Demand for critical minerals is an inevitable byproduct of the energy transition. According to the International Energy Agency’s projection for what will be needed to reach net zero emissions by 2050, the combined market value of copper, lithium, nickel, cobalt, graphite, and rare earth elements — each crucial to batteries and other cleantech components — will more than double by 2040, to $770 billion.
Today, exploration and processing of these minerals and metals is concentrated in China. But the United States and many others, including Australia, are working to develop their own critical mineral markets. And their first order of business is the challenging endeavor of locating the deposits.
This is where Earth AI aims to help. Since years before today’s boom in artificial intelligence, the startup has used the technology to find metals and minerals that are key to the clean energy industry. In November, for instance, Earth AI partnered with the Australian mining company Legacy Minerals to locate one of the country’s biggest greenfield palladium mineral systems.
“Despite the tremendous global need for mineral resources for everything from the energy transition to day-to-day life, new mineral deposit discoveries are notorious for being expensive and time consuming,” said founder and CEO Roman Teslyuk.
Today, the company is announcing its $20 million Series B round, led by Tamarack Global and Cantos Ventures. The oversubscribed round also included investments from Overmatch, Alpaca, Sparkwave Capital, as well as existing investors like Y Combinator and Scrum Ventures.
Operating primarily in Australia, Earth AI uses a proprietary “Mineral Targeting Platform” to locate potential mineral deposits from publicly available data. Since the 1970s, the Australian government has filed all drill results and geological surveys in a central repository. However, much of that data is highly disorganized and chaotic: “Imagine what a 1970s wildcatter submitted to the government when they were told they had to provide their data,” said CFO Monte Hackett.
After years of combing through and enhancing those data sets, the company’s machine learning algorithm has 200 million data points. Armed with that information, Hackett explained, “it takes all this information learning from the past failures and success of explorers and overlays publicly available satellite imaging and remote sensing to create a treasure map of over 600 X’s across Australia that are viewed to be highly prospective for mineral deposits.”
As a result, Earth AI reports striking “economic-grade minerals” at three in four of the locations where it drills, as compared with the typical industry hit rate of one in 200.
The company owns most of its projects outright, but it also partners with two Australian explorers: Tivan, in addition to Legacy Minerals. The business model is to use the algorithm to identify potential deposits, and then use Earth AI’s own drilling technology to prove the size of those deposits before selling them to “large cap miners,” such as BHP or Rio Tinto.
“There are 50 mining companies with a market cap greater than $7 billion,” Hackett said, “and they all need new assets like ones we find in order to grow.”
Expanding the pipeline in a growing market
Earth AI plans to use the new money to develop its AI and drilling technologies, and substantially increase its project pipeline.
At present, Hackett said, the company has seven projects in the works, with run-rate drilling of about 12,0000 meters per year, at a cost of $86 per meter. (He noted that the Australian industry average is about $300 per meter, and the rate in the U.S. and Canada can be significantly higher.) The plan is to increase that to projects at over 50 sites, and increase drilling capacity to 100,000 meters, with costs continuing to hover around $100 per meter.
“Our drilling capacity increases more than our projects, because we likely will spend more time drilling current projects to expand the proven size of the assets than we have in the past,” Hackett explained. “Every time you drill an additional hole and consistently strike metal, that expands the implied area of the mineral deposit and [its] value.”
The arena of AI-enabled mineral discovery is a growing one. KoBold Metals, backed by Bill Gates and Jeff Bezos, is likely the furthest along and best-funded, having just raised a $537 million Series C earlier this month. But other startups include GeologicAI, which uses a “core scanning robot,” as well as the mineral asset generator VerAI. And in 2023, Google’s DeepMind launched a deep learning tool called Graph Networks for Materials Exploration, which can predict the stability of new materials.
