The final days of the Biden administration have witnessed a flurry of activity on AI policy. On Jan. 14, President Biden issued a new executive order (EO) aimed at advancing U.S. leadership in AI by building the next generation of AI infrastructure “in a way that enhances economic competitiveness, national security, AI safety, and clean energy.” The EO followed the Jan. 13 announcement of the “Framework for Artificial Intelligence Diffusion” designed to foster collaboration with U.S. allies while restricting the export of advanced AI chips. Together, the Executive Order (EO) and diffusion framework represent the administration’s final push to position the United States at the forefront of the global AI race.
While ambitious, the EO faces significant challenges in reconciling environmental goals with economic priorities, underscoring the complexities of fostering sustainable AI innovation. As President-elect Donald Trump prepares to take office next week, the ultimate impact of the EO, which is not codified in law, is highly uncertain. Nevertheless, certain elements of the E.O. are likely to persist, particularly its emphasis on geothermal and nuclear power as key energy sources for expanding AI infrastructure in the United States.
Key Provisions in the AI Executive Order
The EO positions the construction of AI data centers as a national security imperative, bolstering U.S. leadership in AI innovation and the clean energy transition. To achieve these goals, the order directs the Department of Defense (DoD), Department of Energy (DoE) and other federal agencies to take steps to advance U.S. national security, promote economic competitiveness, expand clean energy deployment, limit energy cost increases for consumers and businesses and mitigate risks associated with AI infrastructure expansion.
Under the EO, federal land will be leased competitively to private developers to host AI data centers, provided they minimize health and environmental impacts on local communities. Developers building on these sites will be required to meet the electricity needs of their data centers with “clean energy generation resources,” including geothermal, nuclear, solar and wind. By prioritizing these energy sources, the EO seeks to ensure that data centers are aligned with broader climate and decarbonization commitments. The Environmental Protection Agency (EPA) is tasked with providing technical assistance to developers in meeting air quality standards and adopting emissions reduction practices.
The EO further orders the Secretary of Defense, in collaboration with the Secretaries of Agriculture, Commerce, Energy and the Interior, to conduct reviews of the environmental consequences associated with the construction and operation of AI data centers. In coordination with other agencies, the DoE will expand research and development efforts concerning AI data center efficiency, including software, electrical systems, and heating and cooling infrastructure. The DoD and DoE will both require developers on federal sites to pay the total costs of building, operating and maintaining AI data centers, acquire sufficient clean energy generation resources that prevent electricity price increases, as well as purchase a share of U.S.-manufactured semiconductors. The EO also obliges developers to assess the safety and security risks of AI models, in addition to their potential to “advance national security objectives.”
On the international front, the EO directs the Department of State, in consultation with other federal agencies, to develop a strategic plan for engaging with allies and partners on building “trusted” AI infrastructure — data centers that train and deploy safe, secure and transparent AI models. The plan will include measures to mitigate harms to local communities, advance clean energy technologies, and enhance cybersecurity and supply chain protections for AI infrastructure.
Balancing Innovation with Sustainability
The EO comes amid growing concerns about how AI exacerbates climate risks. “AI-powered” products are fueled by multiple energy sources. Data centers consume enormous amounts of energy and water, with projections estimating they will account for 8 percent of U.S. electricity consumption by 2030. A Google search powered by ChatGPT uses roughly ten times the amount of energy as a traditional Google search. The EO’s focus on clean energy technologies, such as geothermal and nuclear, reflects an effort to accelerate advances in AI while addressing its environmental impacts.
There are several provisions within the EO, however, that could undermine this objective. Federal agencies face a deadline of the end of 2025 to process permits and approvals for constructing and operating AI infrastructure on federal sites — a timeline that prioritizes winning the AI race over carefully scrutinizing its impacts. For instance, certain development projects — including those designated as Priority Geothermal Zones or deemed to not “significantly affect” the environment — may be categorically excluded from environmental assessments and impact statements. These carve-outs risk overlooking the broader environmental challenges of AI infrastructure and could impede the sustainability goals the EO intends to achieve.
The EO’s emphasis on geothermal and nuclear energy also raises questions about the viability of its environmental goals. The order includes provisions that would fast-track permits for geothermal projects and facilitate the deployment of additional nuclear powers. While geothermal and nuclear can accelerate the clean energy transition by supporting intermittent energy sources — such as solar and wind — in meeting surging electricity demands, both present certain environmental risks. Developers are increasingly applying hydraulic fracturing or “fracking” — techniques traditionally used in oil and gas extraction — to access geothermal energy, which not only strains water resources, but may also induce earthquakes. Nuclear power faces its own host of challenges, from non-proliferation concerns to the safe storage and disposal of radioactive waste. By expediting permits for geothermal and nuclear energy sources, the EO may overlook the broader repercussions for local communities and ecosystems.
Regional disparities also present obstacles, as varying levels of infrastructure development across the United States could limit the ability to deploy renewable energy technologies equitably. Many areas lack the infrastructure to support large-scale deployment of renewable energy technologies, which could widen geographic inequities. While the EO directs agencies to “prioritize” measures to mitigate electricity cost increases for communities, it offers few concrete mechanisms to achieve this. Without targeted interventions tailored to specific communities, integrating energy-intensive data centers into existing power grids could create negative externalities for local residents.
Ultimately, the environmental regulations outlined in the EO may serve to undermine the competitive AI ecosystem it seeks to establish. Developers bidding for federal sites must meet stringent criteria, including demonstrating high-quality AI training capabilities, maximizing energy and water efficiency, and implementing robust security measures. These requirements inadvertently may favor larger tech companies with the resources to comply, limiting opportunities for smaller players and potentially stifling innovation. Moreover, the added regulatory burdens could deter private-sector investment altogether, slowing efforts to expand AI infrastructure.
What Comes Next
The EO represents a difficult balancing act between advancing AI innovation and prioritizing national security, economic competitiveness, and environmental sustainability that will be difficult to maintain under the incoming administration. Trump has promised to repeal the Biden administration’s October 2023 EO on the “safe, secure, and trustworthy” development and use of AI, arguing that it hinders “innovation.” It is not hard to imagine that this EO will share a similar fate.
Still, Trump repeatedly has emphasized the need to expand energy sources to accelerate AI development, signalling potential common ground. One area of continuity with Biden’s EO governing AI infrastructure may be its emphasis on nuclear power. The Biden administration had set ambitious targets to triple U.S. nuclear capacity, a priority that the second Trump administration, with its strong pro-nuclear stance, may retain and even amplify. In his confirmation hearing on Jan. 15, Trump’s DoE nominee Chris Wright expressed strong support for expanding commercial nuclear power, aligning with trends among big tech companies (including Microsoft and OpenAI) exploring nuclear reactors to power data centers.
Notwithstanding the shared prioritization of nuclear power, Trump’s broader approach to energy policy risks sidelining environmental safeguards and equity considerations when it comes to AI infrastructure. The new Trump administration is expected to streamline regulations and prioritize traditional energy sources, including expanded reliance on fossil fuels, to meet AI’s growing energy demands. This shift risks anchoring U.S. AI infrastructure to a model prioritizing short-term economic gains over environmental stewardship, challenging efforts to ensure that AI development is safe, equitable, and ultimately sustainable.