Why are there so many outages? The November 2026 ledger
Between 10 and 18 November 2026, at least twelve large organisations had a public outage. None has blamed an attack, and nobody has found a supplier they share. What stands out is how little has been explained: one has published a technical report, four have given no cause at all, and the rest have given statements that cannot yet be checked.
Last updated 20 November 2026. Every cause below is the organisation's own account, linked to the page where it was read.
How much has been explained
| What the organisation has said | How many | Which |
|---|---|---|
| A statement that cannot yet be checked | 7 | FAA, Workday, NTT Docomo, JPMorgan Chase, Marriott, NSA, Apple |
| No cause given | 4 | AT&T, AWS, United, PlayStation Network |
| A published technical report | 1 | Microsoft 365 |
What we know
- Nobody has blamed an attack. Microsoft's report describes a software defect.[1] The FAA says there is “no indication that this was a cyberattack”.[2] Chase, United, Marriott, Docomo and Apple each said they found no sign of a cyber incident.[3][4][5][6][7] The Director of National Intelligence said of the NSA outage: “This is not a cyber attack.”[8]
- The most common cause is the organisation's own change. Microsoft (a fix deployed in July),[9] the FAA (a router replacement),[10] Workday (a backup-site test),[11] Apple (an “internal update”)[12] and the NSA (maintenance and patching)[13][14] all describe planned work that went wrong.
- Only one explanation can be checked. Microsoft published a seven-page preliminary report with a timeline.[15] Every other cause in the ledger is a statement from a spokesperson or an official.
- The bank outage was one outage, not several. Chase said other banks were affected only where they were party to a transaction with it. Bank of America, PNC, Capital One and Truist said they had no disruption; U.S. Bank said its systems were not directly affected.[16]
The ledger
One row per outage. None given means the organisation has not said what caused it.
| Outage | What failed, and for how long | Cause the organisation gave | Evidence behind it |
|---|---|---|---|
| AT&T 10 Nov | Network nodes in New York and Ashburn, spreading to other US cities[17] 50 minutes within a two-hour window[18] | None given | None |
| Microsoft 365 12–13 Nov | Exchange Online and Outlook; parts of Teams, Copilot and SharePoint[19] 23 h 50 min[20] | A hidden defect in a fix deployed since July made gateway servers lock up under load. A Teams failover added load while nearly 14% of capacity was out of service, and more than 80% of servers stopped responding.[21] | Preliminary technical report, 15 Nov. Final report due 20 Nov.[22] |
| FAA 13 Nov | The network carrying radar and radio feeds, unstable at 17 of 22 en-route centres. Nationwide ground stop; about 12,462 flights cancelled that day.[23] About 4 h 45 min (about 3:30 pm to 8:15 pm ET)[24] | Data flooded the network “at a rate of 1,000 connection attempts every 4 to 5 seconds”. Two possible sources: a radar site in Amarillo, Texas, and a radio site in El Dorado, Arkansas, where contractor L3Harris was deploying a new router.[25] | FAA questions-and-answers page. Root cause still under investigation.[26] L3Harris says only that it is supporting the investigation.[27] |
| Workday 13–14 Nov | HR and payroll platform, two data centres[28] About 18 hours[29] | Services at the backup site “unexpectedly connected to Production databases”, leaving data inconsistent[30] | Customer notice. No report promised.[31] |
| AWS (us-east-1) 15 Nov | Sign-in, sign-up and nine other services[32] 3 h 16 min[33] | None given | None. The resolution notice gave no root cause.[34] |
| NTT Docomo (Japan) 16 Nov | Nationwide 5G standalone service[35] About 6 hours[36] | Traffic concentrated on the switches serving 5G subscribers. Its third such disruption in November.[37] | Company statement[38] |
| JPMorgan Chase 16 Nov | Transaction processing; delays for other banks' customers transacting with Chase[39] Resolved by Monday evening; no times given[40] | A “software bug in on-premise hardware systems”[41] | Spokesperson statement[42] |
| United Airlines 16 Nov | Departures; FlightAware showed 715 delayed United flights that afternoon[43] About 90 minutes[44] | None given | None from United. A pilots' union memo said the weight-and-balance system could not supply final weights.[45] |
| Marriott 17 Nov | Sign-in on Marriott.com and the Bonvoy app; mobile room keys[46] About 5 hours[47] | A “technical issue”[48] | Spokesperson statement[49] |
| NSA 17 Nov | Classified internal network (NSA Net)[50] Not disclosed | NSA: “temporary degradation during scheduled routine maintenance”.[51] Director of National Intelligence: systems failed to come back after security patching.[52] | Two official statements that differ |
| Apple 17 Nov | iCloud sign-in, App Store, retail systems, business and education tools[53] Several hours, from about 11:45 am ET[54] | An unspecified “internal update”[55] | One-line statement to Bloomberg[56] |
| PlayStation Network 18 Nov | Sign-in, gaming and social features[57] About 1.5 to 2.25 hours[58] | None given | None |
Where the official accounts conflict
- Whether the FAA's upgrade work caused the ground stop. The FAA's page answers “No” to whether its upgrades caused the outage, and blames older equipment. The same page says the disruption “was a result of a routine replacement at a remote facility” and that it has paused installing the new routers.[59]
- What happened at the NSA. The agency's statement says only “temporary degradation during scheduled routine maintenance”.[60] The Director of National Intelligence said patching after a PHP vulnerability left systems unable to come back up: “That is what has happened at the NSA today.”[61] The NSA has not confirmed that account.
Is this more outages than normal?
Probably, for outages at big-name institutions. Not clearly, for outages overall. Four measures exist, and they do not count the same thing.
| Measure | Figure | What it tells you |
|---|---|---|
| Downdetector reports about Microsoft's outage, to 13 Nov | More than 7 million worldwide: the largest outage in 12 months, about two and a half times the next largest[62] | One outage was exceptional. It says nothing about the others. |
| Downdetector US problem reports, week to 16 Nov (as relayed by an analyst column, not in Ookla's own article) | Nearly 6 million, the highest week in a year[63] | Reports were unusually high, but about 3.9 million came from the Microsoft outage alone.[64] |
| Cisco ThousandEyes network outage events, 9–15 Nov | 419 worldwide, down from 450 the week before; 282 in the US, up from 234[65] | The background rate of network outages did not jump. This count leaves out application failures such as Microsoft's. |
| One analyst's count of major outages at nationally important institutions | 9 in six days (12–17 Nov), against no more than 2 in any six-day stretch of the previous 90 days[66] | Points to a real cluster. The author says the search was not exhaustive and the week was picked because of the outages.[67] |
No source has counted the same set of organisations, with the same definition of an outage, over the past year. Until someone does, “more than normal” rests on one informal count.
What is still unexplained
What would settle it
How this page was made
Each row was filled by asking one narrow question per organisation, then checking the answer against the page it cites. Microsoft's report was read in a copy hosted outside microsoft.com.[74] A reported brief Charles Schwab website problem on 17 November is left out because its statement could not be read at source.
Let's look at alternatives:
- Modify the query.
- Start a new thread.
- Remove sources (if manually added).