Why one cloud outage can take down half the internet

What happened in October 2025

On 20 October 2025, Amazon Web Services (AWS), the largest cloud provider, had a major disruption in its US-EAST-1 region in Northern Virginia. According to AWS's post-incident report, a hidden timing fault in the automated system that manages DNS records for its DynamoDB database service left a key address record empty. Applications could no longer find the database, and engineers had to intervene manually.

Network monitoring firm ThousandEyes reported that effects lasted more than 15 hours. As other AWS services that depend on DynamoDB struggled, problems spread to apps such as Slack and Snapchat, games, banking services and smart home devices around the world.

Why one fault spreads so far

• Concentration: a handful of cloud providers host a large share of the world's apps and websites. When one region struggles, many unrelated services stop together.
• Hidden dependencies: an app may use several services that all quietly rely on the same underlying component, such as DNS or a database. See DNS vs DHCP for what DNS does.
• Popular regions: US-EAST-1 is one of the oldest and busiest regions, and some global features depend on it.
• Retry storms: when services fail, apps automatically try again, which can overload systems that are trying to recover.

Is the cloud a mistake, then?

No. Large cloud providers usually have far better security, engineering and uptime than most organisations could achieve alone. The issue is not the cloud itself but putting all your eggs in one basket. Outages like this one are rare, but because so much depends on so few providers, each one is very visible.

What resilience looks like for organisations

• Use more than one region or availability zone for important services, and test failing over.
• Map dependencies: know which services your app relies on, including indirect ones.
• Design to degrade gracefully: keep a basic version working if a feature fails.
• Have offline procedures for critical tasks, such as taking payments or accessing patient records.
• Communicate quickly through a status page hosted somewhere else.

Our technical notes on geo-redundancy and cloud regions and zones go deeper.

What it means for you

• Keep a small amount of cash and a backup payment card in case an app or payment service is down.
• Know how to use physical keys for smart locks and manual controls for heating.
• Keep important documents such as tickets and boarding passes saved offline or printed.
• Check a service's status page before resetting all your own equipment.

Tip: During big outages, scammers send fake "service restored, log in again" messages. Wait for the app itself to work again rather than clicking links.

Outages are not attacks

When a big outage happens, social media fills with guesses about hackers. In this case, and in most large cloud outages, the cause was an internal fault rather than an attack. Waiting for the provider's official explanation, usually published days later, gives a much more accurate picture than early speculation.