Recently Evolvable Hardware (EHW) is widely studied to design effective hardware circuits that can reconfigure themselves according to the environment. However, it is still difficult to apply for complicated circuits because the search space increases exponentially as the complexity of hardware increases. To remedy this problem, this paper proposes a method to evolve complex hardware with a modular approach. The comparative experiments of some digital circuits with the conventional evolutionary approach indicate that the proposed method yields from 50 times to 1,000 times faster evolution and more optimized hardware.