Attwell & Laughlin (2001) calculate rat neocortex to use 40 umol of ATP per gram per minute. The average human brain is 1500 grams, about half of which is gray matter (ref ). And ATP is converted to energy at 30.5 kJoule/mol (ref), yielding a human energy wattage for human gray matter of

40 umol * (1 min/60 sec) * 30.5 kJ/min * 1500/2 = 15.2 J/sec = 15.2 W

This is perfectly reasonable, given that a human at rest uses 60-100 W and the brain uses about 20% of that.

The actual number will be somewhat smaller, since rats have a higher density of neurons (cells/gram) than humans. But on the other hand, human cells are therefore bigger and have more membrane area and synapses per cell to consume energy... so that larger-area effect partially cancels the lower-cell-density effect.

Peter Lennie (2003) uses the Attwell & Laughlin calculations to extrapolate to humans, by tweaking the average size and density of neurons, their synaptic effects, larger brains, etc., along with measured energy use by human cortex. He has lots of unimportant calculations about the fractions of neurons which could be "active" under various circumstances, but the important result of the paper is encapsulated in a single pair of numbers: that a probable average spike rate is 0.16 Hz (once per seven seconds!), and a maximum of 1 Hz.

My conclusion: Average spike rates in monkeys and humans must be in the sub-1-Hz regime. And physiologists ought to measure the average spontaneous activity of the cells they choose for recording, to ensure that the chosen cells are representative of typical cortical cells.