Bandwidth is a bill that never stops arriving

Social media has normalised heavy content. Video and images loaded in bulk are now the default, and the cost of keeping systems running under that kind of media load is a widespread problem. Today the conversation has moved on to how to meet the GPU and cooling demands of AI.

That same appetite for heavy use has spread to media that were never designed for it, and the web is one of them. Ideally a page should load around 2 MB on first visit, but video and high-resolution photography make that hard to hold. The usual answer has been to defer loading media until the visitor scrolls to it: if it never appears on screen, it never loads, and the initial load stays under control.

The problem is then bandwidth. Bandwidth is the maximum amount of data that can travel through a connection in a given period of time. Hosting platforms like Vercel or AWS bill for usage across several parts of the stack, and bandwidth is one of them. It is common to run up large bandwidth costs when video or images are hosted on your own servers, either because the files are heavy or because traffic spikes.

Controlling that cost is a fundamental part of good technical direction. The essential steps are:

1. Compress images properly, in formats such as WebP.
2. Host video on a third-party platform, so you are not paying to deliver it yourself.
3. Cache well, so the server is not queried again for the same thing.
4. Monitor performance and traffic, so any leak shows up early.

If costs suddenly climb, the answer is to go back through those four steps and find which one has slipped.