The outage began on Thursday 12 November, at around 11:17 a.m. Eastern time, in Exchange Online, the system that carries email for Microsoft 365. Microsoft acknowledged the problem at 5:54 p.m. GMT, warning at first that users in the Americas could not reach their mailboxes. Within fifteen minutes the warning had widened to Microsoft 365 users “in any region”. A second incident reference, MO1550821, followed for the wider suite, alongside EX1550543 for Exchange Online itself.
The numbers piled up fast. Outage tracker Downdetector logged more than 140,000 problem reports for Outlook alone on Thursday afternoon, from users in 74 countries. Across all Microsoft services, Downdetector collected more than 7 million reports by Friday morning. Microsoft 365 has more than 450 million paid commercial seats, and the failure reached far beyond office workers checking email.
Staff at the US House of Representatives lost what one aide called “a day of legislative work”. Courts in Los Angeles and Texas warned of delays. Hospitals in Britain rescheduled procedures. Japan’s Digital Agency could not receive email, and Immigration New Zealand paused visa processing. Baltimore County and the city of Austin in Texas told residents their email was down. The outage hit governments, schools, hospitals, airlines and courts in at least a dozen countries.
What broke — and what kept working
Broken or degraded: Outlook and Exchange Online mailboxes (web, desktop and mobile), Teams calendar functions, Microsoft 365 Copilot, SharePoint-connected features, OneDrive, OneNote, Defender, Planner, To Do, Intune, Purview and Power BI. Microsoft’s own list of affected services ran to more than a dozen.
Still working: Microsoft 365 services that do not depend on Exchange Online or email kept running. The company’s Azure cloud hosting business was not part of the incident, even though complaint trackers showed a spike of reports for Azure too.
What Microsoft says caused it: A number of servers became “unexpectedly unavailable” during automated checks, and a process responsible for establishing secure connections became overwhelmed and stopped servers processing requests. The company later blamed a problem that surfaced during an earlier network capacity upgrade. It has stressed, repeatedly, that there is no evidence of malicious or cyber activity. Charles Lamanna, the executive vice president who runs Copilot and Microsoft’s platform business, gave the same assurance in public. A full post-incident report is still to come.
The slow road back
Microsoft’s first fix was to restart the affected servers. It did not hold: the servers went unresponsive again as soon as traffic returned. Engineers then deployed a patch aimed at the authentication and routing component behind the failure, rolling it out region by region, starting with the places entering peak business hours. By the early hours of Friday, Microsoft said it was “seeing positive results for server stability”. Parts of Asia Pacific recovered fully on Friday afternoon.
To protect mail caught in the jam, Microsoft extended the email timeout window to 48 hours, so delayed messages keep retrying instead of bouncing back to senders. By Saturday morning, Microsoft confirmed the services across the affected environments had fully recovered and the backlog of delayed email had been processed, and the incident was remediated — though a full post-incident report explaining the root cause is still to come.
Why this matters
Two things made this outage different from the routine service blips that tech companies absorb every month. The first is the sheer concentration of risk: email, calendars, documents, chat and AI assistance, all routed through one company’s infrastructure, all failing together. When a single vendor carries the working day of governments and hospitals, a bad day for that vendor becomes a bad day for everyone.
The second is the timing. It arrived in the same week as the FAA’s nationwide ground stop in the United States — a coincidence, but one that will sharpen questions about how much of daily life now depends on a handful of systems nobody can see. Microsoft will publish a detailed account of what went wrong; the harder question, for its customers, is what backup plan they have the next time it does.
For now, the practical advice is simple: if a message sent on Thursday or Friday has not arrived, do not assume it was never sent. Check the sent folder, wait for the backlog to drain, and — for anything urgent — pick up the phone.
Related: Jensen Huang resigns as Nvidia CEO — another week of upheaval at the top of big tech. Also see how DeepMind used maths to prove Signal’s encryption.
