assigning, by a computer, an overall energy efficiency rating to a deep neural network model based on utilizing software optimization and hardware accelerators during training of the deep neural network model;

identifying, by the computer, power requirements of each of a plurality of hardware components in each edge device of a plurality of edge devices that comprise an edge inference computing environment;

selecting, by the computer, a predetermined number of hardware components of the plurality of hardware components in each edge device that contributes most to energy consumption on each edge device based on an identified power requirement of each respective hardware component of the plurality of hardware components in each edge device;

assigning, by the computer, an energy weight to each of the predetermined number of hardware components that contributes most to the energy consumption on each edge device based on the identified power requirement of each respective hardware component of the plurality of hardware components in each edge device;

generating, by the computer, an energy score for each edge device of the plurality of edge devices in the edge inference computing environment based on the energy weight assigned to each of the predetermined number of hardware components that contributes most to the energy consumption on each respective edge device;

selecting, by the computer, particular edge devices of the plurality of edge devices that have the energy score within a defined edge device energy score range for the overall energy efficiency rating that corresponds to the deep neural network model; and

deploying, by the computer, the deep neural network model to the particular edge devices that have the energy score within the defined edge device energy score range for the overall energy efficiency rating that corresponds to the deep neural network model.