Artificial metalloenzyme, composed of metal complex(es) and a host protein, is a promising way to mimic enzyme catalytic functions or develop novel enzyme-like catalysis. However, it is highly challenging to unveil the active site and exact reaction mechanism inside artificial metalloenzyme, which is the bottleneck in its rational design. We present a QM/MM study of the complicated reaction mechanism for the recently developed artificial metalloenzyme system, (Rh(nbd)·apo-Fr) (nbd = norbornadiene), which is composed of a rhodium complex [Rh(nbd)Cl](2) and the recombinant horse L-chain apo-Ferritin. We found that binding sites suggested by the X-ray crystal structure, i.e., sites A, B, and C, are only precursors/intermediates, not true active sites for polymerization of phenylacetylene (PA). A new hydrophobic site, which we name D, is suggested to be the most plausible active site for polymerization. Active site D is generated after coordination of first monomer PA by extrusion of the Rh(I)(PA) complex to a hydrophobic pocket near site B. Polymerization occurs in site D via a Rh(I)-insertion mechanism. A specific "hydrophobic region" composed by the hydrophobic active site D, the nonpolar 4-fold channel, and other hydrophobic residues nearby is found to facilitate accumulation, coordination, and insertion of PA for polymerization. Our results also demonstrate that the hydrophobic active site D can retain the native regio- and stereoselectivity of the Rh-catalyzed polymerization of PA without protein. This study highlights the importance of theoretical study in mechanistic elucidation and rational design of artificial metalloenzymes, indicating that even with X-ray crystal structures at hand we may still be far from fully understanding the active site and catalytic mechanism of artificial metalloenzymes.