Tim Oren adds a VC perspective to my Subsidized Wi-Fi AP idea.  He wrote:

[…] the average usage per Starbucks site is three sessions per day. […] No one in any of those models is yet profitable, meaning that a subsidy is just a way of going further into the red.

The AP price is a factor in the initial setup of a public hotspot, but labor costs of the install likely equal or exceed it, unless the location owner is technically sophisticated. On an ongoing basis, the costs will be dominated by maintenance (truck rolls), but even more by backhaul costs.

His conclusion is:

subsidies are a weapon for carriers to buy market share in a period of fast growth. Lacking this growth, and a confidence that new sites will be profitable, a subsidy model isn't likely.

I agree that profitability is a problem.  While reading Tim's post, I thought of the vending machine business.  Where and how a vending machine is placed is similar to the problem of where and how a W-Fi AP gets installed.  No single business model will work in all situations.  Some store owners run their own vending machines, some even pay out of pocket to have them installed, some shares profit.

Thinking of the vending machine business leads naturally to the idea of vending-machines with built-in Wi-Fi AP.  Vending machines are placed where people are.  Power is needed so AP has power.  Connectivity is desirable for real-time inventory, so some aggressive players might opt for pulling DSL lines to vending machines instead of using radios.  It's a stretch, of course.