π DATA CENTER WASTE
π CRITICAL MINERALS
π°π· KOREA ZINC
β
AI hardware upgrade cycles of 3β5 years are quietly creating one of the largest untapped mineral reserves on earth. Data centers are retiring GPU fleets packed with gallium, terbium, and dysprosium β and Korea Zinc is already in talks to reclaim them.
ποΈ 24, May, 2026 | By JoyYoung | cosmos-insight.com
β οΈ Investment Disclaimer: This article is for informational and educational purposes only. It does not constitute investment advice, a solicitation, or a recommendation to buy or sell any securities or financial instruments. All opinions expressed are the author’s own. Investing involves risk, including possible loss of principal. Please conduct your own due diligence and consult a qualified financial advisor before making any investment decisions.
I first came across this idea while tracking Korea Zinc’s quarterly results last year. The company had quietly acquired an electronic waste recycler, and the disclosure sat buried in a footnote. At the time it seemed like a minor diversification move. A few months later, China announced export controls on seven rare earth elements β and that footnote suddenly looked a lot more significant.
The connection most analysts miss is simple: inside every NVIDIA H100, H200, or Blackwell GPU sits a measurable quantity of gallium, germanium, terbium, and dysprosium. These are not theoretical impurities. They are precision-engineered components essential to how the chips function. And hyperscalers β Microsoft, Google, Amazon, Meta β are replacing entire server halls full of these chips every 3 to 5 years, not because the hardware is broken, but because next-generation silicon offers dramatically better performance per watt.
The result is a growing stockpile of mineral-rich e-waste, and a race β still early, still messy β to extract value from it before someone else does.
Why This Moment Is Different
π¨π³ China’s April 2025 Export Control Shock
In April 2025, Beijing tightened export controls on samarium, gadolinium, terbium, dysprosium, lutetium, scandium, and yttrium. These are precisely the elements embedded inside the GPUs and power conversion components that fill modern AI data centers. China controls roughly 90% of global rare earth production and refining β a supply concentration that Western governments have long viewed as a systemic risk, but one that felt distant until those export notices landed. Suddenly, urban mining shifted from an ESG talking point to a strategic imperative.
β‘ The AI Upgrade Treadmill Is Accelerating
NVIDIA’s H100 became the dominant training chip in 2023. By 2024, hyperscalers were already transitioning to H200. By 2025, Blackwell was the new benchmark. Each architecture jump compresses the usable lifespan of the previous generation not because it fails, but because the performance gap is too large to ignore. A data center operator running H100s is facing a compounding competitive disadvantage versus a rival running Blackwell. That pressure means waves of still-functional hardware hitting the decommission queue every few years β hardware whose mineral content has never been systematically recovered.
ποΈ The Scale Is Hard to Overstate
Microsoft, Google, Amazon, and Meta are each committing $50β100 billion annually in data center capital expenditure. Conservative projections suggest tens of thousands of tons of AI server hardware will be decommissioned every year by 2027. The rare earth content per individual GPU is measured in milligrams β but multiplied across millions of units across dozens of hyperscaler campuses, the aggregate mineral value becomes material. The urban mine already exists. The question is infrastructure and economics.
What Critical Minerals Are Inside a GPU?
Each of these elements plays a specific functional role β they are not incidental trace contaminants but engineered components that cannot be easily substituted.
| Mineral | Role in GPU / Server | China Supply Share | Strategic Risk |
|---|---|---|---|
| Gallium | Compound semiconductors, GaN power chips | ~80% | β β β β β |
| Germanium | Transistors, fibre optic transceivers | ~60% | β β β β β |
| Terbium | High-performance magnets, display phosphors | ~90% | β β β β β |
| Dysprosium | High-temperature permanent magnets | ~90% | β β β β β |
| Indium | ITO coatings, display interfaces | ~55% | β β β β β |
| Antimony | Flame retardants, PCB materials | ~50% | β β β β β |
| Gold / Palladium | PCB connectors, contact plating | Low | β β β ββ |
Korea Zinc: Why This Company Is Worth Watching
Korea Zinc (010130 Β· KOSPI) is not a startup chasing a trend. It is a 51-year-old non-ferrous metal smelter β the world’s largest by several measures β with established refining infrastructure, metallurgical expertise, and a balance sheet that generated a record 1.2 trillion won in operating profit in 2025. When its chairman publicly disclosed active negotiations with major US technology firms to recycle data center hardware, it carried a different weight than a press release from a junior mining company.
