Significant transformations are ongoing in automotive powertrains, owing to the restrictive regulations, established by governments around the world, which limit the CO2 emissions into the atmosphere. Car manufacturers are developing, therefore, alternative powertrain systems. Hybrid/Electric vehicles and Fuel Cells powered vehicles are spreading and advanced internal combustion engines technologies are being developed. A common need of these powertrain systems is an effective cooling system. For ICEs, an optimized thermal management allows a significant reduction of frictional losses during warm-up, which is expected to contribute by about 3% to the total CO2 decrease in a homologation cycle. For hybrid/electric vehicles, the power electronic components have specific cooling requirements. Current technology, in fact, is based on the third-generation of insulated-gate bipolar transistors (IGBTs), which, nowadays, produce a heat flux of about 100-150 W/cm2 and is expected to increase to ~500 W/cm2 in the next generation. In Proton Exchange Membrane Fuel Cells (PEM FCs) a close correlation exists between thermal management of the stack and operating performance. Furthermore, it is acknowledged that boiling flow allows a better thermal homogeneity and higher heat transfer rates. In the present work, a methodology for the optimal thermal management of different powertrain devices is proposed, which is based on the careful monitoring of nucleate boiling. The methodology makes use of Model Predictive Control by means of a zero-dimensional model for the heat transfer between the device and the coolant, which defines a metrics for the occurrence of nucleate boiling. The introduction of an electrically driven pump for the control of the coolant flow rate is required. The effectiveness of proposed approach will be presented with reference to an ICE operation. Experimental tests show the advantages of the methodology during warm-up, under fully warmed operation and for the avoidance of after-boiling.