The Semantic Web depends on having everyone who writes web pages also categorize them. Whenever you have 20 different people categorizing content you're going to get 20 different interpretations of what the semantics mean. The vision of a decent, consistent categorization emerging across the web is a pipe dream.
Furthermore the fundamental problem that Google solves is not finding content. It is narrowing in on worthwhile content. In a world filled with people pushing commercial content with a dubious value proposition, you can't depend on content creators to provide an accurate assessment of their quality. That's where Google comes in. They not only categorize, they rank data well enough that end users are happy to accept the results.
I think a lot of the SW hype's overblown - and I've worked in the field - but you're attacking something which doesn't exist.
Sure, their immediate ad-driven revenue model might break, but I very much doubt this would be an overnight change, nor one Google is not already thinking about.
If web pages and web services were to all completely expose their information as RDF and embed rich metadata inside bodies of text the need for Google's fancy AI and Entity extraction would be none.
Google's bread and butter is advertising. That advertising is driven from their search domination - their ability to extract meaning from web pages and match that meaning up to searched terms. Semantic web completely commoditizes that function - if you can crawl, you can search.