There is an increasing interest for genetic selection of dairy cows to improve the efficiency of feed utilization, but the progress has been limited due to difficulties in measuring feed intake in practical dairy farms. Determination of CO₂ production could be an alternative approach to direct DMI measurements for evaluation of feed efficiency (FE), provided that CO₂ production is closely related to heat production (HP) estimated by Brouwer equation. The objective of this study was to evaluate the potential of CO₂ production in estimating HP and consequently FE. The relationships were estimated from respiration chamber data in 100 cows fed a diet based on grass silage and concentrate (55:45 on DM basis). Residual ECM (RECM) yield was calculated from ME intake and requirements for maintenance and production. Residual CO₂ per kg ECM (RCO₂) was calculated by adjusting observed values for differences in ECM yield (−3.75 L/kg ECM) to take into account dilution of maintenance. Feed efficiency was expressed as ECM/GE or ME (g/MJ). PROC MIXED of SAS was used to develop models with period as random factor. There was a close relationship between HP and CO₂ production: HP (MJ/d) = 6.8 ± 1.74 + 0.0185 ± 0.00022 × CO₂ (L/d); adjusted RMSE = 1.2 MJ/d (0.8% of observed mean). This suggests that accurate measurements of CO₂ production can be a useful tool in ranking the cows according to feed efficiency. Residual ECM yield (kg/d) was negatively related to RCO₂ (L/kg ECM): RECM = −0.21 ± 0.80 − 0.173 ± 0.0236 × RCO₂ (RMSE and R² adjusted for random period effect = 2.8 kg/d and 0.64, respectively), i.e., cows producing more CO₂ were less efficient. Also, gross FE (g ECM/MJ GE) was closely related to CO₂/ECM: FE = 441 ± 11.7 − 0.856 ± 0.054 × CO₂/ECM (adj. RMSE = 9.6 g/MJ, R² = 0.88). When the cows are in negative energy balance CO₂ production per unit of ECM can be less compared with cows at zero energy balance due to greater efficiency of converting body energy to milk energy compared with dietary ME. To avoid false ranking of cows for FE, measurements of CO₂ should be made in mid-lactation and/or some proxies (e.g., blood NEFA or milk fatty acids) to identify cows in negative energy balance.