However, energy is energy whether it's heat or velocity and it can be converted (with some loss) from one form to another. Transferring heat from the radiator to the air flowing through it energizes the airflow by heating it up and changing its density. The air in front of the radiator has less energy than the air behind it. With more overall energy, that heat can be converted into exit velocity, making the outgoing flow faster than the incoming flow and voila, you have thrust. Do it right and the total thrust gained from the exit nozzle is more than the total drag incurred from the entire radiator, ducting and cowling assembly.
That's exactly what North American engineers stumbled on in the belly radiator ducting of the Mustang. Pilots reported that, at the right altitude and airspeed, the Mustang was flying faster than the engine should be able to pull it. That's because the radiator ducting had now turned into a thrust nozzle.
So how does this have anything to do with a car? While thrust is good for straight-line speed, downforce is better for the track. Just like free thrust, if the exit flow is turned up slightly, the thrust pushing forward is now also pushing down. With the right design, this becomes net downforce, where the downforce gained is worth more in terms of lap times than the drag incurred by the radiator.
Designing one that works is a whole other story and is typically better with sleek formula cars with side-pod-mounted radiators that vent up and back. But some road cars can take advantage of this, especially mid-engine cars where the hood doesn't have to clear an engine, and air is forced through diverging ducting in front of the radiator into a converging duct out and back. This is why GT cars all have that large, gaping hole in the hood, even if the engine isn't in front.
While I doubt any of us have the resources to develop net thrust, it's a great concept to play with while testing at the track. The angle of convergence and divergence is critical to prevent flow separation, which results in massive drag and reduced cooling. But some basic calculations and ambitious CFD tests by the industrious closet engineer will get you into the ballpark before testing begins. Did anyone say Project NSX?