The critical thickness for two-dimensional layer growth of Si1-xGex on Si and InxGa1-xAs on GaAs is about 1-3 monolayers (ML) for x = 1, beyond which islanding begins. Under certain growth conditions, this thickness t would increase as some power of 1/x. The reason for this is not clear; Snyder et al, argue that, under equilibrium conditions, this critical Stranski-Krastanov (SK) thickness t(c) is independent of x and should remain at 1-3 ML, but that, under nonequilibrium growth conditions, t similar to x(-4). However, Osten et al. showed, experimentally, that even under equilibrium conditions there is an increase of SK thickness with 1/x. We carry out calculations of energetics of large three-dimensional (3D) islands on substrates with varying thicknesses t of the epilayer and different coverages theta. We show that at low theta or when islands are small (or both) then the SK thickness is small similar to 1-3 ML, in agreement with the results of Snyder et al. At increasing coverages, when interisland separation l decreases to the point where l-s (island size), we observe Delta E to decrease for the lower thicknesses t=3,4,5,...8...; until thicknesses t>3 become more favorable. There is considerable tension going deep into the substrate directly below islands. The larger an island becomes, the more favorable it is for a:thicker layer beneath it to be of the same material as the island. It is known that the critical size s(c) at which 3D islands first become favorable varies as x(-2). We argue from this that, at equilibrium, the average 3D island size increases with x(-p), p some exponent, and at high enough coverages, when interisland separation is small, SK thickness t(c) increases. The experimental results of Osten et al. are consistent with exponents 2 less than or equal to p less than or equal to 4.