The strategic logic is clear. Korea Zinc already refines antimony, gallium, germanium, and other critical minerals from conventional ore sources. Transitioning part of that capacity to process e-waste input streams is a natural extension β not a pivot. The company spent two years in quiet R&D before Chairman Yun B. Choi made the first public comment in early 2026.
ποΈ The Tennessee Smelter: A $7.4 Billion Commitment
In December 2025, Korea Zinc announced plans to build a $7.4 billion smelting facility in Tennessee, backed by significant US government funding. The plant is designed to produce 540,000 metric tons of non-ferrous metals annually across 11 critical mineral categories β antimony, gallium, and germanium among them. The target operating margin is 17 to 19%, meaningfully above the company’s Korean refinery benchmark.
Locating in Tennessee is not accidental. It puts Korea Zinc inside the United States supply chain β eligible for IRA incentives, positioned for direct contracts with US defence and technology buyers, and insulated from the tariff and logistics risks that come with cross-Pacific sourcing.
π¬ Two Years of Quiet R&D Before Going Public
Chairman Choi’s exact words were: “We have been quietly researching technologies to extract rare earths over the past two years.” The company has since acquired an electronic waste recycling operator and a scrap metal trader to begin building the input supply chain. Technology partners have not been disclosed. But the pattern β internal R&D, supply chain acquisitions, then a major US facility announcement β suggests this is operational planning, not exploratory positioning.
Other Players in the Urban Mining Race
Korea Zinc is the furthest along among publicly visible participants, but it is not alone. Several players are approaching the same opportunity from different angles.
| Company | Approach | Status |
|---|---|---|
| Korea Zinc (010130 Β· KOSPI) | US smelter + direct talks with hyperscalers | Active 2026 |
| Western Digital (WDC Β· NASDAQ) | Rare earth extraction pilot from Microsoft server waste | Pilot 2025 |
| MP Materials (MP Β· NYSE) | Scaling domestic US rare earth refining capacity | Expanding |
| Critical Materials Recycling | Hydrometallurgical process with Western Digital partnership | R&D Phase |
| Lynas Rare Earths (LYSCF Β· OTC) | Largest non-Chinese producer; US processing plant | Operational |
The Real Barriers β And Why They Matter
Urban mining sounds straightforward until you look at the actual extraction economics. These are genuine obstacles, not minor friction.
| Challenge | Why It Is Hard | Severity |
|---|---|---|
| High Refining Costs | Separating trace rare earths from complex alloys requires energy-intensive hydrometallurgical processes that are expensive relative to virgin ore refining at current scale | Very High |
| Trace Concentration | Rare earths exist in milligrams per device. Collecting, sorting, and aggregating enough volume to justify dedicated refinery throughput is a significant logistics problem | High |
| Data Security Requirements | Hyperscalers cannot hand decommissioned servers to a third party without certified data destruction. This adds process steps and audit requirements that slow throughput | Medium-High |
| China Price Floor | If export controls are eased, Chinese processors could undercut recycled rare earth economics again. The business case depends partly on policy remaining firm | Medium |
| Regulatory Complexity | E-waste recycling rules differ across US states, EU regulations, and Asian markets. Cross-border shipment of hazardous electronic waste is heavily restricted | Medium |
Thematic Exposure: Stocks & ETFs to Research
β οΈ For informational research only. Not investment advice. All investments carry risk of loss.
The following represents a research framework for investors interested in this theme. Positions should be sized according to individual risk tolerance and portfolio objectives.
