Accurately identifying pregnancy status is imperative for a profitable dairy enterprise. Mid-infrared (MIR) spectroscopy is routinely used to determine fat and protein concentrations in milk samples. MIR has also been used to predict fatty acid content, mineral content, body energy status, lactoferrin and methane emissions. Machine learning has been used in a variety of fields to find patterns in vast quantities of data. This study aimed to use deep learning to establish pregnancy status from MIR data. Milk spectral data were from 2 sources: the extensively phenotyped Langhill research herd and National Milk Records (NMR) where there is a large volume of data. The spectral data were standardised and spanned 3–50 μm. First, predictions were trained on the Langhill herd and then used to predict on a subset of Lanhill data. Subsequently, a prediction model was trained on NMR data and predicted on Langhill data. Three neural networks were tested. First was with no hidden layers (ANN) for linear classification and as benchmark, second was a fully connected dense network (DNN) and third was a convolutional, backpropagation network (CNN). Training accuracy, validation accuracy was highest with the CNN and also had the lowest loss value. The CNN fully trained model (training accuracy 0.8951 and loss 0.2967) predicted pregnancy status on 5,371 spectral records consisting of 45 lactations where the animal lost a calf during that lactation. Each record consisted of 225 spectral points. The model successfully predicted 82.61% onsets of pregnancy with an average of 16.233 d after insemination. The model could also identify 73.33% of cases where a cow lost a calf during gestation. Predictions after 350 d in milk became less accurate with most cows having a at least one low probability for pregnancy. Results showed that MIR data contains features relating to pregnancy status and the underlying reproductive hormonal changes in dairy cows and can be identified by means of deep learning. Prediction equations from trained models can be used to alert farmers of nonviable pregnancies as well as verify conception dates.