Miners back US government commodity forecasting model despite AI concerns

Mining companies said a planned new commodities forecasting system backed by the US government and driven by artificial intelligence could be a welcome new window into critical mineral markets.

The initiative, dubbed the Open Price Exploration for National Security (OPEN) program, is intended to develop a commodity structural price prediction and supply and demand forecasting system supported by AI, according to US government program solicitation documents updated in November 2023.

Current pricing and market forecast systems are overly opaque, the US Defense Department's Defense Advanced Research Projects Agency (DARPA) said in the solicitation documents. DARPA is leading the OPEN project in partnership with the US Geological Survey.

Critical minerals such as rare earths and graphite, and even cobalt and lithium, are traded in small quantities compared to base metals such as copper or iron, making accurate price signals difficult. DARPA said it aims to develop a new forecasting system separate from what it considers "flawed" data like futures markets. The agency hopes the improved transparency and forecast accuracy can boost US markets.

Mining companies reached by S&P Global Commodity Insights said they would welcome another data point to consider when transacting their product, though they had some concern about the possibility for errors when using AI-based algorithms.

"While any valid resource that could provide unbiased visibility into future prices would be welcome, our main concern would be the accuracy of an AI-based pricing and supply tool as well as the potential for data to be manipulated," Austin Devaney, chief commercial officer at Piedmont Lithium Inc., told Commodity Insights in an email.

AI at the helm

DARPA did not outline exactly how AI models should be integrated into OPEN in its program solicitation, nor did it specify whether applicants will need to use generative AI of the kind used by OpenAI's ChatGPT. A spokesperson for the trading center The London Metal Exchange Ltd. said AI systems are already in use within metals markets.

"Machine-based learning and AI have long been used to analyze supply-demand fundamentals by participants in our market, and the use of this technology will no doubt continue to evolve," the spokesperson said in an email.

Mineral exploration company Talon Metals Corp. also expressed support for OPEN.

"Competition among pricing mechanisms is a good thing," Todd Malan, chief external affairs officer and head of climate strategy at Talon Metals, told Commodity Insights in an email. "Potential risk to commodities markets is overblown and being advanced by interests that benefit from the status quo and are trying to protect it."

While AI-based models can be useful, their limitations should be understood by their users, said Ellen Wald, nonresident senior fellow with the Atlantic Council's Global Energy Center and co-founder of consulting group Washington Ivy Advisors.

"I think that especially if this program publishes its results and makes them available to the general public, it's always good to have another data set, another piece of information, both for commercial businesses and so forth," Wald said. "But you can't look at it as like 'this is what the price is going to be.'"

Markets for critical minerals are small, and there are no major exchanges for many of them, limiting visibility for producers and prospective producers. CME Group rolled out a new lithium carbonate futures contract in July, one of the few efforts to create a new contract for some lightly traded battery materials. DARPA, in its solicitation documents, acknowledged existing resources but said the ongoing gaps in data across critical minerals markets led to high levels of uncertainty and risk.

"I think the idea is to help companies understand where the pricing is going so that they can make educated decisions for where they want to invest," Wald said.

The National Mining Association, a trade group representing US miners, declined to comment.

Prices for critical minerals have faced volatility as supply looks to find balance amid growing demand for metals related to the energy transition.

Platts price assessments for lithium carbonate and lithium hydroxide, both on a DDP China basis, reached record highs in November 2022 but have since fallen 84% and 85%, respectively. As of Feb. 6, Platts assessed lithium carbonate on a DDP China basis at 96,000 yuan per metric ton and lithium hydroxide at 85,000 yuan/t.

Platts assessed lithium carbonate on a DPP US basis at $15,400/t on Feb. 5.

Looking ahead

DARPA is seeking contractors to develop forecasting prototypes for OPEN under two initial technical areas: prices, and supply and demand.

S&P Global Inc., the parent company of Commodity Insights, and Lockheed Martin Corp. have both applied as contractors, Reuters reported Jan. 29. Lockheed Martin deferred comment to DARPA. S&P Global Commodity Insights, the division of S&P Global that offers Platts pricing assessments, also deferred comment to DARPA. DARPA declined to comment on contractor selection dates.

For price forecasting, contractors will focus on algorithms to generate probability distributions over structural prices for various critical materials. The US Geological Survey's critical minerals list will serve as the basis for critical minerals covered by OPEN. DARPA emphasized that algorithms involved in the program should not mimic futures markets, which seek to predict commodity prices in publicly traded markets.

DARPA defined structural price "as the sum of the input costs incurred to create one additional unit ... of that material at a particular time and geographic location." The agency also expects contractors' OPEN price models will be adjustable for as yet unspecified regions and technologies, including certain production methods.

For supply and demand forecasts, DARPA wants contractors to develop algorithms that can be generalized across critical mineral extraction and refining processes but indexed by time and location.

The agency highlighted a need for forecasting algorithms to be capable of considering uncertainties in demand generation, including battery chemistry innovation and increased material recycling.

For both initiatives, DARPA will provide a data application programming interface and baseline models. Contractors are not expected to rely on any unique data sources they might have as part of their application to contribute to the program, and prediction windows for both systems should span at least three months, the agency said.

Platts price assessments for lithium carbonate DDP China, lithium hydroxide DDP China, and lithium carbonate DDP US are offerings of S&P Global Commodity Insights. S&P Global Commodity Insights is a division of S&P Global Inc.