Skip to content
Komposite News
Markets · Technology · Capital
Thursday, September 10, 2026Global EditionGot a tip?
Cloud

Microsoft moves its outage start six hours earlier

Microsoft's closed record of the Exchange Online outage now runs from 10:05 GMT on Thursday to 16:10 GMT on Friday, 30 hours and 5 minutes. It blames a network capacity upgrade. Our arithmetic shows the SLA credit tier does not change.

Tom Kessler · House byline
November 14, 20264 min read
A fenced data-centre site with low grey buildings behind a black security fence, an access road and a give-way sign in the foregroundJohan Wessman / News Oresund, CC BY 2.0, via Wikimedia Commons
The fenced site of a Microsoft data centre under construction at Staffanstorp in southern Sweden, with a give-way sign on the access road, in a file photograph from 2021.Photograph via Wikimedia Commons

Microsoft's closed record of this week's Exchange Online outage puts its start at 10:05 GMT on Thursday 12 November, six hours and 12 minutes earlier than the time the incident carried while it was open, and its end at 16:10 GMT on Friday 13 November. On those times the outage lasted 30 hours and 5 minutes.

The revised times appear in Microsoft's service-health record for incident EX1550543, as copied from the University of Pennsylvania's Microsoft 365 console, which marked it restored in a final update at 04:15 GMT on Saturday. The companion incident covering other Microsoft 365 services, MO1550821, was given the same start and end. When Komposite read the record on Friday, EX1550543 began at 16:17 GMT and MO1550821 at 18:26 GMT. Microsoft has not said why it moved the start.

For the first time Microsoft also gave a cause. "The incident was introduced by an issue during an earlier network capacity upgrade that surfaced following a recent network change," it wrote, first at 17:57 GMT on Friday and again in the final status. "This was a technical issue not related to cyber or AI / agents." The same update apologised: "We are deeply sorry for the disruption this incident has caused you and your business."

The SLA arithmetic, redone

The earlier start does not change the credit tier for a user who was out from beginning to end. Under Microsoft's Online Services service-level agreement, Exchange Online downtime is "Any period of time when users are unable to send or receive email with Outlook Web Access", and credits step up at 99.9, 99 and 95 per cent monthly uptime. Thirty hours and 5 minutes is 1,805 minutes, which leaves 95.82 per cent of a 30-day November, by our arithmetic. That is below 99 per cent, worth a 50 per cent credit on the month's Exchange Online fee, but above the 95 per cent line that would have meant a full refund.

Komposite's arithmetic on a 30-day month of 43,200 minutes, for a single user unable to send or receive mail in Outlook on the web for the whole period, using the times in Microsoft's service-health record and the tiers in its Online Services SLA. Microsoft measures downtime in user-minutes across a tenant, so most organisations will land in lower tiers.
BasisMinutes outUptime for that user in NovemberCredit tier under the SLA
Start as first logged (16:17 GMT Thursday) to the final end time1,43396.68%Below 99%: 50% of the month's fee
Start as now recorded (10:05 GMT Thursday) to the final end time1,80595.82%Below 99%: 50% of the month's fee
Needed for the 100% tier (below 95%)2,16095.00%Would have required an outage until 22:05 GMT Friday

On Friday we calculated that a user out since the old start would cross the 95 per cent line at 04:17 GMT on Saturday if the outage continued. It did not; on the closed record the full-refund threshold would have needed the outage to last until 22:05 GMT on Friday. Microsoft's agreement also says credits are "your sole and exclusive remedy", and customers must claim them by the end of the month after the incident, so by 31 December for most monthly subscriptions.

Restored on paper, monitored for 12 hours

The recorded end time came before the updates stopped saying recovery was incomplete. At 20:02 GMT on Friday, almost four hours after the 16:10 end time, Microsoft wrote: "We do not have an ETA for full recovery yet. Service access can recover before mail flow fully normalizes." At 22:00 GMT it said: "Due to the severity of impact this incident caused, we're monitoring for an extended period." The final status came 12 hours after the recorded end. By our count, EX1550543 ran to 41 updates and MO1550821 to 37.

The final list of affected services is long. MO1550821's closing text names Exchange Online, the Microsoft 365 admin centre, Microsoft 365 apps, the Apps admin centre, Microsoft 365 Education, Copilot, Defender for Office 365, Defender XDR, Fabric, Graph, Intune, Planner and Power Apps among others, and repeats that "Elements of those services that do not depend on interaction with Exchange Online remained functional." For Copilot, Microsoft said users "may be unable to load Copilot Chat or receive responses", and that its experiences may be unable to reach work email, calendars and files, a reminder of how much of the company's AI product depends on the mailbox.

Read together, the explanations have moved since Thursday. The early updates spoke of servers failing health checks and an overwhelmed connection process; early on Friday Microsoft said it was reverting "a recent network configuration change"; the final status links a recent change to an earlier capacity upgrade. That sequence is consistent with an investigation narrowing, but the final status is still labelled a "Preliminary root cause".

When does the full report arrive?

Microsoft's final status promises "a preliminary Post-Incident Report within two business days and a final Post-Incident Report within five business days". Counting from Friday's end, that means Tuesday 17 November and Friday 20 November, by our calendar. Microsoft's published service description is stricter on the first of those: a preliminary report "within 48 hours of incident resolution", which would have been 16:10 GMT on Sunday.

The report will be read for two things this record cannot answer: why the start moved to 10:05, more than six hours before the first public update at 16:20 GMT, and which customers and regions were affected in those hours. Komposite's account of the outage while it was running, and of how the SLA tiers apply, appeared on Friday. Organisations weighing how much mail to keep in a single provider's cloud have made different choices.

MORE ON THIS STORY