Last updated: November 17, 2026
The Microsoft 365 outage of November 12, 2026, the FAA ground stop, Chase’s banking outage, and Apple’s iCloud sign-in failure were not cyberattacks; each was traced to a software defect, an update, or overloaded equipment. That puts all four in the territory of operations work, the change reviews, capacity margins, and legacy know-how that ops teams carry.

Thursday, Outlook. Friday, the national airspace. Monday, the biggest bank in the country. Tuesday morning, iCloud.
Four outages in five days, and by the end of it half my inbox was some version of the same question. Was it a hack? Is somebody testing us?
No. Not according to anyone who would know. Not even close. Microsoft said its outage was “a technical issue not related to cyber or AI / agents.” Chase said its problem “is not cyber-related.” Federal officials said there was no sign of a cyberattack behind the ground stop, and Apple told Bloomberg an internal update was the cause. Four organizations. Four shrugs at the hacker theory.
Honestly? That’s the less comfortable answer. A hack is something done to you. A bad change, a capacity margin nobody was watching, a forty-year-old copper line on the far side of a migration, those are things organizations do to themselves, usually with nobody available to stop them.
Where I sit, so you can weigh what follows. I co-founded KORE1, and our IT staffing services team gets paid when companies bring in network, cloud, and reliability engineers. I’m not going to tell you these four teams were understaffed either, because nobody outside those buildings knows that, me included. What I can do is read what each organization actually said, line it up against the outage research, and tell you which seats that points to in your own shop. That part I do know.
What Caused the Microsoft 365 Outage?
A defect that had been sitting in production since July.
A latent defect is a bug that ships, passes testing, and runs quietly for weeks or months until some combination of load and conditions finally triggers it. When it fires, the change that caused it is old news, already rolled out everywhere and far harder to pull back than last night’s deploy would be.
Here’s the chain, from the preliminary post-incident report Microsoft gave customers on November 15, as reported by The Next Web and Cyber Security News. In July, Microsoft began rolling out an update built for its Microsoft 365 network gateway servers. It contained a defect that caused lock contention inside an internal Windows process, which could exhaust a server’s CPU and memory. Load testing missed it because the tests ran on weaker hardware with fewer cores, and the problem gets worse on newer machines with more of them. The update ran for more than three months without visible trouble. Months.
Then three things lined up. Separately, none was enough. At 01:07 UTC on November 12, Teams traffic failed over to the gateway that also carries Exchange Online. Nearly 14% of that gateway’s server capacity was already out of service from an unrelated issue the day before. At 16:20 UTC the defect fired, and within two hours more than 80% of the servers behind the gateway were unresponsive at the worst point. Exchange Online and Outlook got the worst of it. Microsoft lists the window as 16:20 UTC on the 12th through 16:10 UTC on the 13th, under incident EX1550543. Call it 24 hours without reliable email for a lot of companies.
Microsoft executive Charles Lamanna wrote that “normal rollback, flighting, and feature flags were unable to quickly revert the change.” Engineers had to write a fix that put the old code path back and apply it to groups of servers in turn. Group by group.
Nothing exotic there. Not one of those three ingredients is. A test rig that doesn’t match production. A failover nobody flagged as risky. Capacity lost to something unrelated and not yet back. Each is the kind of thing an operations review is supposed to catch, and each is the kind of thing that slips when the review is one person deep. The final report is due November 20, and I’ll read it before I’d claim anything more specific.
The FAA Ground Stop: New Equipment Meets Old Copper
At 3:54 p.m. Eastern on Friday the 13th, the FAA stopped departures nationwide. Airspace reopened at 8:15 p.m., per Al Jazeera’s account of the day. Thousands of flights were cancelled, and the counts differ by source and time window, which our breakdown of the FAA ground stop and the router install behind it sorts out. The FAA called it a telecommunications outage affecting air traffic control systems. Transportation Secretary Sean Duffy said switches and routers “became overloaded with data and went down,” and put it bluntly. “We have old switches, old routers. We use 1980s little copper lines.” Officials also said new technology had overloaded antiquated parts of the system.
That last sentence is the whole problem with modernization, in about ten words.
The Government Accountability Office has been writing about this for years. Its September 2026 report on air traffic control modernization says the FAA had replaced 2,560 of 5,170 “old and frail” copper wires with fiber as of May 2026, describes copper cables dating from the 1960s whose technology “is no longer supported by the communications providers,” and repeats its 2024 finding that 51 of the FAA’s 138 air traffic control systems were unsustainable and another 54 potentially unsustainable. Then there’s the line I’d put on the wall of any IT department halfway through a migration. FAA officials told GAO the aging systems have been hard to maintain because of missing parts “and retirement of technicians with expertise in maintaining the aging systems.”
Read that twice. Then read it again. The people who know the old system leave before the old system does. During a migration you need both crews at the same time, the people who understand the copper and the people installing the fiber, and that’s exactly the stretch when budgets assume you need only one.
