Microsoft says a fault introduced during a network capacity upgrade, which surfaced after a later network change, knocked out its cloud services on Thursday.
The company apologized on Friday, saying it was "deeply sorry" to customers and partners. Its statement said the incident was "a technical issue not related to cyber or AI [or] agents," and that it had fixed the underlying problem and was still restoring service to remaining customers and regions.
According to Downdetector, which tracks user reports, Microsoft 365, Teams, Copilot and OneDrive all had problems. Outlook drew hundreds of thousands of reports Thursday afternoon, and still ran above 50,000 every 15 minutes at one check even as reports tapered off. The Microsoft 365 status page was unreachable for at least some users that day. By Friday afternoon it still showed "Service degradation on Microsoft 365 (Business or Enterprise)" and listed 20 possibly affected services, including Exchange Online, SharePoint Online, Intune, Power BI and Windows 365.
Microsoft's account of the response shows how the problem unfolded. It first cited elevated load on some infrastructure, then paused all deployments and moved traffic away. It found that a number of servers had become unavailable after failing health checks, and suspected the component that processes server health. An alternative health check method restarted some of them. The company said it restored service gradually to test stability at each stage, which is why the fix took longer than some might expect.
The sequence matters more than the apology. A change made earlier, then exposed by a later one, is a familiar failure pattern in large systems, and it is hard to catch in testing. The explicit denial of a cyberattack or an AI cause also suggests Microsoft expected those questions.
The promised post-incident report should say what the upgrade got wrong, why the health check component faltered, and why the status page went down too.