Major parts of that aren't true. I think something like 30% of queries have never been seen before, and Google provides low latency on those as well.
Cut him some slack, it was clearly a simplification. Also, you're wrong to declare him inaccurate.
Even if 30% of queries are truly unique (a number that people have quoted from years ago, may no longer be true) caching 70% of the queries is a big win. Also, instant search is heavily relying on Google Suggest, which by it's very nature, are queries which have been performed already, so are trivially cacheable.
I've always used lengthy, quoted search phrases, but what's really interesting is that Google's autocomplete has slowly trained monkey-me to follow the well-worn path and use the search terms it suggests.
It's a great example of using the UI to encourage a desired behavior.
So, let's just turn off caching and triple or quadruple the size of the Google cluster, eh? No big deal. I completely fail to see how you think that disproves my point.
I should probably mention that I work in search quality at Google. The part I took issue with is "Google can take a couple of seconds if you throw a truly novel query at it." which is definitely not true.
Google can't cache nearly as much as you would expect, not because the queries change, but because the results change.
Of course, storing millions of the "common queries" isn't a joke either and that still doesn't answer the question of how they search the internet within a few seconds anyhow; I'm not trying to give a complete answer, just point one aspect out. Google Instant, by design, provides the Google Suggest queries the vast bulk of the time and I guaranteed that those are either largely or possibly even entirely pre-cached, so you're not doing 10 web searches in two seconds, you're doing 10 hash lookups in 2 seconds. Which is still damned impressive, I can't hardly keep my web site loading in less than two seconds for one hit where I work.