The investigation isn’t finished. I’m not claiming the ground stop came from a staffing gap. I’m claiming the GAO says the knowledge to keep the old gear running is retiring, and that the outage happened at the seam between old and new.
Chase and Apple: A Bug and an Update
Monday afternoon it was money. Real money. Downdetector logged nearly 4,500 Chase reports shortly after 1:30 p.m. Eastern and more than 20,000 by 2:22, according to Fortune. Customers couldn’t log in, see balances, or move money, and some said branch transactions and ATMs were affected. Zelle payments involving Chase stalled at other banks too, and Bank of America, Wells Fargo, and others showed up on the outage maps. Chase said other banks were slowed only where they were a party to a Chase transaction. Chase told NBC News it “experienced a temporary issue related to a software bug in on-premise hardware systems.” Around 6:30 p.m. a spokesperson said transactions were posting normally.
On-premise hardware. At a bank that size, that’s the gear you own, rack, patch, and keep running yourself, and nobody’s cloud SLA covers it. Chase hasn’t said more. I won’t guess.
Apple’s turn came Tuesday morning. Its System Status page flagged iCloud Account & Sign In as affecting all users, with the App Store, Apple Music, and AppleCare on Device also listed, and people reported trouble activating new iPhones. Apple’s only explanation, given to Bloomberg and relayed by TechCrunch, was “an internal update caused some outages this morning.” What that update touched, and why it took down sign-in, Apple hasn’t said. So the cause, in any useful sense, is unknown, and I’m leaving it there.
Here’s the week on one screen.
| Outage | When | How long | What the operator said caused it | Cyberattack? |
|---|---|---|---|---|
| Microsoft 365 (Exchange Online, Outlook, Teams) | Thu, Nov 12 | About 24 hours (16:20 UTC to 16:10 UTC next day) | Latent defect in a July gateway update, triggered by a Teams failover and 14% capacity already out | No, per Microsoft |
| FAA nationwide ground stop | Fri, Nov 13 | About 4 hours 20 minutes (3:54 to 8:15 p.m. ET) | Telecom outage; overloaded switches and routers; new technology overloading antiquated parts. Investigation ongoing | No sign of one, per officials |
| Chase online, app and some branches | Mon, Nov 16 | Afternoon; posting normally around 6:30 p.m. ET | Software bug in on-premise hardware systems | No, “not cyber-related” |
| Apple iCloud sign-in, App Store, Apple Music | Tue, Nov 17 | Most services back by afternoon; sign-in still flagged later | “An internal update.” No further detail | No, per Apple’s statement as reported |
If They Weren’t Hacks, What Were They?
Ordinary. Painfully ordinary. That’s what should worry you.
Uptime Institute has tracked outages for years, and its Annual Outage Analysis 2026 reads like a description of this week. Among operators who had a serious network-related outage in the past three years, 41% named configuration or change management failure as a common cause, and 9% named a malicious cyberattack. For IT system and software outages, firmware or software faults led at 53%, change management issues came in at 51%, and cyberattacks or security issues at 15%. Hackers get the headlines. Changes get the outages. Every year.
The human side is sharper.
- 92% of operators said human error contributed at least a little to their most recent serious outage. Only 8% said it played no part.
- Asked how human error contributed, 59% pointed to staff not following procedures, 36% to procedures that were wrong in the first place, and one in four (25%) to insufficient staff.
- In Uptime’s 2025 survey, 87% of operators who had an impactful outage said better management, processes, or configuration could have prevented it.
- And it isn’t cheap. 57% said their most recent major outage cost more than $100,000, and one in five put it above $1 million.
Uptime is careful to call human error a contributing factor rather than a root cause, and I’ll be just as careful. A latent defect is a software problem. It’s also a test environment somebody sized, a failover somebody approved, and a capacity loss somebody was going to get to. Most of the time, that somebody has four other jobs. Sometimes five.

Where Thin Ops Teams Actually Show Up
Not in the outage itself. In the six months before it.
Look at the week again with that lens, and four failure points come out, none of them unique to Redmond or the FAA. I see versions of them in the requisitions we get from 300-person companies.
The test that doesn’t look like production came first, at Microsoft. Its load tests ran on smaller machines than the ones in the fleet. In a mid-market shop the version is a staging environment that hasn’t matched production since the last hardware refresh, because keeping it matched is a full-time job and nobody has that job.
Second, headroom you’ve already spent. Fourteen percent of Microsoft’s gateway capacity was out from something unrelated. Your equivalent is the second firewall that’s been “temporarily” out of the pair since August, or the database replica that died and is on next quarter’s list.
Then there’s the migration seam, which is the FAA’s story and probably yours too. You’re moving from copper to fiber, on-prem to cloud, an old core to a new one. For the length of that project you need people who know the old system cold and people who know the new one, and the old-system people are exactly the ones retiring, leaving, or getting pulled onto the new build.
