The US Is Building a ‘Google Maps for Microbes’ to Break China’s Rare-Earth Monopoly. It Might Just Work

Highlights

• Scientists at National Laboratory of the Rockies and PNNL are using AI and machine learning to identify proteins that naturally bind rare earth elements, creating a “Microbial Rare Earth Element Atlas” for domestic bioprospecting.
• Lawrence Livermore’s “Spicy Lambs” platform can screen 600 protein variants in one month versus 3–5 years traditionally, accelerating engineered biomining solutions that require no toxic solvents or high-temperature processing.
• With China controlling 85–90% of rare earth refining and tightening export controls, bio-based extraction could provide defense-critical domestic supply within a decade, backed by over $1 billion in federal funding.

---

For as long as the global economy has depended on rare earth elements, the United States has faced a simple, brutal math problem: China mines roughly 60 percent of the world's rare earth ore and controls an estimated 85 to 90 percent of global refining and processing capacity. For heavy rare earths, the ones essential for high-powered magnets in guided missiles, MRI machines, and electric vehicles, China's grip is even tighter, with analysts estimating it holds close to all meaningful separation capacity. The US is entirely import-dependent for 16 non-fuel minerals.

That vulnerability has been laid bare repeatedly over the past year. In April 2025, China tightened export licensing requirements for rare earth elements. By October, the controls expanded to include rare earth magnets and processing technologies. Within weeks, global EV production came under pressure, NATO countries began expanding strategic stockpiles, and analysts estimated it would take a decade for the West to build competitive mine-to-magnet infrastructure. A separate, short-lived trade truce on gallium, germanium, and antimony offered partial relief, but the underlying rare-earth problem remained unresolved.

A quiet research announcement made in early May 2026 now suggests a wholly different path forward, one that involves not digging deeper, but mining the microbial world for answers.

A 'Google Maps for Microbes'

On May 7, 2026, the National Laboratory of the Rockies and the Pacific Northwest National Laboratory (PNNL) unveiled (opens in a new tab) an early-stage research initiative that could fundamentally rewrite the economics of rare-earth extraction. Their goal is deceptively simple: rather than prospecting for rare earths with a shovel and a test tube, they will use machine learning to scour the natural world for proteins that can bind to rare earth elements. Combined with the first-ever Microbial Rare Earth Element Atlas, the team plans to build a bioprospecting tool capable of identifying geographic areas rich in these metals and optimizing proteins' abilities to bioaccumulate rare earth elements on US soil.

Behind that description lies a substantial shift in how materials science is being approached. PNNL is already leading the Orchestrated Platform for Autonomous Laboratories (OPAL), a Department of Energy initiative that deploys AI agents and robotics to run biology experiments autonomously, compressing research timelines from years to weeks. The platform integrates mass spectrometry and metabolomics instruments that feed directly into AI systems designed to spot patterns and suggest next steps without human intervention.

"For biology, one of the most underappreciated and difficult tasks in AI training is finding data that's in the right format," said Chris Oehmen, a researcher on the project. "Through OPAL FAMOUS, we are creating agents that ingest raw data and translate it into terms that an AI agent can understand and then act upon."

The 'Spicy Lambs'Revolution

The Rockies-PNNL initiative is not happening in isolation. Across the US national laboratory system, a broader convergence is underway, one that marries the discovery of natural rare-earth-binding proteins with high-throughput, AI-driven engineering.

The poster child for this movement is a protein called lanmodulin, discovered in bacteria that use rare earth elements as a metabolic cofactor. Lanmodulin binds lanthanides with picomolar affinity and extraordinary selectivity. For years, however, studying it was painfully slow.

Traditional protein screening methods test one variant at a time, making large-scale discovery impractical.

That changed in April 2026, when researchers at Lawrence Livermore National Laboratory published a breakthrough in Nature Chemical Biology. Their platform, named SpyCI-LAMBS and nicknamed "spicy lambs," enables parallel screening of hundreds of protein variants in a single 96-well plate run.

"It only took about a month to collect 600 proteins' worth of data with this new assay," said LLNL scientist and first author Patrick Diep. "It would have taken three to five years with the usual process."

The platform works by attaching engineered tags to lanmodulin proteins so they automatically bind to a solid surface, eliminating the complex purification steps that traditionally limited throughput. In a single round of screening, the team identified eight distinct protein clusters with different rare-earth selectivity patterns. One group of more than 200 variants showed improved performance in separating light rare earth elements, a key challenge that has long frustrated conventional chemical separation methods. The data generated is being used to train machine learning models that predict how proteins will behave before they are physically tested, opening the door to predictive, data-driven design of metal-selective proteins.

