatic and must be claimed by 31 December. Check your own contract terms in the admin center. Microsoft says the final report is due by 20 November.
Bottom lineIt's not yesterday's change that takes you down. It's the one from three months ago that has never seen peak load.
Got AI news? Submit news →
FAQ
What caused the Microsoft 365 outage in November 2026?
Per Microsoft's preliminary report, a defect running since July 2026 on its networking gateway servers, triggered by redirected Teams traffic and nearly 14% missing capacity.
How long did the outage last?
Almost 24 hours: from 12 November, 16:20 UTC, to 13 November, 16:10 UTC. Exchange Online, meaning Outlook mail, was hit hardest.
Is there a service credit for the outage?
Not automatically, per TechCentral: admins must claim SLA credits by 31 December. The exact terms are in your own contract.
Source
TechCentral: Microsoft 365 outage: inside the 24 hours that broke email
More sources
- Office Watch – 16.11.2026: Vorfall-IDs EX1550543/MO1550821, 16:20 UTC 12.11. bis 16:10 UTC 13.11., Teams-Umschaltung ~15 h vorher, >80 % in unter zwei Stunden, Copilot-Teile mit Exchange-Zugriff betroffen, Rollback/Feature Flags griffen nicht
- i4u KI-News (13.11.2026) – Früherer Artikel zum Ausfall selbst




