Amazon accidentally turned off large portions of the internet on Monday morning.
A problem at Amazon's cloud computing service disrupted internet use around the world early Monday, taking down dozens of online services, including social media site Snapchat, the Roblox and Fortnite video games and chat app Signal.
About three hours after the outage began, Amazon Web Services said it was starting to recover from the problem. AWS provides behind-the-scenes cloud computing infrastructure to some of the world's biggest organizations. Its customers include government departments, universities and businesses, including The Associated Press.
Amazon pinned the outage on issues related to its domain name system, which converts web addresses into IP addresses so websites and apps can load on internet-connected devices.
"The world now runs on the cloud," and the internet is seen as a utility like water or electricity, Burgess said.
Several major apps were not working. Coinbase, Fortnite, Signal and Zoom faced lengthy outages, as did Amazon's own services, including its Ring video surveillance products.
Millions of companies and organizations rely on AWS to host their websites, apps and other critical online systems. The company has data centers all over the world, and Amazon is said to have at least 30% of the total cloud market.
Amazon did not give a reason for what caused the outage.
https://apnews.com/article/amazon-east-internet-services-outage-654a12ac9aff0bf4b9dc0e22499d92d7
https://techcrunch.com/2025/10/20/amazon-dns-outage-breaks-much-of-the-internet/
(Score: 3, Insightful) by Anonymous Coward on Tuesday October 21, @11:59PM
without all of that pesky government regulation
(Score: 3, Touché) by c0lo on Wednesday October 22, @12:48AM (2 children)
Does an Internet outage still counts as creation?
https://www.youtube.com/@ProfSteveKeen https://soylentnews.org/~MichaelDavidCrawford
(Score: 0) by Anonymous Coward on Wednesday October 22, @12:54AM
(Score: 1, Interesting) by Anonymous Coward on Wednesday October 22, @02:24AM
Hmmm, creative network administration. Not sure how that'll look on a resume. But now that I think about today's (HR) culture, it might be golden.
(Score: 5, Insightful) by SomeGuy on Wednesday October 22, @01:39AM (3 children)
You know, its the cloud, it's magically impervious and has infinite resources, and all that stuff.
Or are we too busy drinking the AI cool-aid now to still believe that?
(Score: 3, Funny) by JoeMerchant on Wednesday October 22, @01:52AM
My AI Kool-aid tastes fine, but GitHub's mermaid diagram renderer of my markdown documents for the AI to implement is still borked.
🌻🌻🌻 [google.com]
(Score: 0, Insightful) by Anonymous Coward on Wednesday October 22, @02:33AM
Greed and gambling go hand-in-hand.
(Score: 2, Funny) by fen on Wednesday October 22, @03:54AM
Add the block chain, and run on a quantum computer.
(Score: 1, Interesting) by Anonymous Coward on Wednesday October 22, @01:59AM (5 children)
Running on AWS makes them pretty useless. As we have just seen, they are too easy to take down
Pretty easy to see why. Either it's embarrassing, or it was intentional... just following orders?
(Score: 5, Interesting) by Reziac on Wednesday October 22, @02:30AM (4 children)
Dave's Garage had an opinion about why it happened, dunno how accurate. Does sound a bit embarrassing.
https://www.youtube.com/watch?v=KFvhpt8FN18 [youtube.com]
And there is no Alkibiades to come back and save us from ourselves.
(Score: 5, Informative) by Thexalon on Wednesday October 22, @03:11AM (3 children)
Based on what I know about AWS as someone who worked for Amazon until recently, so not knowing the exact details of this but knowing the systems and the culture: His explanation sounds very plausible.
Human elements are at play also: There's been a lot of staff losses due to stupid RTO mandates and layoffs, and generally an effort to squeeze the techies more on both pay and working conditions. So it's entirely plausible that inexperienced engineers working on very limited rest either didn't anticipate a problem or accidentally caused one.
"Think of how stupid the average person is. Then realize half of 'em are stupider than that." - George Carlin
(Score: 2, Touché) by Anonymous Coward on Wednesday October 22, @03:18AM (2 children)
Maybe they got a dose of "you get what you pay for"?
(Score: 1, Touché) by Anonymous Coward on Wednesday October 22, @07:31PM (1 child)
Doesn't matter. They only care how well their hedge funds are performing. The store front is a fencing operation to sell counterfeit goods. Notice that nobody goes to jail whenever they get caught.
(Score: 2) by Thexalon on Thursday October 23, @12:52AM
The storefront is basically irrelevant to their profits right now. The big bucks are all in AWS, especially sweet sweet government cash going to it.
"Think of how stupid the average person is. Then realize half of 'em are stupider than that." - George Carlin
(Score: 1) by fen on Wednesday October 22, @02:47AM (1 child)
I worked at a big bank and we had to do everything in west and east (west-2, east-1). We had regular network isolation tests (isolate west or east). I'm sure a buncha people got paged for this one.
(Score: 5, Informative) by fab23 on Wednesday October 22, @01:57PM
The culprit was Amazon DynamoDB which failed in us-east-1, see Multiple services (N. Virginia) - October 20, 2025 [amazon.com] on their AWS Health Dashboard [amazon.com] with a timeline of events with a lot of details including a list of the 141 impacted AWS services at the end.
As far as I am aware, the us-east-1 (N. Virginia) AWS region may for some of their internal workings still be the single point of failure. As this was their first region, some of their core services depend on running services in us-east-1. It is not the first time that something in that region failed. As long as your infra is running on simple things own classic EC2 servers or e.g. from services like RDS nodes / clusters and in others regions it may not have been affected. But access to the AWS Console may not have been possible, and so also the API, e.g. if you are using that to dynamically adjust your infra based on load.
(Score: 4, Informative) by jb on Wednesday October 22, @06:26AM
That's a bit of a long bow to draw. Not convinced that taking one AS offline qualifies as a "large portion of the Internet". Remember, some time ago the 16 bit ASN space was exhausted, so we know for sure that there are more than 64k ASes out there. Granted Amazon is one of the largest of them, but it's still only one.
Only that subset of the world that doesn't care at all about control over its own computing. This should be a wake up call for many. If what you're using it for is no more important than a toy, then yes "the cloud" may well be a suitable home for it. But anything more serious than that requires sufficient direct control: ideally in-house, or failing that at the very least in a third-party data centre of known, constant location, with on-site access to the racks you rent there.
(Score: 5, Funny) by AnonTechie on Wednesday October 22, @07:04AM
Amazon Allegedly Replaced 40% of AWS DevOps Workers With AI Days Before Crash
https://80.lv/articles/amazon-allegedly-replaced-40-of-aws-devops-workers-with-ai-days-before-crash [80.lv]
Albert Einstein - "Only two things are infinite, the universe and human stupidity, and I'm not sure about the former."