Yesterday was a great day for anyone trying to finish their laundry on company time, but a frustrating one if you were trying to do pretty much anything else. Amazon Web Services (AWS) was down, with rippling outages rolling through nearly a third of the internet and affecting thousands of businesses around the globe. In the evening, about 15 hours after the trouble began, Amazon said it had “returned to normal operations.”

What happened? Around 3am ET, outage reports started cropping up. Amazon blamed an error in the DNS system, which does the actual connecting to the site when a user clicks on a link or app. The error came after Amazon pushed a technical update to a database service. Because AWS undergirds ~30% of the web, the problem had a global impact.

Even as Amazon said it was starting to get things patched up in the morning, the AWS outage continued, peaking around midday, according to Down Detector.

Spinning pinwheels: Banks and financial services apps like Venmo and Robinhood weren’t working. Streaming and gaming sites like HBO Max, Tidal, Roblox, and Fortnite struggled. The Snapchat app was no use all day, and even Wordle wasn’t available.

Amazon also felt the pain:

• Its own e-commerce site and services like Ring and Alexa were reported to be offline.
• Early estimates for how much this outage is going to cost top out around $75 million per hour—with $72 million per hour coming from Amazon itself while its site was down, according to the UK-based UX design firm Tenscope (though AWS generated $107.6 billion in revenue last year, so it can likely afford the hit).

Time to say “we told you so.” Experts are using this massive disruption to point out that this is what happens when most of the internet relies on a few major providers. AWS only has two main competitors: Microsoft’s Azure and Google’s Cloud Platform, so if any of them have even a slight hiccup, it becomes everyone’s problem. And it’s not an easy problem to solve, as very few companies could build the infrastructure it takes to power even a fraction of what AWS handles.—MM