The servers running in AWS are mostly metadata.
For non-streaming video web-torrent is probably the best attempt I've seen as it offloads bandwidth to the viewers.
However the real problem that is impossible to get around is storage. To run a moderately large site the largest cost is storage as you will be seeding a copy of the video.
The few competitors (which are about <1% of youtube's views) are built quite poorly IMO.
You won't beat youtube by copying it. You need to build something new that addresses many of the short comings.
I would wage most people won't even know they are being served by the web torrent mechanism when you hit the page. The video player looks like a regular html 5 video player.
You can just point youtube-dl at a video. However the vast majority of viewers just use the site / phone app.
Hard to compete against a free service like YouTube. You'd need to build out your own infrastructure if you want to deliver videos cost-effectively, you'll have to build your own ad network to make any revenue, and even then you'll need some way to draw both content creators and users to your service to have any chance.
We had some competitors, but Vimeo effectively gave up and is marketing for B2B, and Twitch's video upload feature doesn't work since they haven't pushed hard enough for it (a huge marketing push, some UI work, and exclusivity deals on Twitch's side would go a long way if it can get any initial momentum).