When Microsoft 365 moved incoming traffic away from failing servers on November 12, the healthy servers taking over began to fail too. Microsoft's preliminary investigation traces the cascade to a shared network gateway. The worst impact was on Exchange Online, where many users lost mailbox access through Outlook or other clients and mail delivery was delayed.
The outage began on November 12, 2026, at 16:20 UTC. Microsoft's November 15 preliminary report, available through a public mirror, covers incidents EX1550543 and MO1550821. The Verge reported on the outage on November 17, two days after the preliminary report.
Why moving traffic spread the failure
Microsoft describes three conditions that came together. A change introduced in July had left a defect in servers operating the Microsoft 365 gateway. A routine Teams failover brought a different mix of traffic to the gateway shared with Exchange Online. At the same time, nearly 14% of that gateway's server capacity was already unavailable because of a separate issue.
Automatic rerouting sent requests to healthy servers as others stopped responding. Each move concentrated the traffic on fewer servers running the same gateway software. At the low point, more than 80% of servers on the affected gateway were unresponsive, Microsoft says. That figure measures gateway servers. The customer impact varied by region, connection method and time.
Mailbox access and delivery recovered in stages
Teams chats and meetings worked during most of the outage, and some other services could move away from the affected gateway. The scale of Exchange Online kept it on that gateway, Microsoft says. For people whose Outlook and browser sessions both led to the affected service, changing clients could keep them on the same failing path.
Microsoft applied a corrective update and brought groups of gateway servers back gradually. It says active customer impact ended at 16:10 UTC on November 13, after which it continued monitoring and clearing the mail backlog. The UK government told Parliament on November 16 that the outage disrupted health and other public services and that 999 remained operational.
Test the channel that carries the outage alert
For a team that depends on Exchange Online, a useful drill is to have staff locate the incident contact list, receive an alert and reply without opening Microsoft 365. Test the chosen channel with the normal email and sign-in paths unavailable. If email is part of an urgent workflow, also specify who confirms delivery once access returns. Opening a mailbox and receiving a delayed message are separate recovery checks.
Microsoft's preliminary report lists November 20 as the target for its final review and a fuller list of repairs. That review should help customers assess how Microsoft intends to keep rerouted traffic from producing the same cascade.