If your Thursday felt strangely quiet, it wasn’t you. The Microsoft 365 outage that started on November 12 took Outlook, Teams, OneDrive, Copilot, Defender and the Intune admin console offline for a very large chunk of the planet, and kept them offline for most of a day. Not a blip. A full working day of nothing.
I’ve lived through my share of cloud incidents, but this one had a special flavour: Microsoft’s own engineers spent hours apparently not knowing what was wrong. The status page talked about “a large number of servers being unexpectedly unavailable”, which is the corporate way of saying “stuff is down and we’re looking”.
Microsoft 365 outage timeline
It began as a US and Americas problem on Thursday, then spread as the day went on. By the time Europe woke up on Friday, it was everyone’s problem. Doug Madory at Kentik, who watches internet traffic for a living, estimated about 18 hours of downtime before traffic started flowing back; other trackers counted past 27 hours for some tenants, with fixes that kept failing along the way.
Full restoration for all customers only landed on Friday night. Charles Lamanna, Microsoft’s EVP for Copilot, agents and platform, spoke of “encouraging recovery progress” and promised teams were “reviewing what happened”. Microsoft also went out of its way to say the incident was “not related to cyber or AI / agents”, a denial that only exists because people were very loudly speculating that an AI agent had gone rogue in Redmond.
It’s the second big one in a couple of months. Back in September a “core authentication configuration” issue locked users out for days. And if you remember the Outlook SPF validation mess from last week, you’ll agree Microsoft’s email stack has had a rough autumn.
Fax machines, coffee machines and delayed earnings
The anecdotes are where this story gets good. Around 78% of Japanese companies run on Microsoft, so Japan basically lost its office for a day. One worker wrote that the only task left was cleaning the coffee machine. Another company switched its daily PDF reports to fax, with the delighted comment “analog is strong”. Richemont, the group behind Cartier, had to postpone its earnings release because the people who publish it couldn’t talk to each other.
UK health services, airlines and transport operators were hit too, as TechCrunch reported, and so was TechCrunch’s own newsroom. When even the press covering the outage can’t send email about it, you know the dependency runs deep.
What the Microsoft 365 outage should teach you
Here’s my problem. Microsoft sells Defender as the thing that detects malicious activity, then during the outage it couldn’t say for hours whether the incident was malicious, because Defender was among the things that were down. A spokesperson didn’t answer when asked whether the company even had the logs to check. That’s not reassuring from the vendor that holds your email, your files, your identity provider and your device management in one basket.
I’m not going to tell you to self-host Exchange, I’m not a sadist. But I’d do three things this week. Make sure you have a second communication channel that doesn’t depend on Microsoft (a Signal group, a Matrix room, anything). Keep an offline export of the contacts and documents you can’t work without. And ask your provider for a credit; the SLA is there, and enterprise customers who don’t claim it are leaving money on the table.
Cloud concentration isn’t a theoretical risk anymore. We saw it with Yandex Cloud losing customer data after drone strikes, and we’re seeing a softer version of it every time Redmond sneezes. A company that just invited 60,000 employees to leave while pouring billions into AI data centers might want to keep a few people who know how the email servers work.
Eighteen hours of silence, and the loudest thing Microsoft said was “it wasn’t the AI”.
Discover more from TechPassion
Subscribe to get the latest posts sent to your email.





