Abstract : Implicit time schemes reduce the numerical resolution of the Navier-Stokes system to multiple resolutions of steady Navier-Stokes equations. We analyze a least-squares method, introduced by Glowinski in 1979, to solve the steady Navier-Stokes equation. Precisely, we show that any minimizing sequences (constructed by gradient type methods) for a least-squares functional converges strongly toward solutions, assuming the initial guess in an explicit ball dependent of the time step and of the viscosity constant. The resulting method is faster and more robust than the Newton method used to solve the weak variational formulation for the Navier-Stokes. Numerical experiments support our analysis.