Most American policy discussions frame China's dominance in critical minerals as a problem of ownership. China controls the mines, the argument goes, so America must secure mines of its own. That framing is not wrong, but it misses the part of the problem that determines who actually wins, and it leads to responses that cannot succeed on their own.
Consider what the data actually shows. The International Energy Agency's Global Critical Minerals Outlook found that for 19 of the 20 most important strategic minerals, China is the world's leading refiner, with an average market share of about 70 percent. For the rare earth elements that go into permanent magnets, China accounted for roughly 60 percent of global mined production in 2024 but 91 percent of refined output. For gallium, an input to the semiconductors inside radar systems and electronic warfare equipment, Chinese refining approaches 99 percent of global supply.
Read those numbers carefully. The gap between China's share of mining and its share of refining is the whole story. Sixty percent of mining is a strong position. Ninety-one percent of refining is control. And refining is not geology. Nobody's rocks are better at being refined. Refining is chemistry, process engineering, energy management, logistics, and above all the systematic optimization of industrial operations at scale. China's decisive advantage sits in the middle of the value chain, at the processing stage, and the processing stage is an operations problem.
Optimization as national strategy
This did not happen by accident, and it is not slowing down. In August 2025, China's State Council issued its guideline on the Artificial Intelligence Plus Initiative, a directive that treats AI-driven optimization as national infrastructure. The stated targets are explicit: penetration of next-generation intelligent applications exceeding 70 percent by 2027 and 90 percent by 2030, with the intelligent economy positioned as a principal driver of growth.
The manufacturing base is being rebuilt around this thesis. China's Ministry of Industry and Information Technology reports more than 30,000 basic-level smart factories in operation, alongside 1,200 advanced-level and 230 excellence-level facilities spread across all 31 provincial regions and covering more than 80 percent of manufacturing sectors. The ministry's own figures for the excellence tier are worth pausing on: product development cycles 28.4 percent shorter and production efficiency 22.3 percent higher than conventional plants. Whatever discount one applies to self-reported statistics, the direction and scale of the program are not in dispute, and the accompanying policy documents specify the machinery behind it, including industrial AI models for every sector and a thousand high-level industrial AI agents.
The pattern across mineral processing and manufacturing is the same pattern. China's position was built by applying process optimization, industrial engineering, and now machine learning to operations, relentlessly and at scale, for decades. The mines came later and matter less. The moat is operational.
Why buying mines is necessary but not sufficient
This is the uncomfortable implication for American strategy. Stockpiles, equity stakes, and new mines address the ownership problem. They do not address the optimization gap. Ore that cannot be refined domestically at competitive cost still flows through someone else's processing capacity, and processing capacity is won or lost on operational performance: recovery rates, energy per ton, throughput, uptime, quality control.
The stakes of getting this wrong are quantified. The IEA estimates that meeting demand for magnet rare earths outside the dominant supplier requires around 60 billion dollars of investment over the next decade, with refining alone accounting for nearly half, and it puts the potential economic cost of supply disruptions at as much as 6.5 trillion dollars. Every one of those new refining and processing facilities will succeed or fail on the quality of its operations. Which raises the question the investment announcements rarely ask: who is going to run them?
The workforce is the binding constraint
A processing facility is not made competitive by its equipment. It is made competitive by people who can read its sensor data, model its processes, schedule its production, predict its failures, and improve its yield month after month. That is the work of industrial optimization, and the United States does not currently train enough people to do it at anywhere near the scale its own industrial strategy requires.
This is not a controversial claim; it is now official policy. In August 2026 the federal government committed more than 180 million dollars to mining and minerals education, including a Department of Energy initiative whose stated goal is to double the number of graduates in mining, minerals, and supply chain fields within two years, with eligibility extending beyond universities to community colleges, trade schools, and industry partners. That last detail is the significant one. A four-year degree pipeline cannot double its output in two years, because the students who will graduate in that window are already enrolled. The government's own program design acknowledges what the arithmetic requires: the near-term workforce answer must come from high-throughput training pathways outside the traditional university pipeline.
Closing the optimization gap therefore means building training infrastructure, not only processing infrastructure. It means taking the methods that made Chinese processing dominant, which are, at bottom, teachable methods of mathematics, data analysis, and operations research, and putting them in the hands of the technicians, operators, and plant engineers who will staff the facilities America is now committed to building.
That is the premise of Project FORGE. The curriculum starts from zero because that is where much of the industrial workforce starts, and it runs all the way to the optimization and machine learning methods that determine competitive processing operations, taught through the industrial problems where they will actually be used. The optimization gap took decades to open. Closing it is a training problem before it is anything else, and training problems are solvable.
Sources
- International Energy Agency, "Global Critical Minerals Outlook 2025" and "With new export controls on critical minerals, supply concentration risks become reality" (iea.org)
- International Energy Agency, "Rare Earth Elements," executive summary (iea.org)
- S&P Global Commodity Insights, "Rare earth supply bottlenecks set to persist in 2026" (January 2026)
- World Economic Forum data on refined mineral production shares, as compiled by Visual Capitalist, "China's Grip on Critical Mineral Refining" (2026)
- State Council of the People's Republic of China, guideline on the "AI Plus" Initiative, official release, August 2025 (english.www.gov.cn)
- Ministry of Industry and Information Technology smart factory program figures, as reported by China Daily, February 2025
- Center for Security and Emerging Technology, Georgetown University, translation of the "AI + Manufacturing" Special Initiative Implementation Opinions (March 2026)
- U.S. Department of Energy, "Energy Department Launches $100 Million Initiative to Build America's Critical Minerals Workforce" (energy.gov, August 7, 2026)