The processing power of computational devices is limited by their energy supply (Laughlin & Sejnowski, 2003). The division of brain energy use between neuronal resting potentials, synaptic potentials and action potentials is broadly understood (Attwell & Laughlin, 2001), but in general it is unknown how the energetic resources available to the brain are allotted to carry out different parts of a neural computation. We analysed the relationship between energy expenditure and information processing for the cerebellar cortex, a brain area for which the cellular properties and computations performed have been well studied. Most signalling-related energy in the brain is expended on pumping of Na⁺ ions out of the cell (Siesjo, 1978; Attwell & Laughlin, 2001). We estimated the signalling energy expended in the rat cerebellum by using published anatomical and electrophysiological data on the properties of the different cerebellar cells to calculate the Na⁺ influx producing excitatory synaptic currents and action potentials, the Na⁺ influx occurring at the resting potential, and the Ca²⁺ entry driving excitatory and inhibitory transmitter release (Attwell & Laughlin, 2001). As in previous work (Attwell & Laughlin, 2001), these fluxes were converted into values for ATP consumption. We predict the total ATP consumption on signalling (summed over all cell types) to be ~27μmol/g/min, slightly greater than the 20μmol/g/min measured in anaesthetized albino rats (Sokoloff et al. 1977). We estimate that each Purkinje cell uses 1.6×10¹⁰ ATP/s, whereas the much smaller granule cell uses approximately 3×10⁸ ATP/s. However, the much larger number of granule cells results in them consuming the great majority of the cerebellar cortical energy. Thus, most energy goes on granule cells re-mapping the sensory and motor command input arriving on the mossy fibres into a sparsely coded representation used by the Purkinje cells to retrieve motor output patterns. Spatially, signalling energy use was predicted to be split between the granular, Purkinje cell soma and molecular layers in the ratio 30%: 2%: 68%. Comparing the capillary area available for O₂ and glucose supply to the different cerebellar cortical layers suggests that the blood supply to these layers is only approximately matched to the computations carried out in each layer. Our results, together with the calculations of Brunel et al. (2004), suggest that for each Purkinje cell (and associated other neurons and glia) approximately 10¹¹ molecules of ATP/s are used per 5kb of retrievable motor information, corresponding to an energy storage cost of 1mW/GB.