A Parallel Effort: The Microbe-Mineral Atlas

While the Rockies-PNNL team focuses on protein-level bioprospecting, a separate effort at Cornell University is taking a broader and complementary approach: cataloging entire microbial communities. Led by Buz Barstow (opens in a new tab) (interviewed by Rare Earth Exchanges™) (opens in a new tab) and supported by a two-million-dollar grant from the National Science Foundation, the team is creating a Microbe-Mineral Atlas that includes specific genes and catalogs their interactions with minerals. Rather than starting with known proteins and engineering them, the Cornell approach begins with the assumption that nature has already solved the problem of metal extraction in countless ways, and that it is simply a matter of finding the right genetic recipes.

The Atlas will include genes from microbes collected from diverse US environments, with the aim of creating genetically engineered organisms for practical biomining applications. The project has already demonstrated proof of concept: a related effort working with the mineral-dissolving microbe Gluconobacter oxydans (opens in a new tab) improved rare-earth biomining yields by as much as 1,200 percent.

Taken together, the Rockies-PNNL protein-focused Atlas, the LLNL high-throughput lanmodulin engineering platform, and the Cornell Microbe-Mineral Atlas represent the most coordinated application of synthetic biology to the rare-earth supply crisis the world has yet seen.

The Financial Calculus

The global biomining market remains relatively small but is growing quickly. It was valued at roughly three billion dollars in 2024 and is projected to reach eight and a half billion by 2035, a trajectory that reflects mounting pressure to find sustainable alternatives to conventional mining.

The timing of the Rockies-PNNL initiative is also well-suited to the policy environment. In August 2025, the Department of Energy announced one billion dollars in critical materials funding to strengthen domestic supply chains. In April 2026, the DOE issued a new funding opportunity making up to 69 million dollars available through the Critical Minerals and Materials Accelerator Program (opens in a new tab), specifically targeting technologies that advance domestic production and refining. A separate 135-million-dollar funding opportunity has been issued for a Rare Earth Elements Demonstration Facility (opens in a new tab).

After a billion-plus federal deal, private capital is following. USA Rare Earth received a $1.6 billion investment commitment in January 2026 to extract critical mineral feedstock from its Round Top deposit in Texas. In March 2026, Energy Fuels successfully produced the first domestically sourced high-purity heavy rare earth oxides in the US in decades, including one kilogram of 99.9 percent pure terbium oxide and nearly 30 kilograms of dysprosium oxide. The scale remains modest against a global heavy rare earth market measured in tens of thousands of tons annually, but it represents genuine proof that domestic refining capacity is being rebuilt.

The microbial approach, if it scales, could accelerate this trajectory considerably. Traditional rare-earth refining is capital-intensive, environmentally destructive, and geopolitically concentrated. Bioaccumulation using engineered proteins requires no toxic solvents, no high-temperature processing, and no massive tailings ponds. The cost per ton of separated rare earths could fall substantially, potentially making marginal domestic deposits commercially viable for the first time.

The Geopolitical Dimension

None of this is coincidental in its timing. In February 2026, the United States unveiled an initiative to form a critical minerals trade zone explicitly designed to reduce global dependence on China, including the use of tariffs to prevent Chinese price manipulation from undercutting domestic investment.

China's response has been swift. Its April 2025 export licensing controls covered seven categories of medium- and heavy-rare-earth items. By October, those restrictions expanded to include magnets and processing technologies, with extraterritorial provisions applying to products containing Chinese rare-earth content. A CSIS analysis concluded that the restrictions effectively cut off the flow of these materials to Western military manufacturers. Defense contractors face tighter sourcing requirements beginning in 2027, obliging them to secure non-Chinese rare-earth inputs even at above-market costs.

The microbial alternative offers something conventional mining cannot: a domestic supply that does not depend on discovering a world-class ore deposit. Engineered microbes can extract rare earths from low-grade ores, mine tailings, and even industrial waste. Combined with the phytomining breakthroughs reported in April 2026, the broader research portfolio points toward a future in which the US could produce meaningful quantities of rare earths from biological sources located entirely within its borders.

The Double-Edged Sword

For all its promise, bio-based rare-earth extraction faces serious technical hurdles and unresolved environmental questions.

