Human brain is a complex system with powerful abilities of signal processing, which determines our cognition, emotion, consciousness, and behavior. As a computational device, our brain needs to continuously consume metabolic energy to realize above functions. Most of brain's energy usage is consumed on information coding by single neurons, and the subcellular processes consuming metabolic energy include generating and propagating action potentials, maintaining rest potentials, and synaptic transmission. A neuron uses sequences of action potentials as a principal carrier to represent and transmit information. Generating and propagating these electrical signals makes a significant contribution to the overall consumption of metabolic energy in the brain. The biophysical properties of voltage-dependent ionic conductances determine the action potential energy consumption. The cell specificity and spatial heterogeneity of biophysics lead to a high variation in the action potential metabolic efficiency, which brings challenges for understanding the principles, causes, and consequences of metabolic cost of information coding by single neurons. In this paper, we first introduce how the subcellular processes involving in neuronal information coding consume metabolic energy. Then, we present an exhaustive review on the main findings of action potential metabolic cost in recent years, and mainly discuss how biophysical properties and spike shape affect the action potential energy efficiency. Finally, we raise several key issues on the metabolic consumption of neuronal information coding that need to be addressed in the future.