This study presents the influence of the composition of the carrier gas on the growth of GaN by HVPE. Since no hydrogen is introduced in the vapour phase, the deposition is expected to be controlled by Cl desorption in the form of GaCl3, as has been proposed for GaAs. However, our published model predicts much lower growth rates than those observed. We can account for both the observed parasitic deposition and GaN growth rate if we assume that GaCl3 is not at its equilibrium pressure in the deposition zone and where nucleation takes place on the walls as well as on the substrate. This yields a high rate of parasitic nucleation even though the nominal supersaturation is vanishing small. Very little growth takes place on the substrate where the equilibrium pressure of GaCl3 is reached. We describe similar experiments performed with a H-2/N-2 mixture as the carrier gas. In this case, we expect GaN deposition to be controlled by desorption of Cl as HCl, which is known as the H-2 mechanism. It is speculated that the results show the existence of a new growth mechanism.