We report the application of the density-matrix renormalization group method to a spatially anisotropic two-dimensional Hubbard model at half-filling. We find a deconfinement transition induced by the transverse hopping parameter ty from an insulator to a metal. Therefore, if ty is fixed in the metallic phase, increasing the interaction U leads to a metal-to-insulator transition at a finite critical U . This is in contrast to the weak-coupling Hartree-Fock theory which predicts a nesting induced antiferromagnetic insulator for any U > 0.The metal-insulator transition (MIT), also called the Mott transition [1], is certainly one of the most difficult challenge facing condensed matter theorists. Hubbard [2], in a pioneering work, introduced a simple one-band Hamiltonian which has only two parameters, t for the kinetic energy of the electrons and U for the local electronelectron interactions. This model is at half-filling the model of reference for the MIT. In D = 1, Lieb and Wu [3] obtained an exact solution by using the Bethe ansatz. The ground state is an insulator for any U/W > 0, where W is the band width. Thus, the MIT occurs at the critical value (U/W ) c = 0. In infinite dimensions, the dynamical mean-field theory (DMFT) [4,5] predicts a critical point at (U/W ) c ≈ 1.The discovery of layered materials, where the motion of electrons driving the low energy physics is mostly confined in the layers, has raised great interest into the twodimensional (2D) Hubbard model. The physics at large U/W > ∼ 1 is now understood, the charge excitations are gapped, the spin excitation are described by the Heisenberg Hamiltonian which has long-range order (LRO) at T = 0. But for U/W < ∼ 1, the physics is still unclear. Our current knowledge about the weak-coupling regime is mostly drawn from the Hartree-Fock approximation and from quantum Monte Carlo (QMC) simulations [6,7]. The QMC results agree qualitatively with the HartreeFock prediction that the ground state is a Slater insulator for any U/W > 0. However, in most recent studies such as in Ref. [7], even though considerable progress has been achieved in reaching larger systems, in the weak U regime where the eventual gap is small, reliable extrapolations of the QMC data remain difficult to achieve. It would thus be preferable to apply finite size scaling for data analysis instead of relying on extrapolations.More recently, extensions of the DMFT which include non-local fluctuations, the dynamical cluster approximation (DCA) [8] or the cellular DMFT [9,10], have been applied to the 2D Hubbard model. The focus in these studies have mostly been to discuss the nature of the MIT within the paramagnetic solution of the DMFT equations. A systematic comparison of the possible ordered or disordered ground states as function of the cluster sizes is still lacking. Therefore, the issue as to whether or not quantum fluctuations destroy the Hartree-Fock solution in the half-filled 2D Hubbard model in the small U regime remains open.In this letter, we show that insight into this problem...