Serving user generated content is very expensive in terms of infrastructure. More expensive in many ways than streaming studio generated content.
The scales of the two models are very different. Ingesting content is more complicated with user generated content because there's few guarantees about formats (encoding, color, file formats). Serving the content is also more complicated because it's not as friendly to edge caches as studio content. Part of the expense of YouTube is the long tail of content. Popular content might live in edge caches but YouTube serves up old unpopular stuff too.
