Microsoft's mail outage crosses the SLA's 99 per cent line
Exchange Online has been degraded since 16:17 GMT on Thursday, and Microsoft has posted 34 updates without naming a root cause. Our reading of its status record and its service-level agreement, and what the hours mean for customers' credits.
Jeff Wilcox, CC BY 2.0, via Wikimedia CommonsMicrosoft has posted 34 updates on incident EX1550543 since it opened at 16:17 GMT on Thursday 12 November, according to the service-health record that the University of Pennsylvania copies from its Microsoft 365 administration console every 15 minutes. At the 14:01 GMT update on Friday the incident was still a "service degradation", its end time "Not ended yet" and its root cause "N/A".
The Microsoft 365 outage that began on Thursday 12 November 2026 has left users unable to reach Exchange Online mailboxes through Outlook on the web, Outlook desktop, mobile sync and programming interfaces for more than 21 hours. By Friday Microsoft listed 20 dependent services as affected, and said it had found no evidence of malicious activity.
Recovery is under way but not complete. At 13:00 GMT Microsoft reported "broader improvements now observed in North America and sustained improvements across much of Europe"; at 14:01 GMT it said email queues were "significantly reduced and approaching normal levels", while warning that some users may still see intermittent access problems or delayed mail. It said it would publish "a full Post-Incident Report with root cause and defense-in-depth actions after mitigation is complete". Microsoft has also been posting updates on its public status page for customers without console access.
What has Microsoft said went wrong?
Several things, in sequence. Read in order, the updates move from a load problem to failing servers, then to a component handling health checks, then to an overwhelmed process that sets up secure connections, and finally to a patch for an authentication and routing component that began to stabilise servers on Friday morning.
| Time (GMT) | What Microsoft's status update said |
|---|---|
| Thu 16:17 | Incident EX1550543 opens: users may be unable to reach Exchange Online mailboxes |
| Thu 18:19 | The affected infrastructure is under "elevated load across multiple regions" |
| Thu 19:18 | The load is not more traffic but "a large number of servers being unexpectedly unavailable" |
| Thu 19:43 | Affected servers are "failing automated health checks", possibly after a recent change to a component |
| Thu 19:57 | "While we haven't identified any recent changes that correlate with the start of impact", a health-check component is bypassed on some servers |
| Thu 22:00 | Actions "have not provided the anticipated relief"; "a core process on the affected servers is becoming overwhelmed" |
| Fri 01:06 | The process "responsible for establishing secure connections" is overwhelmed; Microsoft begins reverting "a recent network configuration change that correlates with the start of impact" |
| Fri 02:54 | A change is "not taking as expected" |
| Fri 03:34 | "There is no indication this is the result of malicious activity" |
| Fri 04:59 | "A surge in outbound call volume from our routing services to our backend servers" |
| Fri 07:25 | Server stability improves after an "upstream authentication/routing configuration component patch" |
| Fri 09:55 | Email timeout window extended "to 48 hours to help protect delayed messages" |
| Fri 14:01 | Email queues "significantly reduced and approaching normal levels"; still not mitigated |
Two updates sit uneasily together. At 19:57 GMT on Thursday Microsoft said: "While we haven't identified any recent changes that correlate with the start of impact". At 01:06 GMT on Friday it said it had "begun reverting a recent network configuration change that correlates with the start of impact". The second does not say whether that change caused the outage, and investigations of this kind often revise early findings. The post-incident report should settle it. Under Microsoft's own service description, a preliminary report is due within 48 hours of resolution and a final one within five business days.
The mail outage was Thursday's second Microsoft 365 incident. A separate one, TM1550488, disrupted Teams messaging for users routed through North and South America from 14:45 to 15:53 GMT, ending 24 minutes before the Exchange incident opened. Microsoft's preliminary root cause for that one was "Resource contention within the underlying service infrastructure", and it promised a preliminary report within two business days. The two records do not refer to each other.
Which services were hit?
