Several parametric scans have been performed to study momentum transport on JET, AUG, DIII-D, NSTX and C-Mod. NBI modulation technique and RMP coil perturbation have been applied to separating the diffusive and convective momentum transport terms. The magnitude of the inward momentum pinch depends strongly on the inverse density gradient length, with multi machine experimental scaling for the pinch number being ­Rvpinch/cf = 1.1R/Ln + 1.0. There is no dependence of the pinch number on collisionality whereas the pinch number decreases weakly with increasing q-profile. The Prandtl number was not found to depend either on R/Ln, collisionality or on q. The gyro-kinetic simulations show qualitatively similar dependence of the pinch number on R/Ln, but the dependence is weaker in the simulations. Gyro-kinetic simulations do not find any clear parametric dependence in the Prandtl number, in agreement with experiments. Momentum transport coefficients from each device give quite similar dependences and trends, and therefore extrapolation to ITER seems fairly robust. The extrapolation of these results to ITER illustrates that at large enough R/Ln>2 the pinch number becomes large enough (> 3­4) to make the rotation profile peaked provided that the edge rotation is nonzero. And this rotation peaking can be achieved with small or even with no core torque source.