We propose a new microscopic criterion to determine surface wetting: water wets the surface whenever its overlayer has a larger adsorption energy than three-dimensional clusters on the bare or water-covered substrate. This conceptually intuitive criterion is validated by detailed first-principles calculations of the energetics of layers and clusters of water on different metal surfaces. This criterion resolves naturally the current discrepancy between theory and experiment on the wetting behavior of undissociated water on Ru(0001), as well as the hydrophobic nature of the Au(111) surface. It also explains the Stranski-Krastanov ice growth on Pt(111) observed experimentally.