Data Centers, AI, and Decarbonization: Congressional Interest in Powering Clean Innovation

A common assumption among artificial intelligence (AI) optimists is that exponential growth of AI technology is limited only by the available computing resources. That is not the case. Rather, a significant constraint is the ability to meet the substantial electricity needs of such resources, while satisfying the need to maintain reliability and interest in prioritizing clean electricity resources for this energy-intensive technology. This Blog provides an overview of factors policymakers are weighing as they consider the possibility of regulating the power consumption of AI models and the data centers that house them.

Datacenters and Decarbonization Fuel Energy Demand Growth, Policy Interest

In general, demand for electricity nationwide is surging. Leading consultancies project U.S. power demand will grow by 13% to 15% annually until 2030. Data centers are a significant part of this growing demand, with data center investment forecast to exceed $150 billion through 2028. Serving AI functions drives significant elements of that growth. AI companies often do not disclose the energy intensity of their models, but they are expected to consume increasingly outsized shares of data center energy. Goldman Sachs estimates AI will represent 19% of data center power demand by 2028, increasing overall data center power consumption by approximately 200 terawatt-hours per year.

Congressional Hearings

In response to these trends, federal policymakers are considering the implications of on-site energy demand from datacenters and AI development. Recent hearings in both chambers of Congress preview potential legislative action in the 118th Congress and beyond. The hearing highlights below help illustrate the key themes lawmakers have confronted in this area.

Permitting and energy demand modeling: On September 12, 2024, the Senate Energy and Natural Resources Committee held a hearing on the role of the Department of Energy in influencing advanced computing research. During the hearing, Chair Joe Manchin (I-WV) underscored the importance of expanding the domestic energy mix in order to meet the increasing demand for energy spurred by “data centers and a manufacturing resurgence that has resulted from the Bipartisan Infrastructure Law, CHIPS and Science Act, and Inflation Reduction Act.” Chair Manchin argued that failing to streamline the permitting process for new energy infrastructure will cause the U.S. to lose its global influence to China. Sen. John Hickenlooper (D-CO) questioned witnesses about the impact of AI specifically on energy demand, agreeing when witnesses suggested exact models of energy demand are difficult to create without proprietary information from the electricity customers. Other members, including Sen. Angus King (I-ME), suggested AI will be useful for analyzing grid-level data and allocating resources more effectively.

On-site energy storage solutions: The Energy and Natural Resources hearing followed a similar hearing on May 21, 2024, in which the Senate Energy and Natural Resources Committee focused on addressing growing power demand in the U.S. During the hearing, committee members of both parties acknowledged the importance of promoting a competitive AI industry through investments in data centers and computing resources, while at the same time ensuring grid reliability. Committee Democrats underscored the importance of preserving the administration’s clean energy objectives in the face of increasing energy demand. Sen. Mazie Hirono (D-HI) contemplated requiring data centers to install energy storage capacity on site to limit energy consumption during peak demand periods, praising efforts by tech and private equity companies to power their data centers via green energy.

AI power demand in context: The House Energy and Commerce Energy, Climate, and Grid Security Subcommittee held a hearing on June 4, 2024, focusing on the need to bolster the U.S. energy mix to support AI innovation. During the hearing, Republicans and Democrats discussed measures to overcome the permitting delays and supply chain issues preventing the outlaying of new energy transmission and distribution infrastructure to maintain U.S. dominance in the AI space. Committee Republicans highlighted the importance of diversifying the U.S. energy mix — including through investments in nuclear, natural gas, and coal power — to power the AI industry. Committee Democrats disagreed, arguing renewable energy and efficient energy supply management will keep prices low while meeting demand.

Members of Congress did not attribute demand growth entirely to AI and the growth of data centers, however. Noting the other pressures facing the grid, Ranking Member Frank Pallone (D-NJ) avoided blaming AI exclusively for the growth in energy demand, arguing “[t]he long-term trend of vehicle and building electrification over the next 25 years will have an even more outside impact on demand growth.” In keeping with this sentiment, Republicans and Democrats focused on broader permitting reform efforts, rather than AI- and data center-specific proposals. Rep. Scott Peters (D-CA) plugged his BIG WIRES Act (H.R. 5551/S. 2827), which would require FERC transmission planning regions develop the capacity to transfer 30% of their peak energy loads to nearby regions in the event of energy supply shortages. Rep. Kim Schrier (D-WA) underscored the importance of investing in domestic transformer manufacturing capabilities to address existing supply chain issues. Committee Republicans criticized the Environmental Protection Agency’s recent power plant emission standards rule, arguing utilities must be allowed to use a combination of renewable and fossil fuels to meet growing energy demand.

AI power opportunities and possible liabilities: In October, 2023, the House Energy and Commerce Energy, Climate, and Grid Security Subcommittee held a hearing on the power demands of data centers and the importance of preserving U.S. leadership over China in AI innovation. The hearing also echoed partisan divisions regarding green energy’s capacity to power the growing AI industry. Committee members highlighted the promise of AI for increasing the efficiency of power plant operation and energy distribution. Subcommittee Chair Jeff Duncan (R-SC) and others expressed concern that AI’s reliance on semiconductors could create reliability challenges in the future, adding that the “black boxes” surrounding AI decision-making necessitate the continued inclusion of human operators in AI-enabled critical infrastructure sectors. Committee members also underscored the importance of ensuring that AI models do not create new cybersecurity vulnerabilities in critical infrastructure networks. While members did not introduce specific legislative proposals in this hearing, these areas of interest may be ripe for future legislative action.

Conclusion

As Congress turns its attention towards the energy implications of technologies like artificial intelligence and clean manufacturing, policymakers will quickly outline the narratives to characterize the roles of actors in this space. Early entrants have an opportunity to educate policymakers on their perspectives before these narratives become entrenched, defining the terms of their seat at the negotiating table. Companies involved in every step of the innovation supply chain should consider the potential impact of legislation to address imbalances identified by policymakers, and the importance of a first-mover advantage for influencing that legislation. Please contact Arnold & Porter’s Legislative and Public Policy group for details on how to engage.

© Arnold & Porter Kaye Scholer LLP 2024 All Rights Reserved. This blog post is intended to be a general summary of the law and does not constitute legal advice. You should consult with counsel to determine applicable legal requirements in a specific fact situation.