Modern humans don’t realize how dependent they are on the Internet until one day they wake up in the morning… and it’s not working. This is what happened on Monday, October 20, 2025, and it became a unique event that affected the entire planet Earth. The problem with the Internet is that, because it is global, once it goes down… Everyone goes after it as if they were tied to an anchor.
It was a bit of a disastrous day: bank transactions stopped, travelers couldn’t check in for their flights, the Starbucks app didn’t work, and streaming platforms went silent. Although it wasn’t a bad thing for Netflix addicts to get outside and touch the grass, it was a real pain for everyone else who needed to get around. The digital world was completely paralyzed. However, the cause was not a spectacular cyberattack or a natural disaster.
It was due to a tiny internal failure, a lack of synchronization at the heart of the global infrastructure: Amazon Web Services (AWS).
The Root Cause of Internet Chaos
Although we may believe that Amazon is making a fortune just by being an online store, it is the backbone that supports more than 30% of the global Internet.
It is responsible for maintaining the servers of giant companies such as Netflix, Spotify, and Starbucks. Fortunately, the failure did not occur on all servers at once, but on one in particular: US-EAST-1 (Northern Virginia). Although this is the largest neural center in the cloud, it would have been much worse if all AWS servers had gone down at once.
Good heavens, let’s talk about that hour-long outage known as “Amazonk.” There was a failure in the DNS; the Domain Name System (DNS) is the universal “phone book” that makes it possible to translate web names into numerical IP addresses. Thanks to it, we can type Amazon.com and it translates it so that the servers understand the computer language we are asking for. If the DNS fails, computers don’t know where to find the servers—hence why we call it the “phone book.”
DynamoDB, AWS’s NoSQL database, was at the center of this mess. DynamoDB stores data critical to the internal functioning of other Amazon services. The failure began with a “race condition”; simply put, this is when automated programs (executors) attempt to write to the same record at the same time, but for whatever reason, they are not perfectly coordinated.
This is exactly what happened: while the slow Executor was working with an obsolete plan, a faster Executor applied a new plan and, while cleaning up, deleted the reference to the slow Executor’s plan. The result? A primary DNS record in DynamoDB was left empty (without an IP address). It was the equivalent of getting hit in the head with a ball during recess and not remembering your own name. This happened last Monday, October 20, and we also experienced a third of the Internet going down.
Like headless chickens
The inaccessibility of DynamoDB caused a chain reaction: the first service to be affected was EC2 (Elastic Compute Cloud). EC2 was unable to perform any maintenance tasks, as it relies on DynamoDB to store its operating states.
…Then came the collapse of the NLB (Network Load Balancer). The NLB, responsible for distributing traffic, began reporting failures in its “health checks” of EC2 servers. This caused a “flapping” effect, with the NLB constantly removing and adding servers from traffic, as it mistakenly believed that the instances were failing.
This erratic behavior generated connection errors and slowed down the overall recovery. What had been an initial failure ended up impacting 113 internal AWS services. A cyber disaster, indeed!
Consequences of Amazonk
Internet services were disrupted for more than 15 hours in total, affecting nearly 70,000 organizations around the world. Although it is too early to say, cyber risk analysis firm CyberCube estimates the financial impact at between US$38 million and US$581 million.