The most immediate is toxicity. Rare earth elements, particularly when concentrated, pose inhalation hazards that have been understudied relative to their growing industrial use. A systematic review published in April 2026 in Frontiers in Public Health warned that fine REE-containing particles can penetrate deep into the distal lung, where they exhibit high biopersistence. Epidemiological evidence from mining regions links elevated REE burdens to chronic respiratory diseases and interstitial lung disease.

For bioaccumulation, the concern is even more pointed. If engineered proteins concentrate rare earths efficiently enough for commercial extraction, what happens when those proteins, or the microbes producing them, are released into the environment? The US regulatory framework for synthetic biology remains fragmented. The Cornell Microbe-Mineral Atlas project specifically includes a workstream examining whether existing regulatory guidelines must be adapted to accommodate this new form of biotechnology.

There is also the question of scale. The proteins identified by Lawrence Livermore work reliably in 96-well plates under carefully controlled laboratory conditions. Translating that performance to an industrial bioreactor handling thousands of liters of leachate, containing competing ions, variable pH, and unknown organic compounds, is a challenge that has defeated many promising biotechnologies. The first phases of PNNL's Anaerobic Microbial Phenotyping Platform identified key biological pathways for critical mineral recovery under laboratory conditions, but the gap between lab and field is precisely where most such projects stall.

Finally, the fundamental biology remains poorly understood. Despite the discovery of lanmodulin and the emergence of the field sometimes called lanthanide omics, it is still not clear why certain bacteria oxidize rare earths as an energy source, or what selective advantage lanthanide binding confers. Without a firmer grasp of those evolutionary pressures, engineering robust industrial strains will remain partly an exercise in trial and error, even with AI assistance.

What Success Would Look Like

If the Rockies-PNNL initiative delivers, three outcomes become imaginable over the next decade that are not achievable today through conventional means. The first is a domestic supply of separated rare earths sufficient to meet defense requirements. The Pentagon's mine-to-magnet strategy aims to eliminate reliance on Chinese materials for defense applications by 2027, a goal that is essentially unachievable through conventional mining on that timeline. Bio-based extraction, if effectively scaled, could provide a meaningful hedge against complete supply disruption.

The second is a substantially reduced environmental footprint for rare-earth refining. Traditional separation depends on solvent extraction using toxic organic compounds and generates large volumes of acidic waste. Engineered proteins can be immobilized on solid supports and reused repeatedly, with minimal chemical input.

The third is a more durable public narrative. The technology is sophisticated enough to be explained clearly: researchers are finding proteins that evolution has already optimized, then using AI to make them faster and more selective. That story is considerably easier to defend than the proposal to dig new mines in pristine landscapes.

The Rockies-PNNL Microbial Rare Earth Element Atlas will not deliver a domestic rare-earth industry overnight. The researchers themselves describe it as early-stage. The timeline from protein discovery to industrial biomining is measured in years, not months. But it points toward a future that was not available even five years ago: one in which the United States leverages its strengths in biology, computing, and automation to break a resource monopoly that has shaped global geopolitics for a generation. Whether that future arrives depends on sustained funding, consistent policy, and a willingness to accept that the most consequential mining technology of this century may involve no heavy machinery at all, just a microscope, a petri dish, and a machine-learning model trained to read the language of life.

---

## References

1. National Laboratory of the Rockies / PNNL announcement: “You’ve heard of prospecting for gold. But what about prospecting for critical minerals?” LinkedIn (7 May 2026)

2. PNNL OPAL autonomous laboratories: “PNNL uses AI agents to speed up discovery of microbes and plants for critical mineral extraction,” CompleteAI Training (28 April 2026)

3. LLNL SpyCI‑LAMBS platform: “US lab’s ‘Spicy lambs’ protein system cuts rare‑earth screening time from years to weeks,” *Interesting Engineering* (23 April 2026)

4. Cornell Microbe‑Mineral Atlas: “Microbe Atlas: Making Critical Mineral Mining More Efficient,” rawmaterials.net (13 November 2024)

5. Global biomining market: WiseGuy Reports (2024)

6. DOE critical minerals funding: US Department of Energy (30 April 2026)

7. US import reliance on Chinese REEs: Investment Monitor (15 January 2026); CSIS analysis cited in Legis1 (26 March 2026)

8. Energy Fuels domestic heavy rare earth oxide production: Wedoany (26 March 2026)

9. USA Rare Earth domestic investment: Texas General Land Office (27 January 2026)

10. Toxicology of REEs – *Frontiers in Public Health* (15 April 2026)