The amount of power needed would really depend on the device and the application, I guess.
How many mW are we talking about, in the best case scenario ?
http://electronics.stackexchange.com/questions/4664/stealing...
One of the answers has the following math: For a large radio tower (100kW), you'd get roughly 80 W per square meter of receiving antenna area at 100m distance.
Since the typical distance to a nearby radio tower is probably more like 5 miles, we can reduce that to ~12mW for a 1m^2 antenna. For the size of a phone, we're talking more like 0.10mW, which is about 10,000 times less than a typical phone needs (assuming it was charging 24/7 with 100% efficiency).
Although, the answer gives the formula as 100kW/(4 * pi * r^2), which seems to give 0.80 W/m^2 instead of 80 W/m^2, so perhaps divide my other answers by 100 (and make that 1 million times lower than the power necessary to power a phone)
EDIT: I previously stated the formula gave the wrong units which was a mistake.
The biggest problem is going to be PR. So wifi milliwatts of microwave power tens of feet away make hypochondriacs and conspiracy theorists think the government is controlling their thoughts using radio waves... I'm guessing "tens of watts in a ISM diathermy band" is going to substantially encourage them.
For a non-directional beam, it would be P/(4.pi.r^2).
Perhaps the steering could be done cooperatively with the device(s) being charged. I.e., devices announce their location to the charger, and the charger steers the beam towards them.
Maybe that would bring the power required down to a reasonable range.
Wireless power is something that has been researched for years I believe, the fact that to this day there is not a feasible solution (that I know of) means that it's a really hard problem, I guess.