That's peak power, though. Basically no street car is designed to be able to produce peak power for more than a minute or two at a time, and probably very few can do it for more than a few seconds. I'm pretty sure that car doesn't have the cooling capacity to cool itself at full power like this plant would have to.
You can drive your average care at peak power for an hour without problems. The problem with the Veyron is that at top speed the tires will wear off in 15 minutes and you will run out of fuel after online 12 minutes but I think it is not really limited by its cooling capacity - at 400 km/h a LOT of air passes through its 10 radiators.
temperature difference between air entering and leaving the radiator ~ 10degC
==> 1MW
So handling 2.5MW waste heat doesn't seem out of the question from this analysis.
As another ballpark "upper bound" analysis, consider that copper is one of the most thermally conductive materials, with thermal conductivity ~ 400 Wm/(m^2 degC), let's round that up to 1kWm/(m^2degC).
Let's assume that there is a thermal conducting surface of 1m^2 (i.e. the surfaces of the pipes that interface with the radiators). Let's assume that the copper is 1mm thick (= 1m/1000). Let's see what temperature difference would be needed to transfer 1MW of heat across:
As another ballpark "upper bound", the convective heat transfer coefficient for forced air is ~ 100W/(m^2degC), which means that assuming that the radiator fins are 100 degC above the air temperature, in order for the fins to transfer 1MW of power to the air, you would need an area of
1MW == 100W/(m^2degC) * 100 degC * area m^2
==> area ~ 100m^2 of surface area in the radiator, which doesn't seem unreasonable (a radiator 1m^2 area * 10cm deep, made up of thin plates spaced 1mm apart has this total surface area).
So they Veyron dissipating ~ 1MW in waste heat at 400km/h doesn't fail any of these basic sanity tests.
More complicated though is the interaction between the stages considered here. For example, how do we interface our 1m^2 of copper with our 100m^2 of radiator? Making the radiator fins thin lets us pack more surface area into the same volume, but also makes it difficult to keep the "edges" of the fins at a sufficiently high temperature so that they pull their weight transferring heat to the air: since the "copper tubing" has significantly less surface area, it only contacts (and thus transfers heat to) the radiator fins "sparsely".
My point may not have been as strong as I thought, but that tends to support it - airplane and boat engines do need to operate at max power nonstop, so they're designed for it. Consumer automobiles usually just accelerate for a few seconds and cruise at relatively modest speed, and I'm pretty sure their cooling and other related systems are designed around that.
That's because you usually have a speed limit. If you have a car with say about 100 hp your top speed will be about 200 km/h and that is a speed you can drive at for extended period on an Autobahn in low traffic. A more powerful car will of course make it harder to keep the pedal at the metal.
