First-principles total-energy pseudopotential and all-electron calculations predict (001) (GaAs)i-(AlAs)i and (CdTe)i(HgTe)i superlattices to be intrinsically unstable towards disproportionation into compounds. This instability is traced to unfavorable charge redistribution in the system. PACS numbers: 68.65.+g, 61.55.Hg, 68.55.Rt Many disordered 1 or artificially ordered 2 semiconductor systems are manifestly metastable in temperature and composition ranges in which they are usually characterized and utilized. Such are disordered GaAs x Sbix alloys 1 (grown in the range of thermodynamic immiscibility of GaAs and GaSb), and ordered (A m B w )\x -(C\ w ) x alloys 2 (judged by their equilibrium phase diagrams to spinodally decompose). Metastable systems come to exist through kinetic rather than thermodynamic control, e.g., by nonequilibrium growth techniques. 1,2 They owe their thermal stability 1 ' 2 to large reorientation activation barriers, 2 small thermodynamic driving forces, 1,2 and exceedingly low diffusion coefficients at laboratory temperatures. 1,2 Despite extensive study, it is as yet unclear whether artificial semiconductor superlattices (AC) m (BC) n are thermodynamically (intrinsically) stable or metastable. Current understanding can be characterized as follows. Disordered (D) isovalent alloys A X B\-X C are known to have positive enthalpies of mixing 3 AH D (x), so that at a sufficiently low temperature T c the negative entropy term -T C AS D is overwhelmed by the positive AH D , leading eventually 3,4 to disproportionation. Most contemporary theoretical models 3,4 analyze this instability of A X B\-X C alloys via models that do not distinguish them from ordered compounds A m B n C m + n of the same composition. Such are, e.g., elastic models 4 which attribute AH > 0 to the destabilizing role of microscopic strain associated with a mismatch Aa between lattice constants of AC and BC. Since thin superlattices (AC) m (BC) n are most naturally regarded as ordered compounds 5 -e.g., an m=n = \ superlattice in the (001) orientation is crystallographically identical to an ABC2 compound with the simple tegragonal p4m2 space group 5 ' 6 (having a CuAu-I-like A-B sublattice)-these models would judge both alloys and superlattices (having nearly the same Aa) intrinsically unstable at low temperatures. However, Srivastava, Martins, and Zunger 6 demonstrated that AH D > 0 does not require ordered (O) phases to be unstable too because (i) a chemical energy term, neglected by other models, 4 may render AH° negative, and (ii) coherently ordered arrangements of bonds can reduce strain imposed by bond-length mismatch better than do disordered arrangements.Perhaps the best-studied superlattice-(GaAs) m -(AlAs)"-exhibits, 7 " 9 however, a delicate energy balance: It has a nearly vanishing Aa =^A iAs"~^GaAs =0.0009 A at growth temperatures -800 K and consequently a nearly vanishing 3 AH D (hence, ordering offers but a small reduction in strain), yet Al differs (slightly) from Ga in electronegativity (hence, charge tra...