Microsoft Apologizes as a Two-Day Outage Hits Outlook, Teams and Copilot in 74 Countries and Rules Out AI Agents

Microsoft has formally apologized for a global outage that knocked Outlook, Teams, Copilot and a long list of Microsoft 365 services offline starting Thursday afternoon, stretched into a second day, and disrupted millions of users across 74 countries. By Friday afternoon the company said service availability and email delivery were improving across a growing portion of the affected infrastructure, though some users still faced intermittent access.

The blast radius was unusually wide. Downdetector logged close to 130,000 reports on Thursday afternoon and was still receiving more than 50,000 every fifteen minutes at one point. Microsoft's status page itself briefly went down, and by Friday it still listed service degradation for both Microsoft 365 and Microsoft Consumer Products. The affected list ran to more than 20 services: Exchange Online, Microsoft Copilot 365, Teams, SharePoint Online, OneDrive, Word, Excel and PowerPoint among them, plus Defender XDR, Microsoft Fabric, Microsoft Graph, Intune, Purview, Power BI and Windows 365.

The disruption reached well beyond offices. In the United Kingdom, the National Health Service said 111 and 999 services could not refer patient details normally, creating large backlogs, and some police stations asked residents to report crimes by phone. US federal agencies and court systems reported problems, as did local governments including Baltimore County, Jacksonville and Austin. In Australia, the security intelligence organization ASIO and the Australian Federal Police reported knock-on effects, an Indian airline reported customer service disruptions, and even OpenAI and Anthropic hit connectivity issues tied to their own Microsoft 365 dependencies.

On the cause, Microsoft ruled out the two explanations everyone assumed first. There was no indication of malicious or cyber activity, the company said, and the outage was not related to AI or AI agents — an explicit denial notable in a week when agent-driven incidents have dominated headlines. The failure was traced to a technical issue introduced during an earlier network capacity upgrade that surfaced after a recent network change, causing workloads on some servers to spike and others to stop responding during automated checks.

Charles Lamanna, Microsoft executive vice president for Copilot, Agents and Platform, said in a statement that teams are reviewing what happened and how to prevent similar issues. The company said it deliberately restored service in stages, validating stability at each step, which made recovery slower than past incidents but more stable once complete. Email queues, at one point severely backlogged, were described as near normal by Saturday morning Beijing time. A full post-incident report is promised once the review is complete.

Industry observers flagged the scale as rare. Former engineer Gergely Orosz told the BBC it could well be the most disruptive email-related business outage of all time, noting that a 12-plus-hour mitigation for a service with more than 100 million users is almost unheard of. Microsoft's stock slipped less than 1 percent at Friday's open, suggesting investors treated the incident as operational rather than structural — but for an enterprise now betting its AI strategy on Copilot woven through every one of those services, two days of silence is a reputational data point of its own.