It has been over a week since the digital world woke up in panic: fast food apps were not loading, Adobe servers were down in Europe, and important services such as hospital networks were forced to resort to manual processes (as if we were in 1960!). It was chaos and confusion for an entire day. The fact is, we’re not used to mobile banking and streaming platforms freezing up.
This was a massive cyberattack—worthy of the latest Die Hard movie—and it wasn’t caused by a power failure triggered by a solar storm either. Believe it or not, it was a simple software problem that affected the backbone of the Internet: we are talking about the Amazon Web Services (AWS) outage.
Almost everyone with Internet access suffered the inconveniences of this outage to a greater or lesser extent on October 20, 2025. However, there is little information about what happened.
AWS and its collapse
Although we think of Amazon as just the world’s largest online store, it also has followers who invent half of the internet as we know it.
AWS is the invisible plumbing that supports giants like Netflix, Spotify, United Airlines, and millions of other companies. And in the quest for efficient use of resources, it’s much faster to hire Amazon’s services to host your domains than to try to have your own data center. The epicenter of this digital earthquake we experienced last week was located in a single mega-region: US-EAST-1, located in Northern Virginia, the world’s largest neural center for cloud computing.
It was all due to a synchronization error, known in jargon as a “race condition,” which brought the engine of the digital world to a halt.
Why did Amazon servers go down?
It all started with a problem with the Domain Name System (DNS), which is the universal phone book of the internet. Thanks to it, readable names like Amazon.com are converted into the numerical IP addresses that computers need to find the right server. The AWS bug caused the phone numbers for key services to be left blank, meaning that their entries in the phone book had not been deleted.
The AWS central database, called DynamoDB, mentioned a “race condition” in the automated management system of the NS.
Without getting into technicalities, a race condition occurs when automatic programs compete to update the same record without coordinating properly. In this case, two components called executors were trying to update the same record, but one was unusually delayed in its work. While the slow executor was working with outdated information, another faster executor applied a new plan and, as part of its cleanup, deleted the outdated plan that its counterpart was using. The slow executor finished its task, momentarily overwriting the record with old data, and a subsequent cleanup process immediately deleted that outdated data.
This caused DynamoDB’s main DNS log to be completely emptied… the rest is history. This systemic shutdown rendered thousands of applications that depend on that management inaccessible.
Digital Black Monday
People logged on to the Internet on October 20, 2025, and didn’t realize that something was wrong. Popular gaming platforms, Netflix streaming service, the Starbucks ordering app, United Airlines operating systems—everything went down in a chain reaction. It wasn’t just North America that was affected; both Europe and South America saw disruptions in payment methods and access to servers of all kinds. The main outage lasted approximately 15 hours and required manual intervention by engineers to restore the DNS registry.
To avoid repeating the same mistake, AWS has committed to implementing drastic changes, such as disabling DNS automation globally. This automation will remain disabled, at least until the race condition is fully repaired and additional protections are implemented. In the meantime, we will carry some cash around and buy a couple DVDs, just in case there is another outage and we cannot use our CC or access our Netflix account!