The rare earth separation only China masters has become a quantum computing problem

USA Rare Earth, Pasqal and Riven Systems have joined mining, quantum computing and an automated chemistry lab to look for more efficient molecules for rare earth separation — today the real bottleneck in the chain, not mining itself.

USA Rare Earth, owner of the Round Top mine in Texas and of a magnet plant under construction in Oklahoma, announced on September 17, 2026 a three-way partnership with France's Pasqal, which specializes in neutral-atom quantum computing, and Riven Systems, which runs an automated chemistry lab. The stated goal is to develop more efficient extractant molecules to separate mixed rare earth carbonate — MREC — at a stage of the supply chain that today is dominated almost entirely by Chinese processors, regardless of where the raw ore was mined. Riven will run thousands of automated experiments to generate data on how different molecules bind selectively to each rare earth element. That data feeds two parallel paths: classical machine learning models and quantum machine learning models running on Pasqal's neutral-atom processor, compared side by side to see which one finds better candidates. Test feedstocks include ore from Round Top itself, MREC bought from third parties and recycled magnet manufacturing scrap. None of the parties disclosed a commercialization timeline, the amount invested or any technical result obtained so far — the announcement marks the start of the collaboration, not the delivery of a finished technology.

The detail that rarely makes the headlines about the race for rare earths is that China's advantage was never mainly geological — it is chemical and, above all, a matter of process engineering accumulated over decades. The seventeen rare earth elements are chemically almost identical to one another, an effect known as the lanthanide contraction, which makes separating one element from another a problem of extreme precision. The standard method, solvent extraction, typically requires hundreds of cascading mixer-settler stages, each slightly refining the purity of the separation — an expensive process that consumes a lot of reagent and energy, and that Chinese industry has mastered after more than thirty years of optimizing the recipe molecule by molecule, by trial and error at industrial scale. That trial-and-error process is exactly what the partnership is trying to compress. The idea behind quantum machine learning is that simulating how a specific molecule binds to a rare earth element is, at its core, a quantum mechanics problem — electron orbitals interacting with each other — and quantum computers may be more efficient than classical ones at simulating precisely this kind of interaction. If the bet pays off, the result would not just be finding a better molecule, but reducing the number of separation stages needed, which directly shrinks the size, capital cost and energy consumption of a separation plant — the barrier that today keeps most Western attempts from competing with China at this specific stage of the chain, even when the mining itself is already solved. The division of labor also matters as a model: Riven provides real experimental data in volume, Pasqal contributes quantum simulation capacity, and USA Rare Earth ensures any promising candidate is validated under real processing conditions at its research center in Wheat Ridge, Colorado, before any commercial use is considered — none of the three tries to solve the problem alone.

For anyone following the Western effort to reduce dependence on China for rare earths, this partnership points to where the real bottleneck lies: not ore extraction, but the chemical separation that comes after it. The United States, Australia and European Union countries already have or are developing rare earth mines outside China — what is still missing is separation capacity that is competitive in cost and scale outside Chinese territory. A technology that reduces the number of solvent extraction stages needed would make smaller, cheaper separation plants viable outside China, widening the number of players able to enter this specific stage of the chain. That does not mean the outcome is guaranteed — quantum machine learning applied to chemistry is still a research frontier, with no proven commercial cases at industrial scale in any sector so far, and neither Pasqal nor USA Rare Earth has promised a timeline. But the simple fact that a miner, a quantum computing company and an automated lab decided to bet together on this specific combination shows the industry is willing to test unconventional paths to solve a problem that the traditional approach, decades of incremental chemical engineering optimization, never solved outside China.