Microsoft said all affected Microsoft 365 services were working again after a global outage that lasted more than 24 hours and hit products businesses and governments now treat less like software than plumbing.
In a post on X cited by Deutsche Welle, the company said it had “completed” work on the underlying issue and restored service to remaining affected customers and regions. Microsoft said it was not aware of outside interference and blamed server issues.
That matters because the market tends to hear “cyberattack” and “technical fault” as different species of risk. For the people trying to send email or hold meetings, the distinction is less poetic. Teams, Outlook and Copilot all went down. If your office runs on Microsoft 365, “server issues” is simply the modern phrase for work stopping.
Deutsche Welle reported that web testing and network diagnostics company Ookla put the start of the disruption at about 1725 UTC on Thursday. Downdetector logged more than 7.4 million problem reports worldwide. Those figures are not a clean customer count, but they are large enough to make the point without theatre: this was not a local wobble or a fussy login screen.
The outage also exposed who carries the cost of concentration. Germany’s Federal Ministry for Digital Transformation and Government Modernization said federal functions were not heavily affected, but German businesses were significantly hit. Britain’s NHS was reported to have fallen back on paper records. Cloud computing was sold as resilience, efficiency and someone else’s problem. The bill arrives when one supplier’s bad stretch becomes everyone else’s manual workaround.
Microsoft’s scale is what makes these incidents financially interesting as well as operationally annoying. A software suite used across governments, hospitals and large employers can absorb immense trust for years, then remind customers in a single outage that diversification was always more expensive than dependence right up until dependence failed. Big Tech calls that ecosystem strength. Buyers sometimes discover it is just vendor risk in a better suit.
Deutsche Welle described the disruption as one of the longest in Microsoft’s history and the second outage in 2026. That does not by itself prove a broader reliability problem, but it does narrow management’s room for shrugging. One outage can be bad luck. Two in one year, with one stretching beyond 24 hours, starts to look like a denominator.
The company’s reassurance that services are restored closes the incident for users in the narrow sense. It does not close the more awkward question for large customers: how much operational dependence on one provider is efficient, and how much is just laziness with a procurement department.
Gabriel Fenech 
Ryan C 
Dua Mifsud