Forty-nine adult Nellore cows (32 pregnant and 17 nonpregnant) with average initial body weight of 451 ± 10 kg were used in a comparative slaughter study aiming to describe equations and relationships for prediction of net, metabolizable and dietary energy requirements for adult, pregnant and nonpregnant, Bos indicus cows. Feeding control was measured individually and cattle were fed either HIGH (ad libitum) or LOW (restricted feeding 1.2 times maintenance according to the NRC). The 32 pregnant cows were separated at random into 4 groups of 8 cows each (4 cows per each feeding level) and harvested at 136 ± 1, 189 ± 1, 239 ± 1 and 269 ± 1 d of pregnancy. The nonpregnant cows were harvested at different times of the experiment (85 to 216 d of feeding control) to keep them in experiment for a similar amount of time as the pregnant cows. The digestible energy and losses of energy as methane and urine were directly measured to establish the relations between GE, DE and ME. Energy content was analyzed in empty body and pregnant compounds. A set of relationships and equations based in the factorial method from ARC was used to estimate the nutrient requirements. The net energy requirement for pregnancy (NE_p) estimated in this study was about 3/4 of those estimated by NRC. When estimated by a logistic model, the daily requirements for pregnancy showed an exponential increase up to 250 d of gestation and then decreased. However, when an allometric model was used to estimate the daily requirements for pregnancy, the maximum daily requirements were at birth. There were no differences in the dynamics of energy (P = 0.388) in the cow’s empty body weight pregnant free (EBW_np) suggesting that the pregnancy does not affect the requirements for accretion of body reserves in cows. The partial efficiencies for use of metabolizable energy for maintenance, weight gain and pregnancy (k_m, k_g and k_c) were respectively 70, 53 and 12%. The efficiency of transformation of DE in ME was 0.80.