Skip to content

Tech Industry

Microsoft says three combined factors caused the Microsoft 365 outage

Microsoft's preliminary report, dated 15 November, ties the disruption to a latent July bug, a traffic shift and reduced server capacity.

https://breach-signal.com/tech-industry/microsoft-365-outage-three-factors/

Share

Key facts

  • Microsoft said a preliminary post-incident report attributes the global Microsoft 365 outage to three combined factors.
  • The three factors were a latent software defect, increased traffic demand, and reduced server capacity.
  • The outage lasted from 16:20 UTC on 12 November 2026 to 16:10 UTC on 13 November 2026.
  • Microsoft's pre-release testing did not find the defect because test hardware was less powerful than production systems with higher CPU core counts.

Microsoft’s preliminary report, dated 15 November, attributes the global Microsoft 365 outage to three combined factors, with the disruption running from 16:20 UTC on 12 November 2026 to 16:10 UTC on 13 November 2026, or 11:20 a.m. ET on 12 November to 11:10 a.m. ET on 13 November.

Microsoft identified a latent software defect, increased traffic demand and reduced server capacity as the three factors behind the failure. A change deployed to Microsoft’s cloud gateway starting in July created the defect, and the fix had operated worldwide for more than three months without trouble before the outage. About 15 hours before the outage, Teams traffic shifted from the alternate gateway to the primary gateway that also handles Exchange Online. On the previous day, an unrelated issue had taken nearly 14 percent of the gateway’s server capacity out of service.

Together, the added load and reduced capacity activated the July defect during peak U.S. traffic hours. Under certain conditions, the bug caused CPU and memory resources to become exhausted and unresponsive, and server load grew until affected systems used up capacity and stopped accepting new connections. Newer machines with more processor cores suffered worse effects from the defect. Microsoft’s pre-release testing did not find the defect because test hardware was less powerful than production systems with higher CPU core counts.

Microsoft runs two independent gateway systems for resilience, so traffic can move between them during disruptions. The fail-safe shift added load and changed request patterns on the primary gateway. Automatic traffic management moved users to healthy endpoints, which spread the load across fewer machines and caused cascading failures. A separate issue prevented safeguards from containing the failure within individual network sections or regional capacity units, allowing it to spread globally.

At the lowest point, more than 80% of the affected gateway’s servers stopped responding. Restored servers failed again as soon as waiting traffic returned and overwhelmed them. The automated recovery tools could not handle a breakdown of that scale. Because the defect was months old, the standard rollback tools could not quickly reverse it.

Exchange Online, which powers Outlook e-mail, suffered the worst disruption and stayed down longest. The service could not switch away from the faulty infrastructure, while Teams, OneDrive, SharePoint and Office apps could route traffic elsewhere. Teams chat and meeting functions remained operational for most of the outage. Features in Teams, Copilot, Defender XDR, Purview and Intune that depend on Exchange Online failed alongside it.

The outage affected Microsoft 365 and Exchange infrastructure globally and spread to at least 70 countries. Copilot, government offices, courts and hospitals were among the affected sectors. Microsoft described the outage as one of the largest in the company’s history, and the disruption lasted just under 24 hours. Company data stayed intact throughout the outage.

IndiGo and Akasa Air said their customer-support e-mail systems had problems. OpenAI and Anthropic said they experienced problems with Microsoft 365 integrations. In Germany, excessive chlorine in a swimming pool injured 15 hotel guests at a Phantasialand theme park hotel when e-mail alerts about a monitoring malfunction did not arrive because of the outage. Forty-first responders were dispatched, and four guests needed hospital treatment for respiratory irritation.

TechRadar said the German chlorine incident shows serious and unexpected consequences can arise when central e-mail systems go down. Deployflow said a firm running e-mail and threat detection on one platform loses both in the same hour. Deployflow said firms under the FCA’s operational resilience rules cannot pass blame upstream and must keep impact tolerances for important business services even when a third-party provider causes a breach. Deployflow said EU financial entities have similar ICT third-party duties under DORA, and both regimes require a documented, tested workaround for a 24-hour e-mail outage.

Engineers created an update to restore the code path to its previous working condition. They applied a correction, restarted server groups in stages, and brought them back gradually.

A driver stands in the cab of a vintage steam locomotive billowing thick clouds of white steam.
The billowing clouds of steam from the vintage locomotive serve as a powerful visual metaphor for the massive cloud infrastructure and staged restart required to bring complex systems back online. Photo by Géo on Pixabay

Microsoft added server capacity, widened the e-mail delivery window to prevent delayed messages from bouncing, and added monitoring plus limits on traffic redistribution. Additional monitoring remained in place for affected systems through 14 November. Microsoft said full service has resumed.

Charles Lamanna, Microsoft executive vice president of Copilot, Agents and Platform, said on X that this was not a cyber attack or an issue caused by AI tooling. Lamanna apologized for the impact to customers and their normal business and operations. Microsoft ruled out cyber activity and AI agents as causes on 13 November, and listed the final status as a technical issue unrelated to cyber activity or AI agents. The company said neither the defect nor the fix applies to broader Windows deployments.

Microsoft’s first alert about the problem went out at 11:24 pm IST on 12 November 2026 for users in the Americas. Global webmail availability returned to expected levels at 9:17 a.m. ET on 13 November 2026, according to the report. Microsoft tracked the incident under IDs EX1550543 and MO1550821.

A final post-incident report is due by 20 November. Microsoft said customers do not need to make tenant or device changes and promised repair details by 20 November.

Microsoft’s service level agreement promises 99.9% monthly uptime for Exchange Online, Teams and SharePoint. Below 99.9% uptime earns a 25% credit, below 99% earns 50%, and below 95% earns 100%. Affected businesses can request service credits from Microsoft, with claims due by 31 December 2026, when the calendar month after the incident closes. The credits apply only to fees for the affected service and do not cover lost revenue or idle staff time.

Previous Microsoft 365 incidents include a January 2026 outage from elevated load during maintenance lasting more than nine hours. In August 2026, a faulty core authentication setup took down Exchange Online, Teams, SharePoint and Defender XDR for roughly three days. Smaller Exchange Online incidents also occurred in March, April, May and June 2026.

Patch Tuesday two days before the outage fixed an Outlook spoofing flaw tracked as CVE-2026-124194 that attackers were already using. Microsoft’s advisories draw no link between the outage and the Outlook spoofing flaw.

Rep. Rich McCormick, R-Ga., said the outage revealed a weakness in digital systems used by businesses, hospitals, schools and government agencies. McCormick tied his Common-Sense Cloud Security Act to the disruption and called for better cloud risk checks, threat spotting and damage control. He said the nation should strengthen how it evaluates cloud risk, spots threats and contains harm when systems break down.

Was this helpful?

Join the discussion

Please read our Comment Policy before commenting. First-time comments are held for moderation. By posting a comment you agree to our Terms of Use and Privacy Policy, which explain how your name, email address and comment are used.

Your email address will not be published. Required fields are marked *