Attwell & Laughlin (2001) modeled grey-matter signaling energy as 47% action potentials, 34% postsynaptic glutamate effects, 13% resting potential, 3% glutamate recycling

Attwell and Laughlin's "An Energy Budget for Signaling in the Grey Matter of the Brain" (J Cereb Blood Flow Metab 21(10):1133, 2001) used anatomic and physiologic data to model the ATP cost of the components of excitatory (glutamatergic) signaling in rodent cortical grey matter. The paper's own abstract states the four-way split directly: "Action potentials and postsynaptic effects of glutamate are predicted to consume much of the energy (47% and 34%, respectively)," with the remainder attributed to maintenance of resting potentials (13%) and glutamate recycling (3%).

The two dominant components — action potentials (47%) and postsynaptic glutamate effects (34%) — therefore sum to ~81% of the modeled signaling energy budget, leaving the resting-potential and recycling terms to account for the rest (a small residual reflects rounding across four modeled categories). The headline of the estimate is that signaling, not baseline housekeeping, dominates the metabolic cost of grey matter, and that within signaling the spike-and-synapse traffic vastly outweighs the standing cost of holding neurons at rest.

This figure is the load-bearing number behind the vault's system-level observation that the brain, like scaled AI, is signaling/inference-bound (claim-brain-inference-bound-like-ai-at-system-level) — the per-event economics reverse between substrate and silicon (claim-synaptic-plasticity-cheap-fraction-of-transmission-energy, claim-competitive-plasticity-reduces-learning-energy) while the system-level ledger converges on "using dominates the budget." The 2001 split was substantially revised eleven years later by an overlapping author team, which lowered the action-potential share and raised the postsynaptic share (claim-howarth-2012-revised-brain-energy-budget-lowers-action-potential-share); anyone citing the "~81%" figure should date it to the 2001 estimate specifically.