The use of multi-interfaced devices has been steadily increasing due to the popularity of multihomed streaming services in heterogeneous wireless access networks. However, running multiple interfaces simultaneously in a mobile terminal (MT) may cause serious battery drain even when interfaces employ sleep modes. In addition, sleep modes may result in significant degradation of the quality of service (QoS) in terms of packet delay or jitter. This paper examines multihomed MTs and the use of data stripping load distribution across the multiple wireless interfaces of the MTs to minimize their power consumption. This paper then develops a theoretical framework for a power-efficient multipath load distribution that encompasses a dynamic load distribution to each interface employed with sleep mode. For this purpose, the paper first presents analytical models for power consumption and the delay of each interface by considering a medium access control (MAC) operation. Using these models, two simple greedy distribution algorithms are proposed to optimize the load distribution. Extensive simulations in an ns-2 simulator under various practical configurations demonstrate that the proposed algorithms significantly reduce power consumption while satisfying QoS constraints.