Also:
1. Occam’s razor: Could it be that a Google is actually getting good value by paying very high salaries?
Yes. The average revenue per person for Google is about 160G$ / 115kiloemployees = 1M4 per employee. I do note that using an average is silly because the value distribution is not flat, but neither is the employee salary distribution flat. But the figure is so large, average does say something useful for back-of-envelope calcs.
2. Can Google monopolise by buying power alone?
No. Other companies have similarly high returns per employee (Apple is 260G$/160kiloemployees, Netflix is 20G$/20kiloemployees) so for Google to outbid them, it needs to outbid above the marginal revenue per employee, which clearly is well above a 600k$ salary.
3. Can Google corner the market for talent by restricting supply?
No. We know that there are plenty of talented developers because Google isn’t the only company making over $1M revenue per employee. Google employs about 115000 people. Let’s say 50000 of those were “overpaid“ to remove them from the competitors. If the total pool of equivalent talent were as small as 250000, then Google couldn’t monopolise talent. Yet other companies with high revenues per employee have a sum total of employees higher than 250000. Furthermore Google’s returns per effective employee become ~$2.8M/employee, so Google can obviously afford to pay $1M for talent it really wants!
Your sports analogy fails because sports are designed to be zero-sum where there can be only one winner, so the best player can capture more of the winnings.
I don't disagree with your other points, just the first.
However making the argument that the marginal gain from an employee has to exceed the marginal cost is much more difficult, so using an average is just a fair proxy for making the point.
My assumption is that Google are smart enough to know extremely well their marginal gain, which must be an upper bound of what they would rationally pay for an employee (salary plus overhead plus opportunity cost plus variance).
Anyone that thinks a Google is employing people at a loss to cause damage to other businesses is a double plus badthinker IMHO.
Your implication is that the pool of talent is limited and that there is some sort of zero-sum game being played.
The pool of talent is clearly not limited and there are multiple indications of that e.g. a rockstar developer would be employed at rockstar prices of millions (or signed on at millions as per your sports analogy).
If Google were to take a talented person out of the pool, the loss to the competitors is the marginal difference in talent to the next lower talented person.
Put simply, the variation in measurement of talent is large, and training can make huge differences to talent levels. Your premise can’t work without some sort of perfect oracle.
That isn’t to say it doesn’t happen at some microscopic level, or that it doesn’t happen with some particular individuals... Edited: there are just too many sensible reasons why I think it couldn’t happen systematically.