Microsoft’s preliminary investigation into its November 12–13, 2026, global Microsoft 365 outage has identified a combination of software and infrastructure failures that allowed a problem inside its network gateway system to spread across multiple services.
The incident disrupted Exchange Online and affected features connected to Outlook, Teams, Copilot, SharePoint and other Microsoft 365 products. In its preliminary post-incident report dated Nov. 15, Microsoft said the failure involved a latent software defect, a change in network traffic after Teams switched gateways, and a temporary reduction in available server capacity.
The company’s findings offer a more detailed explanation of why the disruption became so difficult to contain. They also make one distinction important: Microsoft attributed the incident to a technical failure, not artificial intelligence or a cyberattack.
A software defect remained hidden for months
Microsoft traced the first major part of the problem to a defect in a software change prepared specifically for its Microsoft 365 networking gateway servers.
The change began rolling out in July 2026. Microsoft said it passed its deployment process and spread globally over several weeks without producing observed problems.
The defect involved lock contention inside an internal Windows process. Under certain workload conditions, competing processes could prevent resources from being handled efficiently, eventually exhausting CPU and memory capacity and leaving affected servers unresponsive.
Microsoft said its usual laboratory load tests did not expose the weakness because the traffic mix and hardware used in testing differed from the conditions that triggered the problem. The issue was more pronounced on machines with more processor cores and newer hardware.
The company also stressed that the defect and the associated fix applied to its Microsoft 365 networking gateway servers, not broader Windows deployments.
That qualification matters. The incident should not be interpreted as evidence that every Windows computer or business server was exposed to the same defect.
Teams traffic added pressure to the gateway
The second factor involved how Microsoft routed traffic across its infrastructure.
Microsoft said it operates two independent network gateway systems designed to provide isolation and availability. About 15 hours before the outage, Teams automatically moved from an alternative gateway to the primary gateway, which also supported Exchange Online.
The failover was part of normal operations. However, it increased the workload and changed the traffic patterns reaching the primary gateway.
At the same time, nearly 14% of the gateway’s server capacity was unavailable because of an unrelated issue from the previous day. Microsoft said that capacity reduction had not caused an outage by itself because the system still had sufficient spare capacity.
The combination proved more damaging than any one factor alone. Increased traffic, reduced available capacity and the previously hidden software defect created conditions in which servers began failing.
Why the outage spread across multiple services
The most serious failure occurred when affected servers became unresponsive and Microsoft’s traffic-management systems redirected users towards other available endpoints.
Those systems are normally intended to protect service availability. During this incident, however, shifting traffic increased the workload on the remaining servers and made them more likely to encounter the same defect.
Microsoft described the result as a cascading failure across its network gateway infrastructure.
At the worst point, more than 80% of the servers were unresponsive. The company also said that a separate issue prevented the normal isolation barriers between some infrastructure partitions from containing the failure.
Exchange Online suffered the most significant impact. Many users could not access their mailboxes through Outlook or other mail clients, while email delivery was delayed.
Other Microsoft products that depended on Exchange Online capabilities also experienced disruption. However, Microsoft said many of those services largely remained online and accessible, with the impact concentrated in features that relied on the affected infrastructure.
This explains why the incident did not necessarily disable every feature of every Microsoft product at the same time.
Recovery became another technical challenge
Restoring the servers was not as simple as switching them back on.
Microsoft said some machines initially returned to service but became unresponsive again once they received waiting traffic. Its normal automated recovery processes were not designed to handle an outage in which such a large portion of the infrastructure had become unresponsive.
The company also said its usual rollback and feature-control mechanisms could not quickly reverse the problem because the underlying defect had been introduced months earlier.
Engineers eventually developed an update that returned the affected connection-handling code path to a previously known working state. They applied it to groups of servers, restarted the machines and gradually returned them to service.
Microsoft’s timeline recorded the first high-priority Exchange Online alert at 4:20 p.m. UTC on Nov. 12. It confirmed that customers were no longer experiencing active impact at 4:10 p.m. UTC on Nov. 13, although extended monitoring and delayed-message recovery continued.
The preliminary report said Microsoft completed recovery actions across portions of the Asia-Pacific infrastructure before global availability returned to expected levels.
What the outage says about AI and cloud dependence
The disruption comes as technology companies expand their investments in artificial intelligence and cloud infrastructure. Microsoft itself has been adjusting its computing partnerships and capacity strategy, a subject explored in WWC’s earlier coverage of its AI cloud infrastructure arrangements and the Oracle agreement.
But the outage report does not establish that AI systems caused the failure. Microsoft specifically described a networking-gateway software defect and infrastructure conditions, rather than an AI model or agent malfunction.
The broader business issue is resilience. A company can operate large networks with automated failover and still encounter situations in which those protections amplify a fault instead of containing it.
That is why the incident is relevant beyond Microsoft customers. Organisations increasingly rely on cloud platforms for email, scheduling, document access, security and business approvals. A failure in shared infrastructure can interrupt several workflows even when individual applications appear to remain available.
WWC previously examined the importance of separating technical failures from security incidents in its report on the US government’s assessment of recent technology outages. The same distinction is central here: a service outage is not, by itself, proof of unauthorised access or stolen information.
The wider question of digital reliability also appears in WWC’s analysis of technology disruptions and public trust.
What happens next
Microsoft’s Nov. 15 document is explicitly labelled a preliminary post-incident report. It said a final report would follow within five business days of full event resolution.
The preliminary report listed several completed recovery measures, including reverting the defective code path, adding capacity, improving monitoring of traffic redirection and connection handling, and introducing safeguards intended to reduce cascading failures.
A comprehensive list of repair items was scheduled for Nov. 20, 2026. That final assessment should provide a clearer picture of the longer-term engineering changes Microsoft intends to make.
For businesses, the immediate lesson is to review how critical operations depend on shared cloud services. Backup communication channels, documented recovery procedures and checks for delayed transactions or messages can reduce the operational consequences when a provider experiences a widespread failure.
Microsoft restored the affected services, but the preliminary findings show that the incident was not caused by one isolated malfunction. Several individually manageable conditions combined to overwhelm a system designed to keep services running. The final report will be important in determining whether the corrective measures address all the weaknesses exposed by the outage.
