Increasing the battery lifetime of power-hungry mobile devices has become a major research target for mobile operators. Based on extensive measurement campaigns with the most recent Long Term Evolution (LTE) devices, we introduce a new Markovian power consumption model, which takes into account the chosen system parameters (such as the number of physical resource blocks) as well as the context of a user in terms of radio channel conditions and service characteristics (non-real-time vs. real time). One key advancement of this generic model is its stochastic nature, which allows for determining the average power consumption of a device based on usage profiles including location information and service statistics. We have conducted comprehensive system simulations using realistic channel characteristics derived from ray-tracing analyses and validated the new model. Finally, we show that the proposed context-aware power consumption model enables quantitative analyses of the trade-off between network resource allocation and enhanced battery lifetime. © 2013 The Authors. Transactions on Emerging Telecommunications Technologies published by John Wiley & Sons, Ltd.