This was true when Google was created.
No one had the processing or memory available on their desktop to search an entire index of the "useful" web.
Not anymore.
How large is a "useful" index of the web today? And can it fit on your laptop? The answer is yes.
Can the entire thing be queried fast? The answer is yes.
As an example take the entire stackexchange and wikipedia dumps in their entirety(including images). Compressed it comes to 50-60 GB range. Think about that number. That's an rough approximation of all known human knowledge. It's not growing too fast. It has stabilized. To query the content you need an index.
So how large is an index to a 100 GB file? Generally around 1 GB. Let's say you use covering indexes with lot of meta data and up that to 5GB to support sophisticated queries.
With today's average hardware you can search the entire thing in milliseconds.
So why aren't we building better local search?
Because everyone is conditioned to believe, thanks to Google's success, we need to do it online. Which means baking in the problem of handling millions of queries a second into the Search problem. Guess what? This is not a problem that local search has.
Every time a chimp or a duck needs to build a protein in it's cell it doesn't query the DNA index stored in the cloud. Instead every cell has the index. Every cell has the processing power to query that index in the nanosecond time scale.
The cloud based search story is temporary.
If you want to index every reference to Taylor Swifts ass that every teenager in Norway, Ecuador and Cambodia are making, then yes you need a Google size index. But for useful human knowledge we are getting to the point where we don't need Google scale.
If you don't believe me look at what is possible TODAY in Dash/Zeal docset search for offline developer documentation search or Kiwix or with Mathematica.