
Microsoft links 365 outage to July gateway software defect
Automated rerouting reproduced the failure across more than 80% of gateway servers after traffic rose on reduced capacity.
Overview
Microsoft 365 users lost access to email, meetings, files and Office applications after a hidden software defect turned a routine into a global outage. The defect, introduced in a July 2026 gateway update, stayed dormant for more than three months. Then shifted traffic to the gateway also serving and , while an unrelated problem had removed nearly 14% of gateway capacity. Under the combined load, exhausted servers’ CPU and memory, making them stop responding. Automated rerouting sent traffic to other servers, spreading the failure until more than 80% were unresponsive.3
Microsoft restored services through , infrastructure recovery and a code rollback. Its preliminary report says testing missed the defect because laboratory traffic and hardware differed from production. The incident shows that redundancy can amplify failures when spare capacity is limited and rerouting systems treat overloaded machines as healthy. Microsoft is now reviewing its load tests, capacity safeguards and automated , though the final investigation could identify further weaknesses.3
Timeline
Microsoft began corrective work on testing, capacity and automated failover
Microsoft said it would continue a final post-incident review and use the findings to improve resilience. The incident focused corrective work on more representative traffic and hardware testing, stronger and closer review of automated rerouting so redundancy does not amplify a localized failure.
Microsoft said production differences allowed the defect to escape testing
The preliminary findings said Microsoft’s test environment did not match or infrastructure closely enough to reproduce the failure. The report indicated that differences in traffic patterns and hardware, including newer machines with more processing cores, helped explain why conventional missed the defect.
Microsoft’s preliminary report linked the outage to three combined conditions
Microsoft said the incident required the July software defect, the that increased gateway load and the unrelated loss of nearly 14% of gateway capacity. More than 80% of were unresponsive at the outage’s lowest point.
Microsoft identified connection negotiation as a main cause of access failures
By about 00:30 UTC, Microsoft had identified problems and competing processes that prevented connections from being handled. The company restored services through , infrastructure recovery and other operational measures while continuing its investigation.
Automated rerouting spread the failure across additional gateway servers
As individual servers became unresponsive, redirected requests to servers considered healthy. Those servers encountered the same defect under the additional load, spreading the failure through the fleet in less than two hours.
Lock contention exhausted gateway servers’ CPU and memory
The increased load and reduced capacity activated the dormant July defect. inside an internal Windows process consumed server CPU and memory, causing gateway machines to become overwhelmed and stop responding.
Teams failed over to the gateway serving Outlook and Exchange Online
About 15 hours before the outage, Teams moved from an alternative networking gateway to the primary gateway that also handled Exchange Online. The increased traffic and changed traffic patterns on the already reduced-capacity gateway.
Microsoft 365 services began failing worldwide
Users experienced widespread disruption across Outlook, , Teams, OneDrive, SharePoint and Office applications. The outage affected access to email and other Microsoft 365 services, with and Outlook experiencing the longest disruption.
An unrelated issue removed nearly 14% of Microsoft’s gateway capacity
A separate infrastructure issue left nearly 14% of the networking gateway’s server capacity out of service. Microsoft said the reduction alone was manageable because routine maintenance and were restoring capacity, and sufficient remained.
Microsoft introduced a latent defect in a networking gateway update
A Microsoft 365 update introduced a defect that could overwhelm servers under specific load conditions. The change passed Microsoft’s existing checks and remained dormant for more than three months without causing a visible service-wide failure.