Functional magnetic resonance spectroscopy (fMRS) allows the non‐invasive measurement of metabolite concentrations in the human brain, including changes induced by variations in neurotransmission activity. However, the limited spatial and temporal resolution of fMRS does not allow specific measurements of metabolites in different cell types. Thus, the analysis of fMRS data in the context of compartmentalized metabolism requires the formulation and application of mathematical models. In the present study we utilized the mathematical model introduced by Simpson et al. (2007) to gain insights into compartmentalized metabolism in vivo from the fMRS data obtained in humans at ultra high magnetic field by Mangia et al. (2007a). This model simulates brain glucose and lactate levels in a theoretical cortical slice. Using experimentally determined concentrations and catalytic activities for the respective transporter proteins, we calculate inflow and export of glucose and lactate in endothelium, astrocytes, and neurons. We then vary neuronal and astrocytic glucose and lactate utilization capacities until close correspondence is observed between in vivo and simulated glucose and lactate levels. The results of the simulations indicate that, when literature values of glucose transport capacity are utilized, the fMRS data are consistent with export of lactate by neurons and import of lactate by astrocytes, a mechanism that can be referred to as a neuron‐to‐astrocyte lactate shuttle. A shuttle of lactate from astrocytes to neurons could be simulated, but this required the astrocytic glucose transport capacity to be increased by 12‐fold, and required that neurons not respond to activation with increased glycolysis, two conditions that are not supported by current literature.