Most cows undergo a state of negative energy balance in early lactation. Body reserves then play a major role in milk fat production, whereas less fatty acids (FA) are synthesized de novo in the mammary gland. This is reflected in the milk FA composition. The objective of this study was to validate if the milk FA profile can be used to predict body weight change (BWC) in early lactating cows in commercial dairy farms. Data originated from 17,067 Danish Holstein cows of 7–35 d in milk across 166 herds in Denmark between March 2015 and March 2017. Body weight was recorded by floor scales in Lely automatic milking systems at each milking. Dairy Herd Improvement milk samples were analyzed by Fourier-transform infrared spectroscopy. Data for BWC prediction included stage of lactation, parity, and test day data for milk production and components (fat, protein, somatic cell counts, and FA concentrations). Daily median BWC was −0.43 (−8.80 to 5.53; 1st to 99th percentile) g/kg of body weight across lactations. Predictions of daily BWC were based on a random forest algorithm with 10-fold repeated cross-validation for which 20% of the herds were randomly withhold for cross-validation. The overall root mean square error of prediction was 1.66 g/kg of body weight at a coefficient of determination after cross-validation of 0.896. The 5 most important variables considered by the model were the short-chain FA group (C4:0–C10:0), C18:1, the medium-chain FA group (C12:0–C16:1), the saturated FA group, and C16:0. While short-chain and some medium-chain FA are synthesized in the mammary gland, C18:1 originate from body reserves (e.g., during energy deficits), and C16:0 and C16:1 derive from either the mammary gland or body reserves. These results suggest that the milk FA profile may be used as an early indicator of BWC in early lactation. Further work is needed to validate the model with North American data, and to investigate possible associations with production, fertility, and health traits.