Sorry, the same universe "runs" bikes and cars in the real world. It's time to take things seriously here. Any "special" calculations required for bikes would improve the car simulation, incidentally.
Even if they used different models, it's really not hard to make them computationally equivalent, or more, or less, or whatever you want (depending on the fidelity you're going for).
It's still fallacious to suggest, arbitrarily, that running bikes is more computationally expensive than cars, and there is no reason to suggest running both models side-by-side on
different entities (so they won't actually be required "at the same time" at all) would be any different. You can run an X2014J on the same track as other cars, and they use vastly different sound models, for example.
Loading lines of code is not really demanding of memory, and the key quantities being calculated are more or less the same in either case (
universal, even) - there's even two fewer tyre contact patches (and suspension items, and brakes and no ARBs etc. etc.) on a bicycle. PD have already done it, anyway.
But here we are, thrashing out the old fallacy once again.