As AI-mania sweeps the world, the technology has already been used to do everything from decoding ancient literary epics and writing soulless breakup texts to generating eerily realistic images of people—though often with an extra finger or two. 

Now, some companies and U.S. agencies are betting that AI can also aid in the United States’ scramble for the critical minerals that are essential to powering the green technologies at the center of the energy transition. 

The world’s electric vehicle batteries, wind turbines, and advanced weapons systems have all been built with vast quantities of raw materials: cobalt, copper, lithium, nickel, and powerful rare-earth elements. Yet the United States has for decades been out of the critical minerals game as a result of a raft of environmental, economic, health, and political concerns—pushing Washington to become increasingly reliant on supply chains dominated by China. 

The global energy transition and energy security concerns have reignited U.S. efforts to slash this dependence and secure new supply chains. Beyond the Biden administration’s Inflation Reduction Act, which included massive subsidies aimed at jump-starting a domestic mining industry, U.S. lawmakers have introduced legislation to try to boost the U.S. mining workforce and develop a comprehensive critical mineral strategy. 

As this momentum builds, companies are increasingly looking into how AI can potentially advance Washington’s critical mineral ambitions, particularly when it comes to mineral exploration. 

“I think right now the use of AI to expedite speed, reduce cost, improve efficiency of exploration is in a pilot phase,” said Gracelin Baskaran, the director of the Project on Critical Minerals Security at the Center for Strategic and International Studies, a Washington-based think tank. “And I think if it is successful, and the faster it moves, and the cheaper it gets, it’s absolutely something we can scale up.”

This push has picked up speed as U.S. lawmakers have intensified their bid to cut Washington’s mineral supply chain dependence on Beijing. In the latest effort, last month, a group of bipartisan lawmakers, including U.S. Sens. Marco Rubio and Mark Warner, introduced legislation intended to boost support for U.S. critical mineral projects. Rubio also proposed a bill that would drastically hike U.S. tariffs on Chinese critical mineral products, including a 150 percent tariff on items manufactured by Chinese-owned entities and even higher tariffs on goods manufactured in China. 

“These measures would boost private sector confidence and allow us, our allies, and our partners to develop a critical mineral supply chain impenetrable to Beijing,” Rubio wrote in an op-ed in the Hill. “They are tough measures, to be sure, but that is precisely what we need.”

The AI push reflects private firms’ efforts to innovate and streamline processes in an industry that is confronting major talent shortages and an expertise gap that emerged after the United States decided decades ago to largely outsource mining. 

It’s not just a workforce squeeze that’s complicating matters, though; Simon Jowitt, an economic geologist at the University of Nevada, Reno, said that the U.S. mining industry is now grappling with massively more amounts of data than it used to. “Over the last 20 years, we’ve seen an explosion in the amount of data that’s become available to a typical exploration project,” he said, citing how advancements and developments in the industry have generated more data in areas including geochemistry, mineralogy, geophysics, and remote sensing. “There’s just an awful lot of information that we’re going to have to deal with.” 

Some U.S. agencies are already looking into how AI can help address the data challenge. “Artificial intelligence (AI) holds the potential to enable an affordable domestic supply of the critical minerals (CM) and rare earth elements (REEs) needed to support America’s transition to clean energy,” according to a 2023 U.S. Department of Energy report, which noted how the technology can, through scouring datasets and other records, offer insights into the location, concentrations, species, volume, and value of mineral resources. 

Last year, the Pentagon also announced plans for an AI-based program that would approximate the market prices of key critical minerals and forecast their supplies. The Defense Advanced Research Projects Agency has run competitions focusing on how AI can enhance surveys of the United States’ existing mineral resources. 

Still, the AI push has largely been driven by the private sector, not the government, said William Xu, a researcher at Stanford University’s Mineral-X program, which focuses on how technological innovation can help build more resilient mineral supply chains. Even thousands of miles away from Washington, D.C., in California, questions of geopolitics still loom over the industry. China is the “undercurrent to every conversation, to everything,” he said. 

The company at the forefront of the AI charge is KoBold Metals, a Silicon Valley start-up that has partnered with Stanford’s Mineral-X program and has been backed by tech titans Bill Gates and Jeff Bezos. The firm has more than 60 projects across four continents, according to its website, and in 2023 invested around $100 million on research and development. All of the firm’s financing has been from private funders, Josh Goldman, the firm’s co-founder, told Foreign Policy. 

“The only way that we’re actually going to have enough critical minerals, and we’re going to have them at prices that can support widespread deployment around the world, is if we get a lot better at finding really high-quality sources of these critical metals,” he said, adding that “the industry’s methods are not improving fast enough.”

KoBold is perhaps best known for its Mingomba project in Zambia’s Copperbelt, an aptly named region known for its copper riches and mining. Mingomba contains 247 million tons of ore, at an average grade of 3.64 percent copper, considered to be high-grade, the firm said in 2022. In February, Goldman made waves when he announced that KoBold discovered that Mingomba would be “one of the highest grade, large underground mines.” AI technology has been “absolutely critical and instrumental throughout our involvement in this project,” he told Foreign Policy. 

“What we’re doing throughout any project like this is we’re trying to predict: Where is there the highest-quality mineralization?” Goldman said. “What we are doing with our technology is we are predicting all of these quantities, we’re quantifying uncertainty, and then we’re using that to guide our decision-making about what to do,” he added. 

The discovery was celebrated by Zambian President Hakainde Hichilema, who said it could be one of the world’s three biggest mines. “It won’t be just the largest mine in Zambia, but it will be one of the largest mines in the world,” he told Bloomberg. “We believe it will produce well over—when it’s fully operational—500,000 to 600,000 metric tons.”

It’s not just KoBold that has turned toward AI, either; another company, Earth AI, touts how it “discovers untapped critical metal deposits at half the cost in a fraction of the time.” Mining giant Rio Tinto has been experimenting with the technology, and U.S. Critical Materials, a private rare-earths firm, has also announced plans to deploy AI in its minerals exploration efforts.

Still, the jury is out on whether the technology will truly be a game-changer in the industry. Previous bids to use AI in mining have drawn criticism for being overhyped, and in the past, other mining companies have also outlined big ambitions for their AI technology—bets that largely failed to materialize. 

“I don’t think we’re at the stage where artificial intelligence, machine learning, [is taking] credit for any kind of discoveries or anything like that,” Jowitt said. “What I think it will have use in is perhaps enabling people to interpret data quicker, and more thoroughly, and basically help people cope with the increasing amounts of data that have been generated.”

“We need to understand the limitations of what this can do and how we should be using it,” he added. 

Baskaran said that the technology would likely be particularly important for exploration in emerging market economies. “Emerging markets are perceived to be more risky; most of your emerging markets are under-mapped or unmapped or on decades-old mapping,” she said. “But what it does is it lets us make progress in a way that we otherwise wouldn’t in really resource-rich jurisdictions.” 

“I think it’s very promising technology,” she said, “but it is still somewhat early stage.”