On November 12, Microsoft Exchange Online, the email platform behind Outlook, stopped working for organizations around the world. Microsoft's preliminary post-incident report dates the start of customer impact to 16:20 UTC, and users in more than 70 countries reported problems. Downdetector logged more than 130,000 Outlook problem reports at the peak. Features in Teams, SharePoint, Copilot, and Microsoft's security tools that rely on Exchange Online were degraded too, although their core functions largely kept working.
Microsoft traces the outage to a latent software defect on the gateway servers that handle connections to Exchange Online, not to cyber activity. Its official impact window ran just under 24 hours, ending at 16:10 UTC on November 13. By CyberCube's count, severe disruption lasted about 18 hours, and the event ran roughly 38 hours in total. CyberCube, which nicknames the event "M364," estimates that global insured cyber losses will most likely be less than $100 million.
What happened
Users began losing access to their mailboxes around midday US Eastern time on Thursday, November 12. Microsoft paused deployments and began restarting and rerouting servers that afternoon, but recovery stalled overnight. Servers that came back online were quickly overwhelmed by backed-up traffic, and early fixes did not hold.
The turning point came at 07:20 UTC on Friday, when Microsoft began deploying an update that returned the affected code on its gateway servers to its previous known good state, applying it in stages across groups of servers. Independent network monitors saw Exchange Online traffic begin recovering from 11:20 UTC, and by 14:17 UTC Microsoft said email queues were significantly reduced. At 16:10 UTC, Microsoft confirmed customers were no longer actively affected. It extended its mail delivery window so delayed messages would not bounce, and it posted its final customer update at 04:15 UTC on Saturday, November 14.
What caused it
On November 15, Microsoft shared a preliminary Post-Incident Report with administrators in the Microsoft 365 admin center. It replaces the explanation Microsoft gave during the incident, which pointed to an earlier network capacity upgrade. The report describes three factors acting together:
- A latent defect. A Windows fix rolled out to Microsoft 365 networking gateway servers in July could, under certain traffic conditions, exhaust the servers' CPU and memory until they stopped responding. The fix had run in production for more than three months without reported issues. Microsoft says the problem was worse on newer machines with more cores than the hardware used in load testing.
- Extra traffic. About 15 hours before the outage, Teams automatically failed over to the gateway infrastructure that also serves Exchange Online.
- Reduced capacity. Almost 14% of gateway capacity was already out of service because of an unrelated issue.
As servers failed, their traffic moved to other servers, which then failed too. Within two hours, more than 80% of gateway servers were unresponsive, and regional safeguards did not stop the spread. Teams, SharePoint, and OneDrive could switch to alternative infrastructure and recover sooner, but Exchange Online could not. Microsoft says general Windows deployments are not affected. It has apologized and expects to publish its final report around November 20.
The fallout
The list of organizations reporting disruption was long and varied:
- Governments: US city, county, and state agencies, courts, and sheriff's offices, as well as government agencies in Australia and Japan.
- Healthcare: hospitals in Canada and parts of the UK's NHS.
- Transport and logistics: airlines including IndiGo, Akasa Air, and Air India (for customer email support), plus Maersk and Deutsche Bahn.
- Financial services: trading firms, fund managers, and companies reporting earnings that week.
The type of disruption matters as much as the breadth. Unlike the 2024 CrowdStrike event, flights weren't grounded, and the reported healthcare impact was to communications rather than clinical systems. The reports focus on inconvenience and lost productivity. Organizations switched to phone calls and other messaging tools, and there's little to no mention of lost revenue.
The outage did expose how much sits on one vendor. Some users were locked out of unrelated systems because their multifactor authentication codes arrive by email, and continuity plans that switch from Outlook to Teams stay on Microsoft infrastructure. Outlook alone has more than 400 million users.
Why insured losses should stay modest
At first glance, M364 looks like the kind of event cyber insurers worry about. It hit a single, widely used provider; it reached every region, and it ran well beyond the eight- to 12-hour waiting periods common in contingent system failure cover. Several factors still hold insured losses down:
- Coverage has to respond. Cover applies only where policies extend business interruption to non-malicious system failure at an outside provider. Some wordings respond only to a security incident, and Microsoft attributes this event to a software defect.
- Revenue was mostly delayed, not lost. Business interruption claims depend on showing income that was lost. As Aon noted after the CrowdStrike event, many businesses hit by an outage see revenue briefly delayed rather than lost, especially where manual workarounds exist. With email down and workarounds widely available, that's the likely pattern here.
- Email isn't in the critical transaction path for most businesses. Organizations could still sell, ship, and serve customers through other systems.
- No reported data loss or exposure. There have been no reports of data corruption or leakage, and Microsoft held queued email rather than returning it as undeliverable, delivering it once service recovered.
- Timing helped. Much of the disruption fell outside US business hours, and the backlog was cleared heading into the weekend.
The closer comparison many in the market reached for is the October 2025 AWS outage, which CyberCube nicknamed "Amazonk." The two aren't a one-for-one match. Amazonk disrupted the cloud infrastructure that companies run their applications and platforms on. M364 was largely confined to email and the Microsoft 365 tools that depend on it.
CyberCube's loss estimate
In a preliminary analysis published on November 16, CyberCube estimates that global insured cyber losses from M364 will most likely be less than $100 million, with a conservative upper bound of $350 million. That's a loss ratio impact of 0% to 2% of global cyber premium.
We expect most outcomes to cluster toward the lower end of that range. The high estimate leaves headroom for information that could still emerge. The estimate is based on CyberCube's catastrophe model and global industry exposure database.
Jon Laux, CyberCube's VP of Products and Analytics, said: "This event once again reminds us how the technology world is concentrated in the hands of a few key providers. But it also reminds us that not all layers of the technology stack are equally critical to business operations. Factors that informed our loss estimates include the relatively short duration of this event, the absence of any reported data corruption or leakage, and the ease with which companies fell back on alternative communication channels to email."
Part of a bigger pattern
Over the longer term, M364 adds to a pattern. CrowdStrike in 2024, Amazonk in 2025, and now M364 have each shown how one provider's failure can reach a large share of a cyber portfolio at once, across industries and regions. Each has also shown that the outcome for insurers depends on which layer of the technology stack fails, and how critical that layer is to the way insureds make money. (Re)insurers that understand which providers their insureds depend on, and for what, will be better placed to tell the next disruptive event from the next costly one.
Preparing for the outage that hasn't happened yet
M364 was disruptive but contained. A longer outage, a failure in a more critical layer, or a malicious version of the same event could look very different. That makes it worth pressure-testing portfolios against more severe scenarios now.
CyberCube clients can identify exposed insureds through the Event Response dashboard in SPoF Intelligence and estimate losses for their own portfolios in Portfolio Manager. Our Concierge team is available to help work through the results.
To see how CyberCube keeps clients informed as major cyber events unfold, learn more about our Cyber Aggregation Event Response Service (CAERS).