A quick scan of Fortune's unicorn list* shows 12 out of the top 25 are from outside of California, and two of the Cali ones are LA, bringing SF down to 11 (after generously counting Palo Alto in SF, because I assume you meant the Bay Area). Without numbers on how many startups are started in various places, I'm not sure what these numbers mean, but it doesn't obviously favor your winner-takes-all argument.
For an area with less than 2% of the country's population, that is overwhelmingly dominant. I'm not sure if you could have underscored my point any more succinctly.
The point is that you still have contributed no idea of what portion of startups are starting in SF. If 95% of startups come out of SF but only 80% of winners do, it still has an overwhelming majority but is by the numbers a terrible place to start a startup.
Of the top 25 international, only about half are American, so stop pretending the denominator is 25 if you've decided that international companies for some reason don't count. There are 2 in LA, one in Boston, and 1 in NYC (what list are you looking at?), giving the Bay Area something like 2/3 of American companies, and about 1/3 of all.
Communication technology hasn't so far enabled, at least in the general case, geographically-distributed tech companies – otherwise we would be seeing an increasing trend of successful tech companies headquartered elsewhere. I don't think we are. Put another way, whatever advantage is conferred upon SF by physically housing an outsized share of tech workers is clearly giving it a competitive advantage that other locations have yet to overcome.
Having a distributed tech team is not hard, but most startups aren't distributed because the VCs aren't distributed. They want face to face meetings.
Sorry, but this is flawed logic.
Tech, more than seemingly any other industry I can think of, has a winner-takes-all model of success. Given this, starting a tech company anywhere other than where the overwhelming majority of the most capable employees live is virtually suicide. And this seems to pan out in the data as well — Twitter, Square, Uber, Google, Facebook, Reddit, AirBnB, Apple, Oracle, Box, Palantir, Pinterest, Lyft, Stripe, CloudFlare… not that these are all of tech, but finding even a fraction of these anywhere (other than perhaps New York) is a stretch.
Given that, it's patently worth it to pay the extra cost per employee, when the relative value of each employee to the company ends up being so high compared to other industries.
This is not to say you can't build a strong tech company in other parts of the US, but you're stacking the odds against you by doing so.