Member-only story
Prime Video cut costs 90% by deleting its serverless architecture. Then the post vanished.
The design hit a wall at 5% of expected load. The fix was one process on one box.
The number and the missing page
In March 2023, an engineer on Prime Video’s Video Quality Analysis team published a post explaining that his team had taken a distributed serverless system, collapsed it into a single application, and cut infrastructure cost by over 90%.
It was called “Scaling up the Prime Video audio/video monitoring service and reducing costs by 90%,” and for about six weeks it was the most argued-about engineering post on the internet.
Go to the original URL today and you do not get the post. You get redirected to a generic Amazon entertainment landing page. The text survives in the Internet Archive and in a hundred reposts, but the canonical version is gone from the site that published it.
Someone in the Hacker News thread predicted exactly that within days of publication. They were right.
Why this one travelled so far
Post-mortems about serverless cost overruns are common. This one was different for one reason: the publisher.
