AI in Mining Exploration
Rapid advancements in technology are reshaping the traditional landscape of commodity markets. As global demand for copper soars, driven by the rise of electric vehicles, generators, and electrical wiring, the need to explore new mineral deposits intensifies. Artificial intelligence is now enhancing traditional exploration methods.
Despite rising copper demand, the number of suppliers of this strategic raw material remains very limited. In 2023, the world’s production of primary copper was estimated at about 22 million tons, an increase from just over 15 million tons in 2006 (a 47% increase over 17 years).
By the end of this decade, global copper needs are projected to reach 30-35 million tons annually (about a 60% increase from current levels). The International Energy Forum estimates that the global industry will require approximately 200 new major copper mines by 2050 to meet demand.
Herein lies the main challenge. Although regions with potential copper ore deposits are relatively well-known (today, over 50% of production is concentrated in just four countries: Chile, Peru, Congo, and China), the methods for geological exploration of new deposits, which have scarcely changed in the last century, are ill-suited to today’s realities.
The more you realize how dependent we are on these technologies, the more you ask: How the hell were we so slow to the fact that we needed vast amounts of raw material to make it all possible?reflects Connie Chan, a General Partner at one of Silicon Valley's largest venture capital firms, Andreessen Horowitz.
Understanding the critical demand for copper in AI development, the California-based company KoBold Metals chose a straightforward and logical approach: if AI needs copper, why not let AI find the necessary deposits?
While partially humorous, this idea underpins a serious initiative. The company has dedicated several years to developing a global AI platform for analyzing geological data and has been involved in mineral exploration projects across 60 countries.
In mid-July 2024, KoBold announced to investors the discovery of a new copper deposit in Zambia, projected to yield 300,000 tons of copper annually. Interestingly, this region was already recognized for its copper mines. Yet, finding a rich vein deeper than 1 km had not been accomplished until now.
KoBold developed the TerraShed database, which aggregates geological survey results from the last century, historical documents, expedition reports, radar data, aeromagnetic survey data, and more. For field exploration, the company utilized muon detectors—subatomic particles emitted by rock formations.
Although muon detectors have been employed in various contexts, such as studying ancient Egyptian tombs and detecting illegal smuggling tunnels, KoBold is pioneering their use in analyzing rock density.
This approach offers a clear advantage: instead of slow and costly traditional drilling methods for sample collection, a small excavation suffices to deploy a device equipped with the detector to the targeted depth.
The main challenge lies in distinguishing required anomalies from a broad spectrum of emissions. Traditional empirical methods demand years of meticulous scientific work. AI's role is to accelerate this analysis exponentially. It took just over a year for the company to locate and characterize the deposit, a process that might have taken decades with conventional methods.
We don’t drill for metals, we drill for information. It puts the science into eurekaremarked Kurt House, CEO of KoBold Metals, indicating that this innovative method of exploration could be extended to other minerals.
KoBold has already secured $2.3 billion in funding for its first copper mine in Zambia and is in discussions with contractors and governments. Notably, the company is hopeful that the U.S. will finance the construction of a railway to facilitate copper exports, viewing Zambia as a strategic counter to China's growing influence in Africa.
Kurt House noted that TerraShed now contains about 3% of all known geological data. Following KoBold’s successful demonstration, it is likely that other companies will emulate this approach. Artificial intelligence could indeed be key to overcoming future raw material crises.
