Among the ab initio methods, second-order perturbation theory (MP2) predicts highly accurate structures and relative energies for water clusters. Researchers will carry out molecular dynamics simulations of water at the MP2 level. However, full MP2 calculations of even modest-sized water clusters are far too time-consuming for dynamical simulations, even on the next-generation Blue Gene. Therefore, a key element of the current approach will be the use of MP2 in conjunction with the Fragment Molecular Orbital (FMO) method. Compared with today’s calculations, researchers will determine the bulk properties at higher levels of accuracy using larger basis sets, larger embedded clusters, and longer dynamics simulations, as permitted by the greater computational capacity available with the Blue Gene. They will target the following bulk properties of water: structure, density, refractive index, diffusion constant, free energy, heat capacity, dielectric constant, vaporization enthalpy, isothermal compressibility, and thermal expansion coefficients. The final eight properties are more difficult to obtain than the first two. While Blue Gene/P gives good estimates, the greater capacity of the next-generation Blue Gene will be critical to establishing convergence of these properties with respect to theory, basis set, cluster size, and simulation length. There have been conflicting reports in the literature about the relevance of “chain” or”ring” networks in water. The high-accuracy simulations on the next-generation Blue Gene will help settle this argument about the structure of liquid water.