Anything that depends on Exchange Online. A companion incident, MO1550821, opened at 18:26 GMT on Thursday naming five services. By our count of each update, the list grew to 20 by 06:54 GMT on Friday, taking in Copilot's access to mail and calendars, Intune device enrolment, Purview compliance tools, Universal Print and some Windows 365 cloud PCs. Microsoft says elements that do not interact with Exchange Online "remain functional", and that Teams calls, meetings, group chats and channel messages were largely unaffected.
| Update (GMT) | Services listed | Added in that update |
|---|---|---|
| Thu 18:27 | 5 | Exchange Online, Teams, SharePoint Online, Copilot, Defender XDR |
| Thu 19:58 | 8 | Planner, Purview, To Do |
| Thu 20:57 | 11 | Defender for Office 365, Intune, Power BI |
| Thu 22:03 | 13 | Microsoft 365 Apps admin center, Universal Print |
| Thu 22:58 | 14 | Windows 365 (Cloud PCs using Customer Key) |
| Fri 00:00 | 18 | Microsoft 365 Education, Microsoft 365 admin center, Microsoft 365 apps, Graph |
| Fri 01:12 | 19 | Fabric |
| Fri 06:54 | 20 | Power Apps |
Microsoft released its monthly security updates on Tuesday 10 November. SecurityWeek noted that Microsoft had not linked an Outlook flaw fixed in that release to the outage. October's release, by Microsoft's own count, covered 1,453 vulnerabilities.
What does the outage mean for service credits?
For a user who could not send or receive mail in Outlook on the web for the whole incident so far, uptime for November would already be about 97.0 per cent, by our arithmetic, which is inside the band where Microsoft's service-level agreement pays a credit of half the month's Exchange Online fees. The October 2026 edition defines Exchange Online downtime as "Any period of time when users are unable to send or receive email with Outlook Web Access", measured in user-minutes across a calendar month.
| Monthly uptime for Exchange Online | Service credit | Full-outage minutes that reach it in November | For a user out since 16:17 GMT Thursday, reached at |
|---|---|---|---|
| Below 99.9% | 25% of the month's Exchange Online fees | 43.2 (about 43 minutes) | 17:00 GMT Thursday |
| Below 99% | 50% | 432 (7 hours 12 minutes) | 23:29 GMT Thursday |
| Below 95% | 100% | 2,160 (36 hours) | 04:17 GMT Saturday, if still out |
Three conditions narrow that. Downtime is averaged across every user in a tenant, so an organisation where only some mailboxes failed, or where service returned early, will sit higher. Credits are not automatic: a customer must file a claim with the time, duration, number and location of affected users, and for an incident in November the claim must arrive by the end of December. And credits are "your sole and exclusive remedy", applied only to the fees for the affected service; the agreement rules out compensation for "lost revenue, operational costs, or any indirect losses". If the incident runs past 04:17 GMT on Saturday 14 November, a user out for the whole period would cross the 95 per cent line that triggers a full refund of that month's fee.
Why does the 48-hour mail window matter?
It decides whether delayed mail is delivered or bounced. Microsoft's documentation says Exchange Online retries delivery for 24 hours, after which messages "will expire and will be returned to the senders in non-delivery reports". At 09:55 GMT on Friday Microsoft said it was "extending the email timeout window to 48 hours to help protect delayed messages". If the clock runs from when a message is first queued, which Microsoft has not specified, mail held since 16:17 GMT on Thursday would now bounce no earlier than Saturday afternoon rather than Friday afternoon.
The record also shows the pace of communication. Microsoft's service description promises updates "on an hourly cadence or, if a different cadence is required, it will be stated". Each update named the time of the next one, and by our comparison three of those promised times were missed by more than five minutes: by six minutes at 01:00 GMT, and by 55 and 61 minutes as Europe's working day began on Friday, when the next updates came at 09:55 and 11:01 GMT.
The outage lands in a year when Microsoft 365's availability has already dipped. Microsoft's published worldwide uptime for the first quarter of 2026 was 99.526 per cent, the lowest figure in its table, equal to about 614 minutes of a 90-day quarter by our arithmetic; the second quarter recovered to 99.994 per cent. Some public bodies were reconsidering their dependence before this week: the Dutch tax administration said in October that it would keep its mail on its own servers rather than move to Microsoft 365, part of a broader pattern we have described as selective cloud repatriation.
Microsoft's next scheduled update on the Exchange incident was due at 15:00 GMT on Friday. The fourth-quarter uptime figure, which will include this week, is the number customers will be able to set against the 99.9 per cent the agreement promises.