A cloud-connected router displays warnings across Microsoft apps, airport radar, a payment card, and flight information.
Chase, Microsoft and United had trouble within days of each other, and The Mary Sue's roundup (drawing on a New York Post interview) asks whether we're entering an era of constant outages. The headline invites a single story: AI, legacy systems and the holiday rush produce one big meltdown. The evidence points to separate failures with separate causes. Even the part that matters most to Windows and Microsoft 365 admins differs from how the roundup tells it.

A cloud-connected router displays warnings across Microsoft apps, airport radar, a payment card, and flight information. The short version​

  • Microsoft 365 (Thursday, Nov. 12 into Friday, Nov. 13): Exchange Online connectivity failed, and services that depend on it were hit too. Microsoft's explanation concerns a network capacity upgrade.
  • FAA ground stop (Friday, Nov. 13): This is a separate aviation-network incident. The FAA's early findings point to a router deployment and legacy equipment.
  • Chase (Monday, Nov. 16): Chase blamed a software bug. The editorial brief for this piece, citing AP, says the bug was in on-premises hardware systems.
  • United (Monday, Nov. 16): Departures were halted by a technical problem. AP, per the same brief, reported that it was resolved after more than 90 minutes.

There is no evidence that these incidents are connected. The roundup says so itself. The suggestion that AI is behind them has no support in the company statements.

Microsoft 365: what actually happened​

Microsoft traced the global outage, which cut users off from Outlook, Copilot and Teams, to a fault left behind by an earlier network capacity upgrade. The company's wording was that the incident was introduced by an issue during an earlier network capacity upgrade that surfaced following a recent network change. Microsoft also stated that this was a technical issue not related to cyber or AI / agents.

That differs from the roundup's phrasing of "a latent software defect combined with high server demand." It isn't wholly wrong, but the details are more specific, and they are still being reconciled:

  • Microsoft's preliminary root cause, as reported by webhosting.today, points to a failure in the systems responsible for establishing connections to the Exchange Online service. Traffic shifts during standard network maintenance and upgrades exposed a latent bug that had stayed dormant until then.
  • As a queue built up, the affected servers exhausted their available processing capacity and could no longer accept new connections.
  • Technology.org reports that Microsoft said a rise in traffic was not the trigger. That cuts against the "high demand" framing. Treat the demand question as unsettled.
  • A secondary write-up notes that an early update said Microsoft had not found a recent change matching the start of the problem, while the later one points to the capacity upgrade. Microsoft's own explanations had not yet been fully reconciled when that was written.

Timeline and impact​

  • Downdetector complaints about Outlook began rising around 12:25 p.m. ET. Microsoft's first public alert at 12:54 p.m. ET concerned Exchange Online mailbox access in the Americas, and a follow-up 13 minutes later warned the problem could reach customers worldwide.
  • Admins following the broader incident found it under MO1550821. The Exchange Online incident ID was EX1550543.
  • Services facing possible connection failures included Microsoft Teams, Exchange Online, Microsoft Defender XDR, SharePoint Online and Microsoft 365 Copilot.
  • Services that depend on Exchange Online went down with it, including Teams, SharePoint Online, Microsoft 365 Copilot, Intune and Defender. Impact varied by region, client and time, so not everyone lost everything for the full duration.

Recovery​

Microsoft was cautious about restoring service. Per the University of Mary Washington's IT notice relaying Microsoft's update, "given the scale of demand, we restored services gradually, validating stability at each stage." Other accounts describe Microsoft pausing deployments, moving traffic away from affected infrastructure and restarting servers (Technology.org). They also describe a rollback of a network configuration change that coincided with the start of the disruption (webhosting.today). These accounts are secondary, so details may shift as Microsoft's own post-incident reporting is read.

The status messages also mention an extended 48-hour email timeout window. Delayed messages were queued rather than bounced, which is why most mail eventually arrived. That window was relayed via the university's notice.

Where the paperwork stands​

One tracker reports that Microsoft set EX1550543 and MO1550821 to "Post-Incident Report published" as of Nov. 16. Earlier reporting had said a preliminary post-incident report was expected within two business days and a final version within five. The report is available to tenant admins in Service Health. If you run Microsoft 365, read your tenant's copy. It is the authoritative account, and it may refine what's described above.

The FAA ground stop: related, but not the same story​

The FAA's Q&A, as relayed by the roundup, says its initial investigation points to two possible sources: a radar site in Amarillo, Texas, and a radio communications site in El Dorado, Arkansas. L3Harris, the agency's network provider, had been deploying a new router at El Dorado. A flood of connection attempts, roughly 1,000 every four to five seconds, destabilized legacy equipment and disrupted radar feeds. Operators canceled about 12,462 flights, though the FAA cautioned that not all were directly caused by the ground stop.

