Cognitive radio is a promising technique for increasing spectrum efficiency in wireless networks. In this paper, we study power minimization problem to allow as many secondary users access the channel of primary users as possible. In this appealing issue, many considerable literatures have been studied to generally remove a subset of the secondary users such that the remaining users satisfy the constraints with power budget and Signal-to-Interference-plus-Noise Ratio (SINR). However, when the outage probability is large, i.e. there are numerous unsupported secondary users in cognitive radio networks, the removal algorithms take interference time. Instead of removing the worst secondary user, we propose a hybrid access algorithm, leveraging the characterization of the feasibility with the spectral radius, to promote the capacity and reduce the interference time. Compared with present known alternatives, numerical results show that our algorithm achieves a better capacity.