Machine to Machine (M2M) communications are a rapidly expanding subset of the Internet of Things (IoT). Example scenarios include small wireless sensor networks, which may only transmit small amounts of data a few times a day; being constrained by a finite and small battery life. This thesis examines the use of 4G (LTE) mobile communications for M2M, and how power consumption may be optimised. Whilst 4G may not be the most efficient way to communicate, the network ubiquity makes ease of deployment a highly attractive option. With this low efficiency, every possible technique should be considered to recover system efficiency, and increase battery life. The approach to this problem involves analysis of LTE signals, measurement and modelling of power consumption in real devices, and evaluation of candidate hardware techniques to improve efficiency. Hardware in the loop methods are used extensively, based on simulated and live networks to verify benefit. For LTE signals, correlation is observed between the signal envelope, and degrees of freedom within the system, such as bandwidth and modulation scheme. This allows the operating point of the power amplifier to be altered to reflect the current signal characteristics without the need to continually monitor the envelope, enabling low cost efficiency enhancement. Energy consumption of LTE modems was measured and modelled for a variety of use cases to produce information aimed at selecting appropriate transmission profiles. In order to remain energy competitive with other long range systems, LTE modems should be used in a high latency mode, accepting an increase in cold start time. Load modulation and dynamic power supply technologies are investigated for low speed modulation based only on MAC layer information in order to increase system efficiency. Improvements of up to ten percentage points in Power Added Efficiency across a reasonable range of output power values have been demonstrated.