Per the roundup, the FAA attributes the instability to its legacy TDM system, noting that towers already upgraded under its modernization effort stayed up. It also says L3Harris blocked both sites and the FAA disabled broadcasting on about 60 similar routers. It plans to reevaluate maintenance timing, since the router swap happened midday at a remote facility. These are initial findings, not a final root-cause determination.

Chase and United: little detail, and that matters​

Chase blamed a software bug, and the roundup says the bank hasn't said whether a software update caused it. The roundup reports that transactions involving Zelle, Bank of America, Capital One and Wells Fargo were affected. According to the editorial brief's reading of AP coverage, Zelle itself was not directly hit. Customers of other banks may have seen delayed transfers involving Chase without the Zelle network being down. That distinction is why "Zelle went down" is likely too broad.

United paused departures because of a technical issue and later said it was resolved. No cybersecurity link was indicated by either company. I won't speculate about which system failed, since neither company has said in the material available here.

The "new normal" claim: commentary, not a finding​

Marva Bailer, CEO of Qualaix, told the New York Post that legacy systems are becoming Frankenstein's monster. Her argument: they combine third-party vendors with AI bolted on, so there are more points of failure. Holiday volume and change-freeze deadlines add pressure. That is a plausible description of risk in complex systems. It is also an opinion, and the facts don't establish it as the explanation here:

  • Microsoft explicitly ruled out AI and agents.
  • The FAA incident involved legacy TDM equipment, which is the opposite of new AI tooling.
  • The roundup concedes there is no proof AI caused these crashes.
  • No outage-frequency data was offered to support "new normal."

Where Bailer's point does carry weight is change management. Two of the four incidents involve infrastructure changes. Microsoft's issue was introduced in a capacity upgrade and surfaced after a later network change. The FAA's trouble followed a router deployment. That is a lesson about changes interacting with dormant faults, not about AI or the calendar. Whether that pattern recurs is unproven.

What admins and businesses can do​

The sources support general resilience practices rather than incident-specific fixes:

  • Know your incident IDs and where to look. Check Service Health in the Microsoft 365 admin center, and use Microsoft's public status page if the admin center is unreachable. Microsoft itself pointed admins to both during this incident.
  • Have an out-of-band communication path. If Exchange Online is down, so may be the mail-based alerts that tell you about it. Keep a phone tree, chat alternative or SMS channel handy.
  • Map your dependencies. This outage reached Teams calendars, Copilot, Defender and Intune because they lean on Exchange capabilities. Document what stops when mail stops.
  • Don't resubmit. The university's guidance was to avoid repeatedly submitting the same message or transaction. Hammering retry can add load or create duplicates.
  • Save work locally during an outage where possible, as that guidance also advised.
  • For payments: check transaction status before retrying. The sources don't say how delayed or duplicate transactions were handled.

Bottom line​

Four incidents in five days make for a tidy narrative, but the facts don't support one cause. Microsoft's outage was a network-layer problem it says is unrelated to cyber activity or AI. The FAA's was a router deployment hitting legacy equipment. Chase and United cited a software bug and a technical issue, with few details. The practical takeaway for Windows and Microsoft 365 shops is less dramatic than "the new normal": read your tenant's post-incident report, test your fallback communications, and treat infrastructure changes, including your own, as risk events.


Update: Homeland Security secretary suggests AI link (November 17, 2026)​

Homeland Security Secretary Markwayne Mullin suggested AI-driven upgrades could be contributing to outages, according to the Washington Times. The theory is not an established cause: Microsoft says its outage was unrelated to AI or agents.

Microsoft’s preliminary incident report adds detail to its earlier account. It identifies a latent defect in a Microsoft 365 gateway-specific fix, a Teams traffic shift onto the primary gateway, and an unrelated capacity reduction of nearly 14%. The defect does not apply to general Windows deployments.

Microsoft said testing missed the defect because contention worsened on newer, higher-core-count machines. At the outage’s worst point, more than 80% of gateway servers were unresponsive. Microsoft expects a final report by November 20.

 

References

  1. Did AI Cause the Microsoft 365 and FAA Outages? What the Evidence Shows washingtontimes.com
  2. Why are Chase, Microsoft, and United all having outages? - The Mary Sue The Mary Sue 2026-11-17T18:45:00+00:00
  3. Microsoft 365 Outage: Cause and Recovery - webhosting.today webhosting.today