AWS Outage, October 20, 2025: How One Region Took Down Half the Internet

If you tried to join a Zoom meeting, send a Slack message or play Fortnite on the morning of Monday, October 20, and got an error screen instead, your home internet wasn't to blame. The problem was Amazon Web Services (AWS), the cloud backbone that runs a huge share of the global internet. On Monday, that backbone failed.

The official AWS Health Dashboard confirmed an "Operational issue" affecting multiple services in the N. Virginia region, showing up as "increased error rates and latencies."

Ground Zero: US-EAST-1

As with many global internet outages, it all started in one place: the US-EAST-1 region in Northern Virginia. It's one of the largest and oldest hubs in the AWS network, which also makes it, well, a regular source of trouble.

Amazon officially confirmed that it was investigating "Increased Error Rates and Latencies" in the region, with the first status dashboard post appearing at 12:11 AM PDT. In cloud speak, that's a diplomatic way of saying "something has gone very, very wrong." The domino effect was immediate, and the failures quickly spread around the world.

The True Scale of the Outage: What Exactly Failed at AWS?

Before we get to the list of apps that went dark, it's worth looking at what broke under the hood. The US-EAST-1 outage wasn't a single-service problem. It was a cascading failure. At its peak, the AWS Health Dashboard showed 65 services directly affected.

The disruption hit some of the most critical building blocks of the cloud, including:

• Databases: The trouble began with "significant error rates" for the Amazon DynamoDB database endpoint. Amazon Aurora and RDS were also affected (mentioned in connection with the EC2 issues).
• Compute: Launching new EC2 instances ran into serious problems, which users may have seen as an "Insufficient Capacity Error." That in turn immediately hit dependent services such as ECS (Elastic Container Service) and AWS Glue.
• Data processing and functions: AWS Lambda saw delays, especially in its critical event source mapping from SQS (Simple Queue Service) queues.
• Global services: The failure in a single region even affected global features that depend on US-EAST-1, such as IAM (Identity and Access Management) updates and DynamoDB global tables.
• Other key services: The list went on and also included Amazon CloudFront (CDN), Amazon S3 (Simple Storage Service), AWS Secrets Manager, Amazon Redshift, Amazon SageMaker, Amazon OpenSearch Service, Amazon Kinesis, Amazon SES (Simple Email Service) and AWS Storage Gateway.

The outage was so severe that customers "may be unable to create or update Support Cases."

From Messaging Apps to Banks: How Far the Disruption Spread

For everyday users, Downdetector lit up. Reports spiked at the same time for services in almost every category. The outage showed how much of our lives, from work to entertainment, we have handed over to a single company.

The list of casualties reads like a who's who of the internet:

• Business and Work: Thousands of companies ground to a halt. Key collaboration tools such as Slack, Zoom and Asana stopped working. Atlassian and Smartsheet services went down too.
• Entertainment and Gaming: Players were kicked off Fortnite, Roblox, Clash Royale and the Epic Games Store. Music fans lost access to Tidal.
• Finance (Fintech): Users of crypto exchanges (such as Coinbase), trading apps (such as Robinhood) and payment services (including Venmo) had some tense moments. Even traditional banks, such as Lloyds and Halifax in the UK, reported problems.
• Everyday Apps: You couldn't order a burger in the McDonald's app, design a graphic in Canva, practice a language on Duolingo, send a message on Signal, look something up on Perplexity, or use Grammarly or Snapchat.
• Amazon's Own Backyard: The outage hit Amazon itself as well. The Amazon.com store and Prime Video had problems, as did Amazon's IoT devices: Alexa smart speakers stopped responding and Ring doorbells failed to send notifications.
• Other Services and IoT: Autodesk and Supabase servers went down, and even Docker.io authentication failed. Users reported that their iRobot vacuums (Roomba), smart door locks (Amazon), smart plugs (Amazon) and video doorbells (Ring) had stopped working.
• Media: Even the biggest news outlets, such as The New York Times and the BBC, as well as Disney platforms, had outages.

What Went Wrong? DNS, the Usual Suspect

According to AWS's official updates, the root of the problem was a DNS failure.

At 2:01 AM PDT, Amazon identified a "potential root cause." The problem was "related to DNS resolution of the DynamoDB API endpoint in US-EAST-1."

Put simply, DNS is the internet's phone book. It translates human-readable addresses (such as google.com) into the IP addresses that machines use. When it fails, services can't "find" each other. Servers try to call one another, but nobody picks up. And because DynamoDB is a foundational database for many other AWS services, its failure took a lot of other services down with it.

Recovery came in several stages:

• 12:11 AM PDT (9:11 a.m. Polish time): Amazon begins investigating.
• 12:51 AM PDT (9:51 a.m. Polish time): Amazon confirms "increased error rates and latencies" in US-EAST-1.
• 1:26 AM PDT (10:26 a.m. Polish time): Amazon confirms "significant error rates" for DynamoDB.
• 2:01 AM PDT (11:01 a.m. Polish time): Root cause identified: a DNS issue affecting the DynamoDB API.
• 2:22 AM PDT (11:22 a.m. Polish time): "Initial mitigations" applied and "early signs of recovery" observed. Customers are advised to retry failed requests.
• 2:27 AM PDT (11:27 a.m. Polish time): Amazon reports "significant signs of recovery," and most requests should now succeed.
• 3:03 AM PDT (12:03 p.m. Polish time): Amazon confirms that global services relying on US-EAST-1 are also recovering.
• 3:35 AM PDT (12:35 p.m. Polish time): Amazon announces that "the underlying DNS issue has been fully mitigated" and that "most AWS Service operations are succeeding normally."

As usual with an outage this large, though, not everything was back to normal right away. AWS acknowledged that some services, such as CloudTrail and Lambda, were still "processing a backlog of events." Even after the DNS fix, launches of new EC2 instances in US-EAST-1 continued to see "increased error rates."

As late as 4:48 AM PDT (1:48 p.m. Polish time), Amazon was still working to fully restore EC2 launches and recommended that customers configure their Auto Scaling groups to use multiple Availability Zones. It also confirmed that the Lambda/SQS delays were affecting, among other things, "Organization policy updates."

At 5:10 AM PDT (2:10 p.m. Polish time), AWS confirmed that Lambda had recovered SQS queue processing but was still working on "processing the backlog of SQS messages in Lambda queues."

AWS also advised customers still having trouble with DynamoDB to "flush their DNS caches."

A Lesson in Dependency (Once Again)

The October 20, 2025 incident was one of the largest global AWS outages in years. It was also another wake-up call, showing the global economy how heavily it depends on a single provider.

AWS holds roughly 30% of the entire cloud infrastructure market. Competitors, media companies, governments and financial systems all run on its infrastructure. Put simply, when AWS sneezes, the whole internet catches a cold.

The outage also raises the question that Niebezpiecznik.pl asked in its article: "What do we do now, how do we carry on?" This is the perfect moment to ask yourself "whether your organization really has a contingency procedure in case a business-critical tool or service becomes unavailable." And whether "every employee knows what to use for online meetings and communication when, for example, the company's Zoom or Signal goes down."

Sources:

• Business Insider Polska
• Reuters
• The New York Times
• Bankier.pl
• SpidersWeb
• Money.pl
• Onet Wiadomości
• NBC News

Aleksander

Read also: The Cloud in Strategic Sectors: Opportunity or Risk? and The Cloudflare Outage: The Internet's Dark Autumn.