On November 15, Microsoft released a preliminary incident report regarding the November 12 event that paralyzed the tech giant's key cloud services and affected millions of users worldwide. According to the document, the root cause of the massive outage was a hidden defect in a Windows update introduced back in July 2026. The error remained undetected for four months as it successfully passed all standard tests. This is reported by RBC-Ukraine with reference to the company's report.

How the error bypassed tests: The powerful hardware issue

According to the preliminary report, the July update was tested primarily on weaker hardware with fewer processor cores. Under these conditions, the defect did not manifest, and the update was released to production without issues. However, on newer and more powerful servers with a large number of cores, the error led to lock contention within an internal Windows process. This resulted in the complete exhaustion of CPU and memory resources on the affected nodes.

"Domino Effect": Two minor factors triggered a cascading failure

The November 12 accident was the result of two concurrent circumstances, neither of which posed a threat on its own: a hidden bug in the code and automatic traffic redirection to a neighboring node upon the first server's loss of connection. When the first server went down, the system automatically shifted the load to a healthy neighboring node, which immediately overloaded for the same reason and failed as well. Within two hours, this "domino effect" spanned the company's entire infrastructure: at the peak of the outage, over 80 percent of servers had stopped responding.

15 hours of emergency response

The first signs of overload appeared on November 12 at 03:07 Kyiv time following a scheduled switch in the Teams service. The monitoring system recorded massive connection problems at 16:03. Over the next 15 hours, Microsoft engineers unsuccessfully attempted to stabilize operations: all planned changes were halted, and reserve capacities were added. Standard automated recovery tools proved powerless as they were not designed for an outage of this magnitude, and a simple rollback was impossible due to the four-month "age" of the July update.

Exchange Online: The most vulnerable service

Most cloud services—Teams, OneDrive, SharePoint, and Office apps—managed to switch to reserve infrastructure relatively quickly. The exception was the Exchange Online mail service and the Outlook client: due to the massive scale of the load, they could not recover from the problematic gateway and remained unavailable longer than others. Only on November 13 at 00:30 did specialists determine the exact cause of the outage—a connection synchronization failure—and at 07:20, they released a special patch that returned the problematic Windows code segment to its previous working state. Full restoration of all services was completed on the same day at 16:10.

Final report due November 20

Microsoft stated that the document published on November 15 is preliminary. The company plans to release a final report with a complete breakdown of causes, a timeline, and recommendations to prevent similar incidents in the future on November 20, 2026.