Who owns the rollback? Lamanna’s line about rollback, flighting, and feature flags not working quickly enough is the one I’d hand to any CTO. A rollback plan nobody has rehearsed is a hope. Not a plan. Rehearsing takes hours, and hours are what a thin team doesn’t have.
None of this says Microsoft, the FAA, Chase, or Apple were short on people. Those are four of the most resourced operations organizations on the planet, and they still had these weeks. The lesson for a company with a six-person infrastructure team is simpler. If it can happen to them with what they’ve got, plan on it happening to you with less.
What to Staff Before Your Own Outage
Start with the seats that close those four gaps, and decide which ones are permanent and which are a project.
Permanent ownership of production belongs with employees. A site reliability engineer who owns capacity, failover, and rollback testing is a direct hire for most companies, because that knowledge has to stay in the building. Project work is different. A copper-to-fiber cutover, a data center exit, a firewall pair rebuild, six months of load-testing a new platform on real hardware. That has an end date, and it’s where contract staffing earns its fee. The contractor carries the project so your permanent people can keep running what’s live.
Here’s what those project seats bill at in our current index. Most of these cells are modeled national numbers, since our own placement count is too small in some of them to publish, and I’ve labeled which is which.
| Seat | What it covers before an outage | Hourly bill rate (25th to 75th, median) | Three months, full time | Basis |
|---|---|---|---|---|
| Senior network engineer | Circuit and switch migrations, restoring the redundant pair, change windows | $115 to $139, median $128 | $66,800 | Modeled, U.S. national |
| Senior cloud or DevOps engineer | Staged rollouts, feature flags, rollback drills, load tests on production-sized hardware | $115 to $139, median $128 | $66,800 | Modeled, U.S. national |
| Senior systems administrator | Keeping the old platform patched and documented while the team builds the new one | $90 to $109, median $100 | $52,200 | Modeled, U.S. national |
| Systems administrator, all levels | Day-to-day coverage so a senior person can be pulled onto the project | $54 to $70, median $57 | $29,600 | KORE1 observed, n=12, primarily Southern California |
Rates come from the KORE1 Contract Rate Index 2026-Q4, effective October 16, 2026. Modeled cells use BLS wage data with a 24% burden and our observed bill bands, and the full method is on our contract rate methodology page. Three months is 520 hours.
Speed matters more than the rate. Usually. Once an offer is accepted, KORE1 contractors start a median of 3 days later (n=1,997), and our average time-to-hire on IT roles is 17 days. Reading this the morning after your own outage? You’re late. During an incident the people who know your system are the ones who fix it, and a stranger arriving mid-fire mostly gets in the way. We wrote a whole piece on when outside engineers actually help during a production crisis, and the honest answer is “usually not in the first hour.”
The better window is now, while your Monday-morning incident review still has everyone’s attention. If the review says you’re migrating something old, add a contract network engineer or a sysadmin for the length of the cutover and budget the old-system expertise as its own line. If the pager is waking the same two people every week, read what your on-call load is telling you about your architecture before you hire anybody.

What People Asked Us This Week
Was the Microsoft 365 outage a cyberattack?
Microsoft says it wasn’t, calling it “a technical issue not related to cyber or AI / agents,” and its preliminary report traces it to a defect in a July update to its gateway servers. The final post-incident report is due November 20, so details may still change.
Were the four outages connected?
Nothing public connects them. Microsoft, the FAA, Chase, and Apple each gave a different cause, on different systems, on different days. Chase’s trouble did spill into other banks through Zelle, but that’s one outage spreading, not two outages linked. A cluster like this is mostly coincidence plus the attention a cluster gets.
How long was Outlook actually down?
About 24 hours for the worst-hit customers, from 16:20 UTC on November 12 to 16:10 UTC on November 13 by Microsoft’s count. Plenty of tenants recovered sooner. Exchange Online took longest because, per the report, it couldn’t fail over the way Teams could.
We’re a 300-person company. Does any of this apply to us?
More than it does to Microsoft, honestly. You have the same failure points, an untested failover, a staging box that doesn’t match production, one person who understands the old system, with a fraction of the staff to cover them. Write those three down for your own environment this week, and put a name next to each.
Should we hire a full-time SRE or bring in contractors?
Both. They do different jobs. Hire full time for whoever owns production, capacity, and rollback over the long run, since that knowledge needs to stay. Use contractors for work with an end date, like a cutover, a decommission, or a load-testing push. Contract seats in our index run about $100 to $128 an hour at the median for senior infrastructure roles.
Before the Next One
The final Microsoft report lands November 20, and the FAA investigation is still open, so some of what I’ve written may need a footnote by Christmas. The pattern won’t. It never does. Each of these four outages began as a change, a migration, or a bug, the kind of work an operations team exists to catch.
Pick the one gap from this week that worries you most. If closing it needs a person you don’t have, tell our team what you’re migrating and by when, and we’ll say straight out whether a contractor fits or whether your own people should take it. And if you’re in the middle of retiring an old system, keep the person who knows it until the last switch is off.

