Data centers form the underpinnings of the global technology revolution that is cloud computing. There is enormous pressure for data center growth and expansion, to meet the computational demands of an increasingly digital world. With energy costs overtaking server costs in data centers, energy is fast becoming a significant bottleneck to data center scale-out. Further, the global data center energy footprint is growing to be a significant burden on the world's energy resources. Yet energy is a signally ill-managed resource in most data centers; average data center energy efficiency is less than 50%. With increasing industry awareness of the magnitude and urgency of this problem, many solutions are cropping up to combat each of the several sources of data center energy inefficiency. The objective of this dissertation is three-fold: First, we examine the causes of data center energy inefficiency from first principles, and identify the challenges involved in addressing them. We find two categories of energy inefficiency: Idle resource energy consumption, and support infrastructure energy consumption. Second, we present solutions to address each form of inefficiency. We describe two ways to combat idle resource energy consumption, and also present a systemic solution to tackle both forms of energy inefficiency. Finally, throughout this dissertation, we examine the related work and literature, and attempt to map them into the solution space to identify how the solutions relate with each other, and what gaps remain to be addressed. The cloud has the potential to enable everything from ubiquitous computing and universal access to knowledge, to smart power grids, greater social connectivity, and near-infinite extensibility of compute/storage power. The cloud turns computation into a utility, and by doing so, has the potential to make it accessible to a much larger part of the world. This dissertation explores ways to enable sustainable scaling of the data centers that power the cloud and enable this vision.