What happened in the Microsoft 365 outage
On Thursday evening, 12 November 2026, Outlook stopped working for businesses around the world, and Teams, SharePoint and Copilot went with it. For many UK offices it ran into Friday. Here is what Microsoft has said so far, how much of the “pulled out by its own health checks” explanation going round social media holds up, and what a small business should do about it.
By Elliot · Published 13 November 2026 · Updated 20 November 2026 · 9 min read
In brief
- Exchange Online, the service behind Outlook, lost capacity worldwide from Thursday afternoon UK time.
- Microsoft says it was a technical fault, not a cyber attack. Its preliminary report traces it to a faulty fix for its gateway servers, rolled out from July.
- Customer impact ended at 16:10 UK time on Friday, Microsoft says. A final report is still to come.
- Check for missing email, warn the team about outage scams, and decide whether to claim a service credit.
Servers dropping out faster than they could be fixed
The first sign was people in the Americas unable to open their mailboxes. Within minutes Microsoft warned that any region could be hit, and by early evening in the UK, Downdetector was taking reports from 74 countries and territories. Outlook on the web, the desktop and phone apps, and anything else that reads a Microsoft 365 mailbox stopped connecting.
Microsoft's early updates explain why it spread. Traffic had not jumped. Instead, a large number of servers had become unavailable, so the ones still running took the extra load. Those servers were failing their automated health checks and being taken out of service, which left fewer still. When Microsoft found a way to bring some back, they stopped responding again as soon as traffic reached them.
Overnight the fault was narrowed to a process on the servers that sets up secure connections, which was being overwhelmed, and Microsoft began undoing a recent network configuration change that matched the start of the trouble. On Friday it told reporters the incident began with an issue during an earlier network capacity upgrade, which surfaced after that later change. Microsoft's preliminary post-incident report of 15 November traced the cause to a fix for its networking gateway servers, rolled out from July, that left them unable to cope under the heavy load of that afternoon.
| 12 Nov, 16:20 | The incident begins, by Microsoft's later report, and its first Service health post goes up under EX1550543. Its monitoring had first flagged connection problems at 16:03. |
|---|---|
| 12 Nov, 18:07 | Microsoft warns that users in any region could be affected, not just the Americas. |
| 12 Nov, 18:27 | A wider incident, MO1550821, opens for people unable to connect to several Microsoft 365 services at once. |
| 12 Nov, 19:20 | Microsoft says the extra load is not more traffic: a number of servers have become unexpectedly unavailable, straining the ones left. |
| 12 Nov, 19:43 | Microsoft says the affected servers are failing their automated health checks and dropping out of service as a result. |
| 12 Nov, 19:58 | Microsoft says it has stopped more servers being taken out of service while it investigates. |
| 12 Nov, 20:57 | A different way of checking server health is validated, and some servers return to service. |
| 12 Nov, 22:09 | Microsoft says its fixes so far have not given the relief it expected, and that a core process on the servers is being overwhelmed. |
| 13 Nov, 01:12 | The overwhelmed process is narrowed to the one that sets up secure connections. Microsoft starts reverting a recent network configuration change that lines up with the start of the problem. |
| 13 Nov, 03:34 | Microsoft says it has no indication that the outage is the result of malicious activity. |
| 13 Nov, 04:36 | Partial recovery reported across regions. |
| 13 Nov, 07:20 | Microsoft starts applying an update that reverts the faulty change, to part of its capacity first, then region by region. |
| 13 Nov, morning | Microsoft extends how long its systems keep trying to deliver queued email to 48 hours, so delayed messages are not returned to sender. |
| 13 Nov, 16:10 | Microsoft confirms customers are no longer seeing active impact, and moves to monitoring and clearing the email backlog. |
| 13 Nov, 16:40 | Microsoft apologises, reports broad recovery and stability, says the cause was technical and not a cyber attack, and promises a full post-incident report. |
| 14 Nov, 04:15 | After extended monitoring, Microsoft posts its final updates and closes both incidents. |
| 15 Nov | Microsoft publishes a preliminary post-incident report tracing the cause to a fix for its networking gateway servers rolled out from July. A final report is to follow. |
Did Outlook get pulled by its own health checks?
Partly. It is a fair description of the first few hours, and it is not the cause.
A health check is an automatic test a big service runs against each server, many times a minute. If a server fails, it is taken out of rotation so customers are not sent to a broken machine. Normally that is what keeps an outage small. On Thursday it worked the other way: so many servers failed at once that taking them out left too few to carry the load.
What Microsoft did say
Servers were failing automated health checks and becoming unavailable. The failures might come from the component that processes server health. It stopped further servers being removed, and brought some back by checking their health a different way.
