Any simulation must cause a passably similar analog of whatever it simulates, which in our case is being in a universe governed by the law of physics. However, just to store the value of each calculated variable, even in a one bit representation, will require its own memory register. It would thus take the numerical equivalent of a whole universe of subatomic particles in discrete memory registers just to store one bit of information on every such particle, let alone calculate those bits or do anything with the results of those calculations.
It will of course be objected that not every subatomic particle in a remote galaxy would need to be individually represented, even if simulated scientists might use quantum methods to generate random numbers to decide which remote galaxy they shall wormhole to in order to check the state of a randomly chosen subatomic particle in a random star.
But while that could be good for reducing hardware requirements, something else is rather bad for it, because the perceptual effect of countless objects governed by quantum and relativistic phenomena cannot be generated in a coherent way, even to Newtonian approximations, without being calculated in advance or real time with many if not all subjective and simulated objective variables featuring in many if not all of each other’s simultaneous equations.
So even if every decision made by scientists is entirely scripted in boring ways such that only a tiny bit of quantum physics or astronomy is ever done, simulating any universe well enough to cause every interrelated subject to experience a lifetime of coherent experience as of an objectively existing universe governed by the laws of physics is an engineering problem that would appear to require more resources to solve in practise than any non-simulated planets or amalgams thereof could suffice to provide.
Yet all simulation must ultimately be run by some objective existence.
Hence we are not living in a simulation.