Blog / Tech, New & Old /
The AWS outage: one region, half the internet
On Monday 20 October 2025, a problem inside Amazon Web Services knocked a long list of apps, games and websites offline or into a crawl. People noticed it the way you always do: not through a news story, but because their usual apps stopped working at the same moment.
Here’s what happened, why one fault reached so far, and the part that applies to anyone with a website.
What happened, in short
The trouble was in AWS’s US-EAST-1 region, in northern Virginia. It’s one of the oldest and busiest of AWS’s regions, and a great many services use it. AWS reported that the root of the problem was a fault in DNS affecting the address of a core database service, DynamoDB. Because that service sits underneath many other things, the effects spread to other parts of AWS, and from there to the apps that rely on them.
Service was restored during the day, though some things took longer to recover while a backlog cleared. AWS has said it would publish a fuller write-up afterwards, so some details may still change. I’ll stick to what’s been stated so far.
The analogy: one road into town
Imagine a city with dozens of roads leading in, but most of the delivery companies built their depots on the same side of town, off the same main junction. As long as that junction works, nobody notices. The day it closes, the shops in every part of the city run out of stock, even though most of the roads are fine.
AWS has many regions around the world. But in practice, lots of companies put important pieces in one of them, often the biggest and oldest. And inside AWS, some of its own services also lean on that region.
Why one region matters so much
- It’s the default. For years, US-EAST-1 was where many things were started, and moving later is a hassle.
- Some services are anchored there. Certain global features rely on it, so a problem can reach customers who think they run elsewhere.
- Dependencies are hidden. An app might run on another provider, yet use a login service, a payment tool or an analytics script that sits on AWS.
- DNS is a single point of failure. If the system that turns a name into an address fails, everything looks “down”, even if the servers are fine.
That last point is worth remembering. DNS problems feel like total outages, because nothing can find anything. I wrote about the basics in DNS, SSL and Cloudflare: what site owners should know.
Who felt it
Many popular apps, games, banking and payment services and smart-home gadgets reported problems at some point. Which exactly depends on how each one is built, and I won’t list names I can’t verify. The pattern is what matters: lots of unrelated services failing together is the sign of a shared dependency.
What it means if you run a website
First, the good news: if you host a small WordPress site with an ordinary hosting company, this outage probably didn’t affect you directly, unless your host or a service you rely on uses AWS.
Second, the real lesson: your site depends on more than your host. Think through the chain:
- Your DNS provider.
- Your CDN or firewall.
- Your payment provider.
- Your email-sending service.
- Your analytics, chat and booking tools.
- Any third-party script that loads on your pages.
If any of these breaks, your site can look broken to visitors, even when your server is perfect. I wrote about the way big platforms communicate during these events in status pages and outages, and I’ve covered the general picture in cloud outages explained and why outages make headlines: the fragile internet.
What you can practically do
Don’t try to build a perfect, multi-cloud setup. For most small sites, that’s expensive over-engineering. Sensible steps are cheaper:
- Map your dependencies. Write down every outside service your site needs. It takes twenty minutes and is quite revealing.
- Find the critical ones. Which would stop a sale or a contact form? Those get attention first.
- Make failure gentle. Non-essential scripts shouldn’t block the page. If a chat widget is down, the page should still load.
- Monitor from outside, including your forms and checkout, not only the homepage. Uptime isn’t enough: monitoring forms, checkout and email explains how.
- Keep a status message ready. A short note on social media or an email to clients: “A major provider is having problems; we’re watching it” is better than silence.
- Know your backups and your way out. If a provider is down for days, could you move? Backups stored in the same place as the site aren’t much help.
A word on what to say to customers
If your service was affected, be plain and quick. Say what’s broken, whether it’s your fault or a provider’s, what customers should do, and when you’ll update again. Don’t blame a provider by name unless it’s necessary and you’re sure of the facts; people care about what happens next more than whose fault it was. I wrote about this in hosting outages: what I tell clients.
What to watch next
- AWS’s own written explanation of what failed and why.
- Whether providers change how critical services depend on one region.
- How many of your own tools turned out to depend on it.
Key takeaways
- A DNS-related fault in AWS’s US-EAST-1 region caused wide disruption on 20 October 2025.
- One region mattered so much because so many services, including AWS’s own, rely on it.
- Hidden dependencies mean your site can fail even if your own host is fine.
- Map dependencies, protect the critical ones, monitor from outside and have a message ready.
Related reading
Do you know which outside services your website would struggle without?