Lead-based halide perovskites with the chemical formula of APb(II)X3 (A = Cs, CH3NH3, or CH2NH=CH2; X = I, Br or Cl) have attracted enormous interest for solar cell applications, primarily due to the high photoelectric conversion efficiency (PCE) up to 19.3%. However, the stability and toxicity are the major issues restricting the commercialization of these lead perovskites based PVs. ASn(II)X3 perovskites are lead-free, but they are extremely sensitive to the ambient air. To resolve the issues, an air-stable class of perovskite materials with the chemical formula of A2Sn(IV)X6, a defect variant of the ASn(II)X3, has been very recently introduced to PV applications. Particularly, Cs2SnI6 exhibits an ideal bandgap of 1.26 eV, high carrier mobility, and have demonstrated high PCE up to 7.8%. On the other hand, the electronic structure and the chemical bonding nature of Cs2SnI6 have not been understood properly; i.e., Sn in Cs2SnI6 is expected to be the +4 charge state upon assumption of Cs+ and I-, which would lead to expectation that the conduction band minimum is formed by Sn 5s similar to SnO2. Here, we performed hybrid density functional theory calculations with the HSE06 functionals for Cs2SnI6. We found that the band gap is formed mainly by anti-bonding and bonding states of I 5p ï¿½ I 5p slightly hybridized with Sn 5s because the Cs2SnI6 structure is composed of isolated Cs ions and [SnI6] clusters. Unexpectedly, the Sn 5s state is very deep, ~7 eV deeper than the I 5p valence band (VB), and has little contribution to the VB; while, Sn 5s has more contribution to the conduction band. As a consequence, the charge state of Sn is not +4, but much closer to +2 like in CsSnI3 and SnO. Then, the apparent charge state of I is a bit smaller than -1 and to be of -4/6, which difference comes from the direct covalent bonds between the adjacent I atoms in the [SnI6] cluster. We also studied the defect physics and revealed the origin of ambipolar conduction in Cs2SnI6. These results break the conventional common sense about chemical bonding nature in ionic semiconductors and provide a guiding principle to design new pervoskite-based PV materials.