peer reviewed ; Recent development of equations based on mid-infrared (MIR) spectrometry for the prediction of milk fatty acid (FA) contents allows their measurement on a large scale during performance recording. The objective was to show that a genetic evaluation for milk FA in dairy cattle is feasible in the Walloon region of Belgium and to report first results. Estimated breeding values (EBV) and associated reliabilities (REL) were computed using a multi-trait test-day animal model similar to the one used for the routine genetic evaluation for yield traits. Studied traits were first lactation test-day milk, fat and protein yields, fat (FAT) and protein contents, and content of saturated fatty acids in milk (g/100g of milk, SAT). More than 6,700,000 records were available for common production and content traits and 194,000 records were used for SAT. Used variance components were estimated using REML. The average SAT content was 2.79% with a standard deviation (SD) of 0.50%. A total of 1,707 Holstein bulls used in Walloon Region had REL superior to 0.49 for all studied traits. REL for SAT ranged from 0.53 to 0.99. A total of 1,217 bulls had REL superior to 0.74. SD of EBV for SAT was 0.20%. The maximum and minimum SAT EBV values were 0.89% and -0.69%, respectively. In order to have a direct measure of the part of FAT that is not due to SAT, a new trait (dSAT) was post-evaluated and defined as difference between expected SAT EBV for a given FAT EBV and the estimated EBV for SAT. This new trait can be assumed to be a direct predictor of the content of unsaturated fatty acids in fat. The interest is that this trait cannot be accurately predicted directly by MIR. The maximum and minimum EBV for dSAT for the 1,707 bulls were -0.28% and 0.24%, respectively. Based on these results, a genetic evaluation for milk fatty acids is feasible. In the bull population used recently, a genetic variability for dSAT exists and could be used to improve the milk fat composition.