Quick Facts

  • Microsoft 365, Workday, the FAA, AWS, JPMorgan Chase, United Airlines, and Apple all suffered major outages between Nov. 12 and Nov. 17, 2026.
  • Each organization cited a separate, unrelated cause, from a July software bug to a disaster recovery test gone wrong.
  • 2026 is already a record year for major cloud failures, with 19 incidents recorded across AWS, Azure, Google Cloud, and Cloudflare through Nov. 16.

Seven major technology platforms went down in six days last week, and none of them failed for the same reason. The cluster of failures, spanning enterprise software, cloud infrastructure, aviation, and banking, has forced a hard question on technology leaders: if your systems are reliable but cannot absorb disruption, are they actually resilient?

Zeus Kerravala, principal analyst at ZK Research, argues the answer is no. High uptime figures, he writes, tell organizations how often systems work. They say nothing about what happens when systems stop.

What Broke and Why

The streak began on Nov. 12 when Microsoft 365 went dark for 20 hours and 15 minutes. Microsoft's post-incident report identified three compounding problems: a software bug deployed silently in July, a routine Teams failover that sent traffic surging, and reduced server capacity. More than 80% of the servers in the gateway fronting Microsoft 365 became unresponsive. Outlook, Teams, SharePoint, OneDrive, and Copilot all failed for users worldwide. It was the company's third major Microsoft 365 outage of 2026.

On Nov. 13, Workday pulled an entire production environment into emergency maintenance after a disaster recovery exercise went wrong. Services running inside Workday's DR environment had connected to live production databases, leaving customer data in what the company described as an inconsistent state. More than 10,000 companies were affected.

The same day, the FAA ordered a nationwide ground stop that lasted nearly five hours. The agency said a router replacement triggered excessive data traffic that destabilized legacy telecommunications equipment, with the network receiving approximately 1,000 connection attempts every four to five seconds.

AWS reported authentication, sign-up, and connectivity problems in its US-East-1 region on Nov. 15. JPMorgan Chase followed on Nov. 16, blaming a software bug in on-premise hardware that slowed transactions including some Zelle transfers. The bank stated there was no evidence of a cyberattack. United Airlines also went down on Nov. 16, grounding departing flights for about 100 minutes after applications used for crew communications and final aircraft weight data failed. Apple attributed several iCloud disruptions on Nov. 17 to an internal update.

One Pattern Across Separate Causes

Every organization that explained its outage pointed to a change: a software update, a network reconfiguration, a hardware replacement, a DR test, or a maintenance window. None confirmed a cyberattack. Online speculation about a coordinated AI-driven attack found no support in any published post-incident report.

The pattern matches broader data. Control Plane's outage tracker shows that configuration or deployment changes caused 20 of 63 major cloud outages since 2020, or 32%, making them the single most common trigger. Hardware failure ranked lower.

Scale of the Problem in 2026

The November week pushed 2026 into record territory. The tracker counted 18 major outages through Oct. 26, already ahead of the 16 recorded in all of 2025. The count has since risen to 19. The median outage lasted 5 hours and 57 minutes.

Geography offers less protection than many architects assume. Thirty-one of 62 tracked outages hit more than one region or a provider's global services. Among AWS, Azure, and Google Cloud alone, 18 of 46 outages crossed regional boundaries, meaning multi-region deployments within a single provider do not guarantee isolation.

The Microsoft 365 outage, at 20 hours and 15 minutes, is the longest recorded major cloud outage in the tracker's history.

The Business Implication

Kerravala's argument points to a gap that service-level agreements do not capture. A system can hit 99.9% annual uptime and still take 20 hours to recover when the wrong component fails. The organizations that fared worst last week were not necessarily those with the worst track records. They were those whose recovery systems either did not activate or made conditions worse.

Cloud-first resilience strategies increasingly treat recovery as an engineering discipline separate from availability metrics. Last week's events suggest more technology leaders need to treat it that way before the next outage, not during it.

This article was written by an AI agent. Spotted an error? Send a correction and we will fix it.