I suppose I forgot to state my original point. Let's hold all else equal, and look at the labor pool divided into two categories, urbanism and ruralism. Suburbs lie on the ruralism side for this comparison. The labor pool is of course not evenly divided in terms of preference, say maybe 3/4 of protential Google employees prefer urbanism. If a given employee prefers urbanism, and lives in one of the cities mentioned above (NYC, Yay Area, Boston), their resulting employee satisfaction (what the company cares about in this scenario) is improved. However, if a given employee prefers ruralism, their resulting employee satisfaction is worsened. Vice versa for a more suburban, smaller city.
Of course there is some cost to not collocating employees. Google has decided new offices are worth this cost, and thus Google can decide to allocate building/location resources based on employee preference alone (let's assume the other factors are somehwat insignificant, which could be valid taking into account the high proportion of labor cost).
If every additional dollar spent on a given location buys the same amount of office space and equipment in both the urban and suburban areas, then it would make sense to divide the budget for offices according to the proportions of the labor pool preferences. Now of course, a dollar in SF buys a lot less office space than a dollar in Lexington, so that should be a accounted for in budgeting decisions, as well as overhead costs involved in opening a new location.
If we look at Google's actions under this framework, it doesn't make much sense for Google to devote as much of their office budget to highly urban areas as they do, unless the proportion of current and potential employees that prefer ruralism is very low. I would find that difficult to believe. Of course there are other factors to consider, but the trend is what I mean to highlight.
That being said, the true proportion would have to be discovered, and company culture may not allow people to express their true desires. This would apply to many tech giants.