When a technology makes its way to Congress, it’s usually a sign of burgeoning national intrigue — as well as a leading indicator of future regulation.
In the wake of ChatGPT’s debut, artificial intelligence has gotten particular attention on the Hill. And on Thursday, applications of the technology for “powering America’s energy future” got their moment in the spotlight.
A House Energy and Commerce Subcommittee heard testimony that showcased lawmaker concerns including data privacy and cybersecurity, the energy intensity of data centers, and the extent to which humans will remain in the loop of an AI-enabled grid.
“We really don’t know a lot about this subject,” Energy, Climate, and Grid Security Subcommittee Chair Jeff Duncan (R-SC) told panelists.
Duncan and others questioned witnesses from both the energy and tech sectors about how to manage the immense power needs of AI-enabled data centers in light of an aging and already strained power grid.
Sreedhar Sistu, vice president of artificial intelligence at Schneider Electric, told the subcommittee that AI demands around 4.3 gigawatts of power today, but Schneider projects that to grow to between 13.5 and 20 GW by 2028.
Major infrastructure investments geared at AI-enablement will be essential to the energy transition, Sistu said in his written testimony: “Investments must be made to ensure data centers can retrofit to meet AI power needs, adopt both air- and liquid-cooling to support AI clusters; and leverage software tools and AI to optimize data center operations.”
Unlike residential consumers, data centers tend to have a constant power demand, explained Paul Dabbar, who was formerly the Department of Energy’s Under Secretary for Science. That likely means the industry will have to turn to energy sources that don’t rely on weather — like nuclear energy, hydro energy, and combined-cycle natural gas — to power them, Dabbar said. Those energy-intensive centers are yet another use case for longer-duration energy storage.
“Hopefully, some of the new battery technologies like Form Energy and others can drive the cost down, because lithium-ion is not good enough to do that cost-effectively,” Dabbar added.
In addition to energy intensity, lawmakers sought insight into national security challenges posed by both hardware and software. All four panelists voiced cybersecurity and privacy concerns, and recommendations ranged from banning “all chips, software, electric vehicle components, and power grid controls from China today” (Dabbar), to developing industry-wide standards in collaboration between public and private sector stakeholders (Sistu.)
Lawmaker questions also homed in on just how big of a role artificial intelligence could play in grid management — as congresswoman Kim Schrier put it, deciding “where AI stops and human decisions begin.”
