The aim of this thesis is to determine the natural boundary of meromorphy (when it exists) of an Euler product of n variables associated to a polynomial h \in \mathbf{Z } [X_1....,X_,n] satisfying an hypothesis of analytic regularity. Precisely it consists in finding the boundary of a maximal domain on which a meromorphic extension exists. We present in this thesis some methods which permit to extend in the multivariable case, under an hypothesis of analytic regularity which is mostly satisfied, the well-know result of Estermann concerning the maximal domain of meromorphy of an one variable Euler product \prod_{p}h(p^{-s}) associated to a polynomial h with integral coefficients (such that Sh(0)=1S). We also precise the sense which we can give to the concept of "natural boundary" with regard to the real or complex dimension of a possible continuation beyond this boundary. As an application, we determine the natural boundary of a class of Euler products associated to a projective toric variety. A second application consists in the determination of the natural boundary of a class of Euler products of the form \prod_{p}h(p^{-s_l },...,p^{-s_n},p^{-c }) where c is an integer (positive or negative). In particular we solve a problem of N. Kurokawa and H. Ochiai concerning the natural boundary of meromorphy of the multivariable lgusa zeta function Z^{\textrm{ring} }(s_1,\dots,s_n; \mathbf{Z}[T,T^{-1}])
