What's Happening?
A significant portion of the planned data centers in the U.S., particularly those driven by the boom in generative artificial intelligence, may not be constructed. According to estimates by Bernstein Research, only about 35% of the announced data center capacity
in the U.S. is likely to be built. This skepticism stems from various factors, including a lack of guaranteed electricity supply, necessary permits, or confirmed customers for many proposed projects. The issue is particularly pronounced in Texas, where applications for large consumers to connect to the electricity grid have reached 474 gigawatts, more than five times the state’s record peak electricity consumption. Approximately 90% of this declared capacity in Texas is associated with data centers. Analysts have coined the term 'bragawatts' to describe ambitious claims by developers who may submit multiple applications without the intention or capability to follow through on all of them, leading to inflated forecasts of future demand.
Why It's Important?
This discrepancy between announced and actual data center construction has significant implications for U.S. infrastructure planning and energy policy. Overestimated demand forecasts can lead to inefficient investment in electricity networks, potentially resulting in unused equipment and increased costs that could be passed on to consumers. Conversely, underestimating actual growth could leave existing infrastructure unable to support new data centers, hindering technological advancement. The International Energy Agency (IEA) forecasts that global data center electricity consumption will nearly double by 2030, reaching 950 TWh, with AI-specific data centers tripling their energy use. In the U.S., data centers are projected to account for nearly half of electricity demand growth by 2030. The challenges in securing power, permits, and customers highlight a critical bottleneck in the expansion of AI infrastructure, impacting the pace of technological development and the reliability of the national grid.
What's Next?
Regulators, such as those in Texas, have begun requiring more detailed information from developers regarding electricity supply, water sources, cooling, and funding to distinguish genuine investment projects from speculative proposals. This increased scrutiny aims to ensure that infrastructure planning is based on realistic projections. The actual pace of data center development will be influenced by factors such as the cost of loans, the availability of electricity and microchips, and the ease of obtaining planning permission. While major technology companies continue to expand computing capacity due to strong demand for new services, the industry will likely focus more on completed construction, actual connected electrical capacity, installed server equipment, and data center utilization rates as key indicators of growth, rather than initial announcements. This shift will help in more accurately forecasting energy needs and preventing misallocations of resources.
Beyond the Headlines
The phenomenon of 'phantom' data centers reveals a deeper challenge in managing rapid technological growth and its environmental footprint. The energy intensity of AI and data centers raises ethical questions about sustainable development and resource allocation. The reliance on diesel generators for backup power, for instance, contributes to environmental concerns, even as the industry strives for efficiency. The long lead times for grid connections and the increasing trend of data centers building their own on-site power generation, often gas-fired, underscore the systemic pressures on energy infrastructure. This situation highlights the need for integrated planning that considers not only technological advancements but also the environmental and economic sustainability of supporting infrastructure, potentially accelerating the adoption of renewable energy solutions and more efficient cooling technologies to mitigate the impact of this energy-intensive sector.
