Joe Biden’s administration wants to forge ahead with its artificial intelligence goals, but those seem somewhat in conflict with their climate change mitigation efforts.
The US has a significant lead in the global AI race, with many of the most powerful AI models and chips necessary for running them deriving from the US, and thus subject to the full force of US law. But training and running AI models often involves accessing data centers that require copious amounts of electricity to run them, and water to keep them cool, presenting a challenge for an administration that set a goal of a net-zero emissions economy by 2050.
This conflict was at the center of a recent White House roundtable where executives from tech leaders such as Anthropic, Google, OpenAI, Microsoft, Nvidia, and Google met with senior administration officials to discuss the energy infrastructure for AI in the United States, particularly the powerful data centers needed to train and run generative AI models. According to projections from the Electric Power Research Institute, data centers could consume up to 9% of US electricity generation annually by 2030, up from 4% in 2023.
The White House meeting focused on AI's energy usage, data center capacity, semiconductor manufacturing, and grid capacity. In response, the administration has announced a new Task Force on AI Data Center Infrastructure that will monitor the country’s data centers, develop new permitting standards, and aim to transition these electricity guzzlers to clean energy sources.
Landon Derentz, senior director of the Atlantic Council's Global Energy Center, told GZERO that the administration is both pursuing a comprehensive strategy to accelerate US global leadership in AI and maintaining its commitment to sustainability. “While net-zero targets … by 2050 remain the central tenet of the White House’s domestic and foreign policy agenda, winning the race for superiority in AI is an imperative today," he said.
Derentz noted that the administration isn’t just focused on allaying the demands on the US electric grid, but also on how AI could improve grid efficiency and energy use in the future.
“AI has tremendous potential to improve grid reliability and resilience as our energy systems become more complex,” he said. “It can also accelerate research and development of new technologies and materials that can enhance the United States’ strategic advantage in the energy sector.”