Highlights
- DOE announced two AMD-accelerated supercomputers, Lux and Discovery, through a new public-private partnership model, investing $1 billion to accelerate American AI and scientific computing leadership.
- Despite supercomputer advances, the U.S. remains vulnerable due to China's dominance over critical minerals refiningโincluding rare earths, gallium, cobalt, and graphite essential for AI infrastructure.
- While U.S. investments in MP Materials, Ucore, and Energy Fuels show progress, analysts warn that 'buying equity is not building capacity' without domestic metallurgical infrastructure spanning extraction to magnet production.
The U.S. Department of Energyโs October 27 announcement (opens in a new tab) of two AMD-accelerated supercomputersโLux and Discoveryโmarks the administrationโs latest attempt to fuse public-private partnerships into national competitiveness. Secretary of Energy Chris Wright (opens in a new tab) heralded it as a model of โcommonsense innovation,โ where shared funding and computing infrastructure supposedly compress timelines from years to months.
On paper, itโs a narrative of speed, scale, and AI sovereignty. Yet for all the talk of supercomputers, silicon, and scientific leadership, the announcement reveals a strategic imbalance at the heart of U.S. industrial policy: Washington is betting big on computation, but is there sufficient bets being made on materials security?ย That is, those ย rare earths, critical minerals, and midstream refining capacity that actually underpin such systems?
Table of Contents
Note that the new AMD-accelerated artificial intelligence (AI) supercomputers are located at Oak Ridge National Laboratory (opens in a new tab) (ORNL); and, one of the assets will be built at record speeds thanks to this new public-private partnership model according to the government media entry.
The aim? The DOE declares that such supercomputers will help expand Americaโs leadership in scientific computing, strengthen national security, and drive the next generation of Gold Standard Science and innovation.ย
Missing Metal in the Machine
While DOEโs partnerships with AMD and HPE sound transformative, the core of American vulnerability remains the sameโChina dominates the refining and metallization steps for the very elements inside these processors. AI infrastructure depends on rare earths, gallium, cobalt, nickel, and graphite, yet domestic investment in separation, alloying, and magnet manufacturing still lags.
Yes,, the U.S. has taken a prominent equity position in MP Materials, and also is spreading funding around to the likes of Ucore, and Energy Fuels, and has financed projects through the Defense Production Act and DOEโs Loan Programs Office and theย Export-Import Bank of theUnited States (EXIM) (opens in a new tab)ย and its collaboration with theย U.S. Department of Commerce, which together support American businesses in exporting their goods and services.
However, these efforts pale in comparison to Beijingโs vertically integrated control of the entire value chain. REEx analysts have long warned that, since our website launch on October 15, 2025, โbuying equity is not building capacityโโcapital alone doesnโt close a technical gap if refining IP, reagent technology, and waste treatment systems remain foreign.
The Gold Standardโor a Distraction?
The DOEโs $1 billion AI co-investment is emblematic of Washingtonโs fascination with digital sovereignty, but the supply chain foundation beneath it remains porous. ย Yes, President Trump just completed a whirlwind tour as well, signing up multiple Asian nations for mineral and element deals. Plus, the establishment of what appears to be somewhat of a yearโs reprieve with China. ย But is this all enough?
A hard, critical question to ponder.ย Does the rhetoric about โAmerican energy and national securityโ in reality skip over the fact that these supercomputersโand the broader electrification and defense ecosystemsโstill depend on foreign-processed magnet metals and cathode materials, without a national metallurgical strategyโspanning ionic-clay HREEs, solvent extraction, alloy smelting, and magnet sinteringโthe U.S. risks owning the worldโs fastest computer that canโt model its own materials shortage.
The REEx Take: Progress Without Parity
Credit where due: DOEโs public-private model is faster, leaner, and more collaborative than previous procurement regimes. ย And the Trump administration, to its credit, is doing far more than previous ones to address the critical need to re-industrialize America (to be fair Joe Bidenโs CHIPs and Science Act (opens in a new tab) also appeared to show potential for impact).
Yet until equivalent urgency and scale are applied to midstream infrastructure, this remains a technological shell atop a hollow supply chain. Washingtonโs approach needs less fanfare and more furnaces at scale.
Citation: U.S. Department of Energy (2025). Energy Department Announces New Public-Private Partnership Model, Two Supercomputers, to Accelerate American Dominance in Science and Technology.
ยฉ 2025 Rare Earth Exchangesโข โ Accelerating Transparency, Accuracy, and Insight Across the Rare Earth & Critical Minerals Supply Chain.
0 Comments