Stochastic linearization produces a linear system with the same covariance kernel as the original nonlinear system. The method passes from factorization of finite-dimensional covariance kernels through convergence results to the final input/output operator representation of the linear system.