The rapid rise of AI data centers has surfaced a complex web of challenges and opportunities. Addressing these requires strong partnerships and continued discourse across industry, government, and academia. Based on the ACES Symposium discussions, we have identified the following critical trends shaping the future of energy and water infrastructure.
Grid Evolution and Load Growth
-
Accelerated Load Growth
After decades of stagnant load, the electricity grid started to see demand increase in the last couple of years. Due to electrification and the growth of AI data centers, the demand trend may look more like a hockey stick in the next few years. Generation, transmission, and distribution assets will need to be expanded to accommodate this load growth, while ensuring affordability and reliability for all grid users.
-
Aging Infrastructure
Many grid assets are nearing the end of their lifespans. While this offers opportunities for new technologies to bring new efficiencies to the grid, the cost of upgrades presents challenges to energy affordability, even without load growth. At the same time, load growth spurs the need for infrastructure upgrades and investments in ways that stagnant load may not.
-
Grid Resilience
The frequency and impact of severe weather events have increased significantly in the last two decades. These events, as well as man-made risks from cybersecurity threats, pose an emerging resiliency challenge for the electricity grid and energy systems more broadly.
-
Energy Transition
Traditional thermal and synchronous generators are being replaced by inverter-based resources (solar, wind), which exhibit greater variability and uncertainty. This complicates generation dispatchability.
-
Electrification
Even in the absence of load growth from AI data centers, electricity demand is projected to grow in the years ahead due to electrification. The timescale for load growth from electrification is slower than the significant uptick we are currently experiencing from data centers coming online. However, both growth trends are underlying the increasing electricity demand over the next two decades.
Market Volatility and Supply Chain
-
Capacity Price Spikes
PJM, the electric grid operator for Pennsylvania and its region, has experienced large capacity price spikes recently. PJM’s capacity July 2024 auction (for delivery in 2025) cleared at roughly $270/MWh/day, up from $30/MWh/day the year before. The July 2025 auction reached the market cap of nearly $330/MWh/day. These spikes are attributed to data center demand growth and accelerated retirements of generation assets, which are not being offset by new entrants.
-
Energy Affordability
Currently, 1 in 6 electricity customers has difficulty paying their bills on time. Approximately 30% of US households experience some form of energy insecurity due to rising natural gas and electricity prices. Increased demand could exacerbate these issues.
-
Supply Chain Constraints
While expanding operationally flexible generation capacity seems like an obvious way to address load growth, there are concerns about how quickly new natural gas power plants can be brought online due to supply chain constraints. Large manufacturers of combined cycle natural gas turbines, both in the US and abroad, have their queues booked almost through 2030.
-
Accelerated Retirements
Recently, retirements of generation assets have accelerated. Some retirements are age-related and some are market-driven. But the bulk are based on policies. The pace of retirement is not keeping up with labor force growth, and new entrants are not offsetting accelerated retirement.
Data Center Technical Demands
-
Growing Complexity and Power Demands
Power densities of advanced AI racks have grown sharply and will continue to do so. Drawing on average 25kW-40kW two years ago, these racks now have power densities exceeding 200kW, with the aim of producing 1MW racks in the near future. The data center load growth, therefore, comes from both new data centers coming online and newer racks being integrated. Thus, for data center developers, the primary challenge is in identifying solutions to meet their expected and growing energy needs.
-
Power Quality & Stability
In addition to their high density, AI data center power demands are highly variable and nonlinear. This poses a challenge for the power quality delivered to other local grid customers. Additionally, data centers use uninterruptible power supply systems to disconnect from the grid if they sense minor disturbances. These disconnections, particularly of large-scale data centers, can pose serious challenges for the grid.
-
The Water-Energy Trade-off
Data centers are moving toward reduced water use due to concerns about water scarcity, even though more energy-efficient cooling systems are highly water-intensive. The industry is focusing on developing technologies that use closed-loop systems, alternative liquid cooling, or waterless innovations such as immersion cooling.
Policy, Siting, and Global Competition
-
Permitting and Siting Challenges
Data center developers cite challenges with permitting and zoning as obstacles to the timely and scaled deployment of their infrastructure. There are also challenges for queued generation assets needed to support these infrastructures.
-
Water Load Stewardship
While the electricity grid has not seen load growth over the last several decades, regional hydropower systems have experienced similar demand ramp-ups in recent years, particularly related to shale gas drilling. Over the last 10 years, this ramp up led to improved efficiencies and environmental stewardship by drilling companies. A similar trend could be expected for the current demand growth driven by data centers. In this respect, caution is warranted to avoid over-predicting demand growth and overbuilding.
-
The Geopolitical AI Race
The growth and dominance of AI technology have been positioned as the geopolitical race of our times. An all-hands-on-deck approach will be needed to compete in this race, grow national AI capacity, and build supporting infrastructures. The AI industry transforms quickly, and growth is needed at a fast pace. Innovations to address the energy and water challenges of AI data centers must align with this deployment and advancement timeline, i.e., speed to market.
Emerging Solutions
-
AI for Energy
The recent emergence of AI can improve the efficiency of energy systems. Newer technology solutions for grid operations, maintenance, forecasting, and decision-making incorporate AI- and learning-based approaches to improve system efficiency and reliability.