The United States is rushing to reduce its dependence on foreign sources of important minerals.
Raw materials are essential inputs for the latest technology. Military weapons like smartphones, 5G networks, and fighter jets are built using these materials and continue to be used to promote innovation. Lithium drives electric vehicle batteries, copper continues to run data centers, and silicon forms the foundation of semiconductors.
Market research firm Kings Research says that demand is growing rapidly, with global key mineral markets projecting to reach nearly $500 billion by 2030. However, the US remains heavily dependent on imports. As of 2024, the country imported 100% of 50 designated important minerals, including graphite, manganese and gallium, according to the US Geological Survey. Many of them come from China.
Now, under the second President Trump, the United States is pushing to reestablish its control in mineral production. The administration has made domestic manufacturing a national priority, issued executive orders to promote mining in American soils, and imposed a 50% tariff on imported metals such as steel and aluminum.
Restructuring of domestic mineral supply could strengthen the economy and improve national security. However, it can also lead to supplying gaps, potentially increasing the cost of materials for power innovation.
To prevent bottlenecks, startups and legacy technology companies are turning their eyes to artificial intelligence. Their AI tools promise to speed up mineral discovery and reduce supply chain risks in unstable geopolitical climates.
However, in slow, highly regulated industries, some experts have questioned whether AI can fulfill its promise.
Startups are racing to rebuild mineral exploration
Startups are betting on AI to discover new mineral deposits – and some are seeing early results.
Earth AI uses prediction software and proprietary drilling hardware to find, verify and sell billion-dollar mineral projects. The algorithm is trained on decades of historical data from Australia, including past successes and failures in mineral discovery, to identify hydrothermal systems.
After identifying promising sites, the company will train using its own rigs and analyze rock samples to check for metal presence. Once proven, the site will be sold to large mining companies.
EarthAi said the success rate was 75% – Well above the industry average of less than 1%. Over the past 12 months, the company said it has already made three discoveries in Australia. One of them is indium, a rare metal used in touchscreens and semiconductors for AI hardware. They also discover undeveloped reserves rich in minerals. In late July, Earth AI software identified a large underground nickel and palladium land on Australia's east coast.
Using AI, Earth AI told Business Insider that it can reduce its mineral discovery timeline from years to months.
“I think we can create the most value by digging into the ground. Monte Hackett, CFO of Earth AI.
Terra AI is also betting that AI can speed up the industry's slow discovery process.
“Even with decades of investment in sensors and data, we're getting worse every year,” said John Mern, co-founder and CEO of Terra AI. “The amount of metal added to global supply this year is 90% lower than in 1990.”
Terra's software is used using AI to ingest layers of geological data such as magnetic field measurements and seismic activity to generate thousands of underground maps to identify the most promising locations.
Mern said the AI-first approach is already being piloted by rare earth projects in the US and mining companies in the US, Africa and Europe. He added that Terra's platform can cut the average mine development timeline for 17 years by half.
Investors see mining AI applications as a major opportunity. UK-based venture company Founders Factory recently partnered with Mining Giant Rio Tinto to launch an accelerator that supports 12 startups a year, including Terra AI.
Founders Factory investor Jack Kennedy is looking at the $2 trillion industry and mining “untouched” by Tech Innovation. He compares mining to a “waste management business,” where a large amount of earth travels to extract small amounts of metal.
“AI is basically a way of trying to process a large number of different data points to increase efficiency,” Kennedy told BI. In doing so, he adds, translated into reducing waste, costs and environmental impact.
Legacy players use AI to protect their supply chain
Legacy companies are also taking part in the action.
Exiger, a supply chain management software provider, helps governments and Fortune 500 companies track and protect their critical mineral supply chains. Its AI model breaks down the product into digital twins – a detailed virtual version that maps internal materials – tracks the material composition of a product using a database of 10 billion transaction records.
The database includes commercial datasets purchased from custom brokers and invoice processes, financial data, engineering specifications, build-to-print drawings, material declarations, and manufacturing process documents.
Using AI to assess a company's supply chain allows clients to visualize vulnerabilities within their mineral supply chains that may be overly dependent on specific countries and geopolitical risks. Second, clients can make informed decisions when adjusting their mineral supply chain strategies.
In one case, Exiger was able to identify ways to extract germanium, a rare earth mineral used in fiber optics and chips from US coal ash and smelter waste, potentially reducing foreign dependence.
“When China restricted exports in rare earths, it exposed its customers to price volatility and geopolitical uncertainty,” Exiger CEO Brandon Daniels told BI. “Our platform helps clients navigate that risk at a level of accuracy that they have not achieved previously.”
AI Limitations
Still, AI is not a magical solution.
Rajive Ganguli, a professor of mining engineering at the University of Utah, said that with decades of experience applying AI in the field, the technology is as good as the data he is trained. High-quality hard data, such as drill hole information and physical samples, is often small, expensive and difficult to obtain.
“Bad numbers of AI don't provide a good answer,” Gangari told BI. He also points out that many “AI discoveries” occur in areas already known to geologists. He says the technology doesn't work well in untapped, data-lacking areas.
Additionally, the startup who spoke to BI said it was skeptical that mining would adopt new technology and make AI adoption a difficult battle in the critical mineral discovery process.
That said, Gangari believes that the biggest obstacle to expanding mineral production is systematic rather than technical. In the US, businesses often wait 10-15 years to approve a permit.
Despite the Trump administration's move to allow quick trucks, the process remains a bottleneck in the short term. Despite early consultations with US clients, Earth AI has not started exploring domestically due to delays that could be dragged for years.
Experts agree that AI is not a replacement for humans. Geologists and engineers remain crucial to interpret the output of AI and make final decisions about where to train.
“This is not a laboratory issue,” Ganguli said regarding mineral discovery, adding that “domain experts” are important to understand how machines work, what data means, and how sites work in practice.
Still, companies believe that AI can play a key role in strengthening the US mineral supply chain.
But even with the best tools, the US is unlikely to do it alone.
“The reality is that a large part of our supply chain will cross our borders,” said Terra AI CEO Mern. “We need a responsible international partner to ensure that.”
