We present results on the energy balance of the Deep Impact experiment based on analysis of 180 infrared spectra of the ejecta obtained by the Deep Impact spacecraft. We derive an output energy of 16.5 (+9.1/−4.1) GJ. With an input energy of 19.7 GJ, the error bars are large enough so that there may or may not be a balance between the kinetic energy of the impact and that of outflowing materials. Although possible, no other source of energy other than the impactor or the Sun is needed to explain the observations. Most of the energy (85%) goes into the hot plume in the first few seconds, which only represents a very small fraction (<0.01%) of the total ejected mass. The hot plume contains 190 (+263/−71) kg of H 2O, 1.6 ± 0.5 kg of CO 2, 8.2 (+11.3/3.1) kg of CO (assuming a CO/H 2O ratio of 4.3%), 27.9 (+25.0/−8.9) kg of organic material and 255 ± 128 kg of dust, while the ejecta contains ∼10 7 kg of materials. About 12% of the energy goes into the ejecta (mostly water) and 3% to destroy the impactor. Volatiles species other than H 2O (CO 2, CO or organic molecules) contribute to <7% of the energy balance. In terms of physical processes, 68% of the energy is used to accelerate grains (kinetic energy), 16% to heat them, 6% to sublimate or melt them and 10% (upper limit) to break and compress dust and/or water ice aggregates into small micron size particles. For the hot plume, we derive a dust/H 2O ratio of 1.3 (+1.9/−1.0), a CO 2/H 2O ratio of 0.008 (+0.009/−0.006), an organics/H 2O ratio of 0.15 (+0.29/−0.11) and an organics/dust ratio of 0.11 (+0.30/−0.07). This composition refers to the impact site and is different from that of the bulk nucleus, consistent with the idea of layers of different composition in the nucleus sub-surface. Our results emphasize the importance of laboratory impact experiments to understand the physical processes involved at such a large scale.