What Microsoft did not say
That the servers taken out were healthy. Servers that came back stopped responding again once traffic reached them, because a process on them was overwhelmed: a real fault under load, not a false alarm. Nor did it blame the checks. Its preliminary report of 15 November puts the cause in a faulty fix for the gateway servers, rolled out from July, which choked them under heavy load.
Where the phrase comes from
Academics interviewed by ABC News in Australia and Al Jazeera described Microsoft's checks pulling servers out of service. It is a fair picture of the first few hours. It is commentary on the status updates, not Microsoft's explanation of the cause.
A global outage that reached the NHS and Ofsted
Downdetector counts problem reports from the public, not affected users. Microsoft has not said how many customers were hit.
74
countries and territories with Downdetector reports within about an hour
Source: Downdetector, via Al Jazeera
130k+
Outlook problem reports in the US alone at the peak, on Thursday afternoon US time
Source: Reuters
2
incidents to read in your admin centre: EX1550543 for email and MO1550821 for the rest
Source: Microsoft
In the UK, Downdetector reports climbed sharply from about a quarter to six. The Guardian reported hospitals and GP surgeries working around the outage, with pharmacists warning of prescription delays, while NHS England said local services were using their long-standing plans for IT outages. Ofsted could not receive email, and the Metropolitan Police asked people with time-critical but non-emergency reports to ring 101. The Bank of England said its critical operations were running normally.
Admin tools went down with the mail. As the incident grew, Microsoft added Intune, Purview, Power BI, Planner and its own admin centre to the list, so some IT teams also lost the tools they would normally use to manage devices and follow the incident. An early US estimate from Anderson Economic Group, reported by CNN, put the cost in lost work at around $7 billion. No method has been published, so treat it as a rough guide rather than a measurement.
Six things to check this week
The first two are for everyone. The middle two are for whoever looks after your Microsoft 365 account. The last two make the next outage hurt less, because there will be a next one.
Check the email you were expecting has arrived
Microsoft extended how long it keeps retrying queued mail to 48 hours, so most delayed messages should land late rather than vanish. Look for anything important that is still missing, such as a quote acceptance, an order or a supplier invoice, and ask the sender to send it again. Check your own sent items too: any bounce message from Thursday evening or Friday means that email never arrived and needs sending again.
Treat outage emails and calls with suspicion
Big outages bring a wave of scams: emails saying your account must be “re-verified” after the outage, fake Microsoft sign-in pages, and callers claiming to be IT support. Treat any message asking you to sign in again because of the outage as a scam until you have checked it another way. Remind the team today, and report anything odd rather than clicking it.
Admins: read the incidents in Service health
In the Microsoft 365 admin centre, go to Health, then Service health, and open EX1550543 and MO1550821. They hold Microsoft's own updates, and the full post-incident report will be attached there. While you are in, set Service health email notifications to go to an address that is not on your Microsoft 365 email, or the warning lands in the inbox that is down.
Decide whether to claim a service credit
Microsoft's service level agreement for Exchange Online promises 99.9% uptime each calendar month, which is about 43 minutes of downtime in a 30-day month. Below that a paid subscription can earn a credit of 25% of the month's fee, 50% below 99%, and 100% below 95%. Credits are not automatic: an admin has to claim them, with evidence, and for a November incident Microsoft must receive the claim by 31 December. If you buy through a reseller, ask them whether they will raise it for you.
Write a one-page plan B for the next one
Thursday showed how much a small business runs through one supplier. Agree now how staff reach each other if Teams and Outlook are down (a phone tree or a group chat on a separate service, kept to non-sensitive matters), who can still reach the bank and key customers, and where the most-used phone numbers are written down outside Outlook.
Make sure customers can still reach you
If your website's contact form only sends an email to a Microsoft 365 inbox, every enquiry during an outage waits in the same queue as everything else. A form that also keeps its own copy of each enquiry, and a phone number in plain sight on every page, mean a customer is never left wondering whether their message got through.
Where the accounts disagree
When it started. Microsoft's report puts the start of the incident at 16:20 UK time, with its monitoring first flagging problems at 16:03. Most coverage says around 18:00, which is when public reports peaked and Microsoft's wider incident was about to open. Both are right about different things: the fault came first, the flood of complaints about an hour and a half later.
Whether a change caused it. On Thursday evening Microsoft said it had found no recent change that matched the start of the problem. Overnight it began reverting a network configuration change that did, and on Friday it pointed to a capacity upgrade and a later network change. The Register noticed the shift too. The preliminary post-incident report of 15 November mostly settles it: the change was a fix for the gateway servers that had been rolling out since July without trouble, and only failed under that afternoon's load. Microsoft's final report may add detail.
