Microsoft says it has completed its fixes after a global Microsoft 365 outage that disrupted Outlook email, Teams, SharePoint Online, Microsoft 365 Copilot and Defender XDR in more than 70 countries. The failure began on Thursday afternoon, November 12, and lasted at least 18 hours, with broad recovery reported late on Friday, November 13, and reports on Saturday saying users should no longer encounter problems.
The scale became clear quickly. Downdetector logged about 140,000 reports of Outlook problems on Thursday afternoon, from 74 countries by one BBC-cited count, and at one point it was receiving almost 20,000 reports a minute about five key services. Microsoft says more than 430 million people use Microsoft 365 applications.
The list of those affected showed how deeply the service is woven into daily operations. Staff at the United States House of Representatives lost email access, and city and county governments including Baltimore County, Austin and Jacksonville reported disrupted mail and applications. The Singaporean government was caught up in the outage, as were police forces, schools and doctors' offices in the United Kingdom, where the BBC reported disrupted NHS email and some cancelled appointments. In Colorado, a sheriff's office said the outage delayed people posting bond. The industrial technology company Workhorse Group postponed an earnings call, and Richemont delayed results.
Microsoft said the cause was a fault introduced during an earlier upgrade to its network capacity, which surfaced after a more recent network change. The company said a number of servers became unexpectedly unavailable, straining the remaining infrastructure, and that a core process on affected servers became overwhelmed. An early fix brought some servers back, but it failed under load once traffic returned, and on Friday the company acknowledged that its repair efforts were not taking as expected. Microsoft executive Charles Lamanna, executive vice president for Copilot, Agents and Platform, said there was no indication that the incident was the result of malicious or cyber activity, and the company said it was not related to AI agents.
The outage also exposed a newer dependency. Microsoft 365 Copilot was out alongside the services it reads from, and rival assistants felt the failure through their connectors. Anthropic's status feed reported elevated errors on Claude's Microsoft 365 connector for about five hours on November 12, affecting requests for mail, SharePoint, Teams chat and calendar, and OpenAI said ChatGPT's Microsoft connectors were not working as expected. The models themselves stayed up. What failed was the pipe between them and people's work data.
Microsoft restored services gradually, a step it said took longer but helped ensure a stable recovery, and it extended its email timeout window to 48 hours so delayed messages would not bounce. By 17:50 UTC on Friday it reported broad recovery and said the underlying issue was fixed, while warning that some users could still see intermittent access problems or delayed delivery. The company has promised a full post-incident report with the root cause and the steps it will take to prevent a repeat.
The lesson is about concentration. One vendor now supplies the email, meetings, documents, security tooling and AI assistant for a large share of the world's offices, so one network fault can halt thousands of organizations at once, and can quietly disable other companies' AI tools that depend on the same connectors. The outage is a reminder to map those dependencies and test how work continues when the main platform goes dark.
Photo: Building 92 on the Microsoft campus in Redmond, Washington. Photograph by Jiaqian AirplaneFan, via Wikimedia Commons, CC BY.