What was going to be the “Next big thing” of web dev but failed?

Structured data. There were a bunch of flavors (web schemas, semantic markup, micro formats) but they've all pretty much failed. They didn't fail in the sense of not working (they're great and they do what they're supposed to really well) but they've not been adopted to an extent where they'd actually be useful for anything besides very basic SEO.

CSS + HTML5 the goal was you were going to write your document in semntic HTML5 and then use CSS to generate beautiful web pages for any screen or any device... need something printed? no problem print.css has you covered.

But as the web has shown it's almost impossible to generate a good looking website without a bunch of div sprinkled all over your html.