AI, Energy, and Climate: What’s at Stake? Hint: A lot.
DC’s first-ever Climate Week brought with it many chances to discuss the hottest-button topics in climate innovation and policy. FAS took the opportunity to do just that, by hosting a panel to explore the intersection of artificial intelligence (AI), energy, and climate issues with leading experts. Dr. Oliver Stephenson, FAS’ Associate Director of Artificial Intelligence and Emerging Technology Policy, sat down with Dr. Tanya Das, Dr. Costa Samaras, and Charles Hua to discuss what’s at stake at this critical crossroads moment.
Missed the panel? Don’t fret. Read on to learn the need-to-knows. Here’s how these experts think we can maximize the “good” and minimize the “bad” of AI and data centers, leverage research and development (R&D) to make AI tools more successful and efficient, and how to better align incentives for AI growth with the public good.
First, Some Level Setting
The panelists took their time to make sure the audience understood two key facts regarding this space. First, not all data centers are utilized for AI. The Electric Power Research Institute (EPRI) estimates that AI applications are only used in about 10-20% of data centers. The rest? Data storage, web hosting capabilities, other cloud computing, and more.
Second, load growth due to the energy demand of data centers is happening, but the exact degree still remains unknown. Lawrence Berkeley National Lab (LBNL) models project that data centers in the US will consume anywhere between 6.7% and 12% of US electricity generation by 2028. For a country that consumes roughly 4 trillion kilowatt hours (kWh) of electricity each year, this estimation spans a couple hundred billion kWh/year from the low end to the high. Also, these projections are calculated based on different assumptions that factor in AI energy efficiency improvements, hardware availability, regulatory decisions, modeling advancements, and just how much demand there will be for AI. When each of these conditions are evolving daily, even the most credible projections come with a good amount of uncertainty.
There is also ambiguity in the numbers and in the projections at the local and state levels, as many data center companies shop around to multiple utilities to get the best deal. This can sometimes lead to projects getting counted twice in local projections. Researchers at LBNL have recently said they can confidently make data center energy projections out to 2028. Beyond that, they can’t make reasonable assumptions about data center load growth amid growing load from other sectors working to electrify—like decarbonizing buildings and electric vehicle (EV) adoption.
Maximizing the Good, Minimizing the Bad
As data center clusters continue to proliferate across the United States, their impacts—on energy systems and load growth, water resources, housing markets, and electricity rates—will be most acutely felt at the state and local levels. DC’s nearby neighbor Northern Virginia has become a “data center alley” with more than 200 data centers in Loudoun County alone, and another 117 in the planning stages.
States ultimately hold the power to shape the future of the industry through utility regulation, zoning laws, tax incentives, and grid planning – with specific emphasis on state Public Utility Commissions (PUCs). PUCs have a large influence on where data centers can be connected to the grid and the accompanying rate structure for how each data center pays for its power—whether through tariffs, increasing consumer rates, or other cost agreements. It is imperative that vulnerable ratepayers are not left to shoulder the costs and risks associated with the rapid expansion of data centers, including higher electricity bills, increased grid strain, and environmental degradation.
Panelists emphasized that despite the potential negative impacts of AI and data centers expansion, leaders have a real opportunity to leverage AI to maximize positive outcomes—like improving grid efficiency, accelerating clean energy deployment, and optimizing public services—while minimizing harms like overconsumption of energy and water, or reinforcing environmental injustice. Doing so, however, will require new economic and political incentives that align private investment with public benefit.
Research & Development at the Department of Energy
The U.S. Department of Energy (DOE) is uniquely positioned to help solve the challenges AI and data centers pose, as the agency sits at the critical intersection of AI development, high-performance computing, and energy systems. DOE’s national laboratories have been central to advancing AI capabilities: Oak Ridge National Laboratory (ORNL) was indeed the first to integrate graphics processing units (GPUs) into supercomputers, pioneering a new era of AI training and modeling capacity. DOE also runs two of the world’s most powerful supercomputers – Aurora at Argonne National Lab and Frontier at ORNL – cementing the U.S.’ leadership in high-performance computing.
Beyond computing, DOE plays a key role in modernizing grid infrastructure, advancing clean energy technologies, and setting efficiency standards for energy-intensive operations like data centers. The agency has also launched programs like the Frontiers in Artificial Intelligence for Science, Security and Technology (FASST), overseen by the Office of Critical and Emerging Tech (CET), to coordinate AI-related activities across its programs.
As the intersection of AI and energy deepens—with AI driving data center expansion and offering tools to manage its impact—DOE must remain at the center of this conversation, and it must continue to deliver. The stakes are high: how we manage this convergence will influence not only the pace of technological innovation but also the equity and sustainability of our energy future.
Incentivizing Incentives: Aligning AI Growth with the Public Good
The U.S. is poised to spend a massive amount of carbon to power the next wave of artificial intelligence. From training LLMs to supporting real-time AI applications, the energy intensity of this sector is undeniable—and growing. That means we’re not just investing financially in AI; we’re investing environmentally. To ensure that this investment delivers public value, we must align political and economic incentives with societal outcomes like grid stability, decarbonization, and real benefits for American communities.
One of the clearest opportunities lies in making data centers more responsive to the needs of the electric grid. While these facilities consume enormous amounts of power, they also hold untapped potential to act as flexible loads—adjusting their demand based on grid conditions to support reliability and integrate clean energy. The challenge? There’s currently little economic incentive for them to do so. One panelist noted skepticism that market structures alone will drive this shift without targeted policy support or regulatory nudges.
Instead, many data centers continue to benefit from “sweetheart deals”—generous tax abatements and economic development incentives offered by states and municipalities eager to attract investment. These agreements often lack transparency and rarely require companies to contribute to local energy resilience or emissions goals. For example, in several states, local governments have offered multi-decade property tax exemptions or reduced electricity rates without any accountability for climate impact or grid contributions.
New AI x Energy Policy Ideas Underway
If we’re going to spend gigatons of carbon in pursuit of AI-driven innovation, we must be strategic about where and how we direct incentives. That means:
- Conditioning public subsidies on data center flexibility and efficiency performance.
- Requiring visibility into private energy agreements and emissions footprints.
- Designing market signals—like time-of-use pricing or demand response incentives—that reward facilities for operating in sync with clean energy resources.
We don’t just need more incentives—we need better ones. And we need to ensure they serve public priorities, not just private profit. Through our AI x Energy Policy Sprint, FAS is working with leading experts to develop promising policy solutions for the Trump administration, Congress, and state and local governments. These policy memos will address how to: mitigate the energy and environmental impacts of AI systems and data centers, enhance the reliability and efficiency of energy systems using AI applications, and unlock transformative technological solutions with AI and energy R&D.
Right now, we have a rare opportunity to shape U.S. policy at the critical intersection of AI and energy. Acting decisively today ensures we can harness AI to drive innovation, revolutionize energy solutions, and sustainably integrate transformative technologies into our infrastructure.
As federal uncertainty grows and climate goals face political headwinds, a new coalition of subnational actors is rising to stabilize markets, accelerate permitting, and finance a more inclusive green economy.
This DOE Office has been achieving DOGE’s stated mission of billion dollar savings for decades. Now government leaders may close its doors.
To fight the climate crises, we must do more than connect power plants to the grid: we need new policy frameworks and expanded coalitions to facilitate the rapid transformation of the electricity system.
The stakes are high: how we manage this convergence will influence not only the pace of technological innovation but also the equity and sustainability of our energy future.