Expanding Josh's comments above...
The AC power consumption is (usually, and the minimum required by the safety regs), the 1/8-rated audio power AC power consumption. This assumes maximum undistorted audio.
This figure is also the average number of BTUs drawn from the wall during average use. Now, multiply this number by 20% efficiency and this is the number of BTUs that must be dissipated as heat (average) Because there are some short term thermal excursions, normally we double this number to be sure we never get into a runaway condition.
Removing the heat depends in part on the surfaces radiating the heat and this is a non-linear function because heat transfer increases as the temperature differential grows. The balancing act is that in some instances there are de-rating factors that must grow with temperature so we like to limit the temperature rise of critical areas to avoid unnecessary de-rating.
Using the Subway amp and an example, the average power is 400 watts (this is because the amp is designed to drive and support greater than the bare minimum 1/8-rated audio power, in fact the number is slightly greater than 1/3 which is a substantial difference) so the heat dissipated is roughly 400 x 0.2 = 80 watts, but we also design around an environmental temperature of ~120 deg F and though the AVERAGE temperature of the air inside the chassis may only be ~140 degrees (20 deg delta T is a good rule of thumb for air temps), and there must also be enough air passing over specific components and heat sinks to limit the surface (and die) temperatures and must overcome any static pressure drop through the system. This is where the two theories kind of diverge.
So, the 80 watts represents the average thermal energy EXCHANGED from the cabinet, or 80 watts x 3.4 BTU/hr = 272 BTU/hr and the CFM would be (272 x 1.08)/20 = ~14.5 and generally the static pressure for the Subway and is less than the rated SP for the fan so this is generally the worst case.
Now that we know the worst case conditions, measurements of critical spots are made and we will then fine tune the numbers based on the actual air flow paths and the temperatures of the critical components under these worst case conditions.
This is a good example of why just substituting a quieter fan that ends up with a lower CFM and/or rated static pressure can get you into big trouble. A lot of thought goes into a reliable cooling system.