The headline numbers. Figures such as hundreds of millions of users affected, or billions in lost work, come from report counts and outside estimates, not from Microsoft. Some coverage says NHS operations were cancelled; the reporting we could check mentions cancelled meetings and delays, not operations. Microsoft's preliminary report says most users were affected when using anything that relies on Exchange Online, but gives no count. We will update this page when its final report is published.
Microsoft outage questions, answered
What caused the Microsoft 365 outage on 12 November 2026?
Microsoft's preliminary post-incident report of 15 November says three things combined. A fix made specifically for the networking gateway servers in front of Exchange Online and other Microsoft 365 services, rolled out from July 2026, had a hidden defect: under certain loads, lock contention in a Windows process used up the servers' processor and memory. About fifteen hours before the outage, Teams had moved onto the same gateway, adding load, and nearly 14% of the gateway's capacity was already out of service after an unrelated issue the day before. Servers stopped responding, their traffic moved to the rest, and the failure cascaded until more than 80% were unresponsive. A final report is to follow. At the time, on 13 November, Microsoft had pointed to an earlier network capacity upgrade and a later network change.
Was the Microsoft outage a cyber attack?
Microsoft says not. Overnight it said it had no indication of malicious activity, and on 13 November it said the cause was a technical issue, not a cyber attack and not related to AI. Nobody credible has reported otherwise.
Is Outlook working again?
Yes. Microsoft's report says customers were no longer seeing active impact from 16:10 UK time on 13 November, after which it monitored the service and cleared the queued email, and it closed both incidents at 04:15 on 14 November. If anything still looks wrong for your organisation, check Service health in the admin centre, where EX1550543 and MO1550821 are kept in the incident history.
Were emails lost during the outage?
Mostly not. Email sent to you during the outage normally waits in a queue and is retried, and Microsoft extended that retry window to 48 hours. Messages can still bounce if a sender's system gave up sooner, so check for missing replies and for bounce messages in your own sent items.
Why did Teams fail if the problem was with email?
Microsoft listed Teams in the wider incident, MO1550821, alongside Exchange Online, SharePoint Online, Microsoft 365 Copilot and Defender XDR. Its preliminary report says Teams could move off the failing gateway, so Teams chats and meetings kept working without issue for most of the outage, while Exchange Online could not move for reasons of scale. The parts of Teams that rely on Exchange Online, such as the calendar, were affected with it.
Can we get money back from Microsoft for the outage?
Possibly, as a credit against your bill. Microsoft's service level agreement for Exchange Online gives 25% of the month's fee below 99.9% uptime, 50% below 99% and 100% below 95%. You have to claim it, with details of the downtime and the users affected, and for November the claim must reach Microsoft by 31 December. Free services are not covered.
Sources & further reading
The guidance on this page is grounded in primary sources. Follow the links to read them in full.
- Microsoft: Preliminary Post Incident Report for Microsoft 365 (EX1550543/MO1550821)
Microsoft's preliminary post-incident report, dated 15 November 2026: the cause, the timeline in UTC and the repair items. This is a public copy via next.ink; the original is in the Microsoft 365 admin centre under Service health, in the incident history.
- Microsoft 365 Status on X
Microsoft's public updates during the outage, including the admission on Thursday night that its fixes had not yet worked.
- ENow: Service health updates for MO1550821 and EX1550543
A copy of Microsoft's admin centre entries: the health checks, the overwhelmed connection process and the reverted network change.
- The Register: Microsoft struggling to fix massive cloud outage
Thursday night's account, including the contradiction over recent changes.
- BleepingComputer: Massive Microsoft 365 outage disrupts Outlook, Teams, Copilot
The services listed in each incident, the separate earlier Teams incident, and the 48-hour email retry.
- RTÉ: Microsoft says outage caused by network capacity upgrade
Microsoft's stated cause on 13 November, its apology and its promise of a post-incident report.
- ABC News: Microsoft outage affects users worldwide
Countries affected, expert comment on the health checks, and advice on outage scams.
- Al Jazeera: Microsoft hit by global outages, what happened and why
Downdetector's 74 countries and territories, and Microsoft's statement that it was not a cyber attack.
- Reuters, via The Korea Times: Microsoft outage drags on for hours
Downdetector's peak of more than 130,000 US Outlook reports.
- The Guardian, via AOL: Microsoft outage hits hospitals and GPs
The UK picture: NHS services, pharmacies, Ofsted, the Metropolitan Police and the Bank of England.
- SecurityWeek: Microsoft 365 outage disrupts Outlook, Exchange Online and Teams
Microsoft's overnight statement that it had no indication of malicious activity.
- Security Magazine: Microsoft outage could lead to $7B in losses
The Anderson Economic Group estimate first reported by CNN.
- Microsoft: Service Level Agreements for Online Services
The October 2026 SLA: Exchange Online's 99.9% commitment, the credit tiers and the claim deadline.