This article discusses the costs and challenges of cloud service data centers, emphasizing the need for optimization to improve efficiency and reduce expenses. Cloud data centers represent a significant investment in capital and ongoing costs. The cost breakdown reveals the importance of optimizing work per dollar invested. However, resources inside data centers often operate at low utilization due to resource stranding and fragmentation. To address this, the authors propose increasing network agility and providing appropriate incentives to shape resource consumption. Cloud service providers are building geo-distributed networks of data centers, which can lower latency and increase reliability. However, without proper design and management, these networks can raise service costs. The authors suggest joint optimization of network and data center resources and new systems for geo-distributing state. The article also highlights the importance of reducing power consumption and improving network efficiency. It discusses the challenges of current data center networking, including static network assignment, resource fragmentation, poor server-to-server connectivity, and proprietary hardware that scales up rather than out. The authors propose design objectives such as location-independent addressing, uniform bandwidth and latency, and security and performance isolation to achieve agility. Incentives for desirable behavior are also discussed, focusing on shaping resource consumption through economic incentives. The article emphasizes the need for geodiversity to reduce latency and improve system availability. It outlines the challenges of optimal placement and sizing of data centers, considering geographic diversity, size, and network cost. Finally, the article concludes that improving data center efficiency requires increasing internal network agility, designing algorithms and market mechanisms for resource consumption shaping, and geodistributing data centers to enhance performance and reliability. The authors also emphasize the need for joint optimization of data center and network resources and new systems for managing geo-distribution of state.This article discusses the costs and challenges of cloud service data centers, emphasizing the need for optimization to improve efficiency and reduce expenses. Cloud data centers represent a significant investment in capital and ongoing costs. The cost breakdown reveals the importance of optimizing work per dollar invested. However, resources inside data centers often operate at low utilization due to resource stranding and fragmentation. To address this, the authors propose increasing network agility and providing appropriate incentives to shape resource consumption. Cloud service providers are building geo-distributed networks of data centers, which can lower latency and increase reliability. However, without proper design and management, these networks can raise service costs. The authors suggest joint optimization of network and data center resources and new systems for geo-distributing state. The article also highlights the importance of reducing power consumption and improving network efficiency. It discusses the challenges of current data center networking, including static network assignment, resource fragmentation, poor server-to-server connectivity, and proprietary hardware that scales up rather than out. The authors propose design objectives such as location-independent addressing, uniform bandwidth and latency, and security and performance isolation to achieve agility. Incentives for desirable behavior are also discussed, focusing on shaping resource consumption through economic incentives. The article emphasizes the need for geodiversity to reduce latency and improve system availability. It outlines the challenges of optimal placement and sizing of data centers, considering geographic diversity, size, and network cost. Finally, the article concludes that improving data center efficiency requires increasing internal network agility, designing algorithms and market mechanisms for resource consumption shaping, and geodistributing data centers to enhance performance and reliability. The authors also emphasize the need for joint optimization of data center and network resources and new systems for managing geo-distribution of state.