That doesn't address the issue of storage and compute power for streaming to the absurd amount of users.
There's been attempts before and it all comes down to file transfer time and storage (because at the time the servers weren't transcoding for streaming the file. Secondary issue of buy in, like what we see with niche communities staying on reddit instead of moving to the fediverse.
There already exist a number of projects out there like peertube. Take a look at how even the most popular instances are doing. It's not well.
The closest thing was around a decade ago, the popcorntime or popcornflix or whatever it was called app/program that was just a nice front end for torrenting videos and watching them before they finished downloading. Each individual user was responsible for their own storage, network connection speed, and compute power to render the video for themselves. Each end user was also contributing back through helping others to download the file via standard torrenting p2p stuff.
So now you need a front end to host the magnet links to the files, and a robust set of seed servers so no video is ever truly lost. That still doesn't cover a significant portion of youtube's functionality like reccomendations, comments, allowing creators to edit/adjust videos after the fact.
Unlike reddit, youtube is technologically complicated and impressive. Hell, read up on some of the stuff Netflix has had to do to achieve reasonable streaming quality and speed on an insanely smaller curated library.
A decentralized federated solution is possible, but there's a shit ton more that would have to go into this than just appealing to the concept.