Shortly before noon on 12 November, a large share of the working world lost everyday Microsoft tools. Outlook would not open, calendars went blank and meeting invitations hung unresolved. The trouble began in Exchange Online, the service behind Outlook, when users in the Americas could not reach their mailboxes. Within about two hours, Microsoft said the problem could reach users in any region.
Teams meetings mostly continued, but scheduling them through calendars failed. Copilot could not reach email or files, and security tools that check links clicked in Outlook and Teams could not always verify them.
As 13 November began in other time zones, the effects spread. Schools in Australia texted parents to phone in absences, and live video feeds from courtrooms went dark. Police forces in Britain could not take crime reports online and asked people to call instead. Pharmacists warned that prescriptions could be delayed because National Health Service email was affected. Several Norwegian ministries lost email, Singapore’s government email went down, and Japanese officials fell back on the telephone. Some Hong Kong banks reported partial disruptions, and an aluminum company there could not receive messages from its American customers. Air India rerouted its email while its planes kept flying. In the United States, at least one public company postponed its quarterly earnings call, and some office workers went home early.
About three hours into the outage, Microsoft engineers froze every deployment and configuration change across the service, redistributed traffic across worldwide infrastructure and updated their diagnosis in public. They first suspected a surge in load, then concluded that servers were dropping out because a process that sets up secure connections was becoming overwhelmed and stopping. Microsoft said early and repeatedly that it had no indication of malicious activity, and the author of the opinion piece, Steven Goodman, wrote that the outage was not caused by artificial intelligence running out of control or by a rush of Copilot users.
When a first wave of repaired servers went unresponsive again as soon as traffic returned, Microsoft said its changes were “not taking as expected.” Engineers restarted servers, added capacity, rolled back a recent network change that had coincided with the start of the trouble, and tested the fix on a small group of servers before a broader rollout. Goodman wrote that the company posted roughly 20 updates on X and worked directly with governments, which let police departments, hospitals and ministries focus on getting work done.
By the morning of 13 November, Microsoft was reporting broad improvement in North America, sustained gains across much of Europe and encouraging signs in Asia-Pacific, with email backlogs approaching normal. A full root-cause report was still to come.
Goodman, senior director of technology for the Klingler College of Arts and Sciences at Marquette University and a co-founder of WiscNet, argued that organisations should keep a lifeline off the main platform, such as a text-message list, a phone tree or a status page hosted elsewhere. He noted that many would otherwise try to announce an email failure over Teams, which shares the same calendars and dependencies.
