Energy costs of neural communication and computation

Previous research has suggested that the human brain consumes an estimated 20 watts of glucose.

The researchers attempted to uncover how energy is allocated to various functions carried out by the human cerebral cortex, which is the outermost layer of neural tissue in the brain.

The results revealed that cortical computation consumes only 0.1 watts of adenosine triphosphate (ATP), the brain’s energy currency. By contrast, the energy cost for long-distance communication between neurons is 35 times greater, or 3.5 watts of ATP. The authors also calculated that the total energy cost of the human cortex is 4.9 watts of ATP.

Therefore, communication costs account for 71% of the total energy budget of the human cortex, compared with 2% for computation. However, both costs are required to predict the number of synapses, or junctions between neurons.

Additional results suggest that neuronal computation is 108 times less energy efficient than the optimal limit ordained by physics.

According to the authors, the energy costs of communication might be necessary to ensure adequate speed of neural signaling and information processing. 

https://www.pnas.org/content/118/18/e2008173118