Outages

A day without Outlook: what you can know before the cause is known

On 12 November Outlook, Teams and Microsoft 365 failed for users around the world, in an outage that ran across two working days. Microsoft’s account of the cause changed as it learned more. What a team can find out for itself in the first hour.

14 November 2026 · 3 min readAll posts

Illustration: a pile of unopened envelopes, the one on top stamped and addressed.

On Thursday 12 November 2026, shortly before one in the afternoon in New York, Microsoft said it was looking into reports of people who could not open their mailboxes. Within an hour more than 100,000 problem reports had reached Downdetector. It was the start of an outage that ran across two working days.

What happened

ABC News and CNBC followed it as it unfolded. Outlook email was hit hardest. Teams and other parts of Microsoft 365, among them SharePoint and Copilot, failed for some users as well, and Microsoft warned that users in any region could be affected.

On Friday morning, nearly 24 hours in, Microsoft said mail queues were “approaching normal levels”, while warning that access could still come and go and that mail could arrive late. It also said it would extend its email timeout window to 48 hours, so that delayed messages would not be lost. On Saturday it said the underlying problem was fixed and service was back for the remaining customers and regions.

An explanation in stages

The cause was not known at the start, and the account of it changed. At first Microsoft pointed to a surge of traffic, to heavy load in several regions, and to a fault in a component that tracks the health of its servers. By Friday it said the root was “an issue during an earlier network capacity upgrade”, which came to the surface after a later change to the network.

It said there was “no indication” of an attack, said it was “deeply sorry”, and promised a full report once its review is finished. When we wrote this, that report was not out yet.

The first hour is the hard one

None of this was known at one o’clock on Thursday. What every affected team needed then was simpler. Is it us, our network, or them? Is it everyone, or only some of our people?

Those questions have answers long before the cause does. They are questions about reach: from which countries, and over which networks, does the service answer right now?

What an outside check can and cannot see

  • It sees the door. A check from outside tells you whether a service’s address answers, from where, and how fast. When it stops answering from every country at once, the problem is not your office network.
  • It does not see inside. A sign-in page can load while the mailbox behind it does not. A reachability check will not catch that, and should not claim to.
  • It sees recovery where your people are. This recovery came region by region. “Restored” in a provider’s update is not the same as working again for your team in their country.

With Hopbits you can put a monitor on a public web address, including one that is not yours. For a host you have not verified it checks from datacenter nodes every five minutes, and tells you by email or webhook when a country stops getting an answer and when it comes back. For your own verified domains the checks can also run from home connections. It will not tell you why a provider failed. It will tell you, within minutes, that it is not just you.

Sources