π°π· Korean Stocks (KRX)
| Company | Ticker | Thematic Link |
|---|---|---|
| Korea Zinc | 010130 Β· KOSPI | Direct β data center waste recycling + Tennessee smelter |
| Young Poong | 000670 Β· KOSPI | Korea Zinc major shareholder; indirect exposure |
| LS Materials | 294870 Β· KOSDAQ | Non-ferrous metals recycling infrastructure |
| Sungeel Hitech | 365340 Β· KOSDAQ | Battery recycling with critical mineral recovery |
π Global Stocks (US-listed)
| Company | Ticker | Thematic Link |
|---|---|---|
| MP Materials | MP Β· NYSE | Only producing US rare earth mine; domestic refining scale-up |
| Lynas Rare Earths | LYSCF Β· OTC | Largest non-Chinese rare earth producer globally |
| Energy Fuels | UUUU Β· NYSE | Uranium + rare earth co-processing |
| Western Digital | WDC Β· NASDAQ | Active pilot extracting rare earths from Microsoft server waste |
π°π· Korean ETFs (KRX-listed)
| ETF Name | Code | Theme |
|---|---|---|
| KODEX 2μ°¨μ μ§μ°μ | 305540 | Battery + critical minerals supply chain |
| TIGER μμ¬TOP10 | 396520 | Korean materials sector including strategic metals |
| KODEX νμν¨μ¨κ·Έλ¦°λ΄λ | 375760 | Circular economy; recycling-aligned green investment |
| HANARO Fnν΄λ¦°μλμ§ | 385560 | Clean energy materials and infrastructure |
| TIGER κΈλ‘λ²μμμμ°κΈ°μ | 448540 | Global resource producers including rare earths |
| KBSTAR κΈλ‘λ²ν΄λ¦°μλμ§ | 395160 | Clean energy + critical mineral producers |
| ARIRANG κΈλ‘λ²ν¬κ·κΈμ | 269540 | Global rare and specialty metals basket |
π Global ETFs (US-listed)
| ETF | Ticker | Focus |
|---|---|---|
| VanEck Rare Earth/Strategic Metals | REMX Β· NYSE | Most direct pure-play rare earth ETF available |
| iShares MSCI Global Metals & Mining | PICK Β· NYSE | Broad diversified metals and mining exposure |
| Sprott Critical Materials | SETM Β· NASDAQ | Critical minerals supply chain focused |
| Global X Lithium & Battery Tech | LIT Β· NYSE | Battery metals demand including rare earths |
| Invesco Clean Energy | PBW Β· NYSE | Clean energy materials and recycling value chain |
| First Trust Nasdaq Clean Edge | QCLN Β· NASDAQ | Clean energy technology supply chain |
| KraneShares Electric Vehicles & Future Mobility | KARS Β· NYSE | EV adoption driving structural rare earth demand |
My Read: What Makes This Interesting β and What Could Go Wrong
What I find genuinely compelling about this setup is the convergence of three independent forces arriving simultaneously. The China export controls were not triggered by urban mining β they were a geopolitical response to US semiconductor restrictions. The AI hardware upgrade cycle is not being driven by mineral economics β it is being driven by model competition. And Korea Zinc’s Tennessee investment is not primarily a recycling play β it is a response to US industrial policy and IRA incentives. Three unrelated dynamics are pointing at the same asset: e-waste mineral content.
The risk I watch most closely is the China policy variable. If Beijing decides to ease export controls as a negotiating concession in a broader trade deal, the economics of Western rare earth refining β urban mining included β deteriorate quickly. The business case for expensive hydrometallurgical processing only works if Chinese supply remains constrained and priced high.
Korea Zinc’s established refining infrastructure and 51-year track record differentiate it meaningfully from junior operators. But the Tennessee smelter is a 2027β2028 story at the earliest. Investors are pricing in execution that has not yet occurred. That gap between strategic logic and operational reality is where the risk lives.
π Five Things to Take Away
| 01 | AI hardware refresh cycles of 3β5 years are creating a predictable, growing pipeline of mineral-rich e-waste that has never been systematically recovered |
| 02 | China’s April 2025 export controls on seven rare earth elements transformed urban mining from an ESG initiative into a strategic supply chain priority |
| 03 | Korea Zinc brings 51 years of refining expertise, record profitability, and a $7.4B Tennessee commitment β this is not a speculative junior miner |
| 04 | High refining costs, trace mineral concentrations, and data security requirements are real barriers β the economics work only if rare earth prices stay elevated |
| 05 | The biggest risk is a China policy reversal β if export controls ease, the entire Western urban mining business case weakens materially |
β οΈ Full Investment Disclaimer: This article is published solely for informational and educational purposes. Nothing in this article constitutes financial, investment, legal, or tax advice. References to specific stocks, ETFs, or financial instruments are for research discussion only and do not represent recommendations to buy, hold, or sell any security. All investments carry risk, including the potential loss of principal. The author may or may not hold positions in securities mentioned. Always consult a licensed financial advisor before making investment decisions. cosmos-insight.com accepts no liability for any investment decisions made based on content published on this site.
urbanmining datacenterwaste rareearthminerals criticalminerals KoreaZinc GPUrecycling ewaste chinaexportcontrols gallium germanium terbium dysprosium antimony circulareconomy AIinfrastructure datacenters REMX MPMaterials WesternDigital greenmining sustainabletech investingthemes criticalmineralsinvesting commodityinvesting supplychain techstocks KOSPI010130 mininginvesting Lynas EnergyFuels rareearth
About the Author
