In the existing virtual network embedding algorithms, the minimization of energy consumption for mapping virtual nodes is ignored. By means of the optimization theory and the resource consolidation, we construct an optimization model for energy efficient virtual node embedding. The minimization of energy consumption is solved as the minimal product of energy cost per CPU and allocated CPU from virtual nodes to substrate nodes. Based on the minimal element, an algorithm of energy-efficient virtual network embedding is proposed. Simulation results show the energy gains of the proposed algorithm over the existing heuristic algorithms.