AI Infrastructure Emerges as Third Structural Driver of Rare Earth Demand Alongside Defence and Electrification, Sprott Warns of China's 94% Magnet Chokehold

Introduction

For the better part of a decade, the rare earth demand narrative has rested on two structural pillars: the energy transition, anchored in electric vehicle motors and wind turbine generators, and defence modernisation, anchored in precision-guided munitions, fifth-generation fighter jets, and advanced sonar and radar systems. In a July 17, 2026 analysis, Sprott Asset Management added a third pillar that few demand models had incorporated with any rigour: artificial intelligence infrastructure.

The five largest hyperscale data-centre operators collectively committed approximately $400 billion in capital expenditure during 2025 alone, a 72% increase on the prior year according to Bloomberg data cited by Sprott Managing Partner and Head of ETFs Steve Schoffstall. Rare earth magnets are embedded throughout these facilities in cooling motors, storage drives, semiconductor fabrication equipment, and communications hardware. Cooling systems alone account for roughly 20% of a data centre's total electricity costs, and high-performance permanent magnets are central to keeping those systems running efficiently.

The International Energy Agency, whose April 2026 report on rare earth supply chains was released to inform discussions under France's G7 Presidency, projects that AI-driven data centres will account for 3% of global magnet rare earth element consumption by 2030. That figure may appear modest in isolation, but it represents a structurally new demand category layered on top of already-growing EV, wind, and defence volumes at a moment when supply chains remain dangerously concentrated in a single country.

China accounts for approximately 94% of global permanent magnet production. The United States mines roughly 13% of global rare earth output and still depends on China for around two-thirds of its requirements. Against that backdrop, Schoffstall's characterisation of the current environment as a "new level of resilient demand" is simultaneously an investment thesis and a geopolitical warning.

The Three-Pillar Demand Architecture: AI, Defence, and Electrification

Sprott's analysis frames the current demand environment as the convergence of three independently powerful forces, each of which would be significant on its own. Taken together, they create what Schoffstall describes as a demand structure less sensitive to traditional economic cycles and more anchored in long-term structural transformation. Total investment across defence, the energy transition, and AI has risen 78% since 2020, according to Sprott's April 2026 investor presentation.

The electrification pillar remains the largest single driver by volume. High-performance neodymium-iron-boron (NdFeB) magnets, composed primarily of neodymium and praseodymium with dysprosium and terbium added to maintain performance at elevated temperatures, underpin every commercially viable EV drivetrain and direct-drive wind turbine generator. Demand for magnet rare earths has doubled since 2015 and is projected by the IEA to grow by more than 30% by 2030, with overall demand expected to exceed 120,000 tonnes by that date and reach 175,000 tonnes by 2050, a growth of over 90% under existing policy trajectories.

Defence is the second pillar, and it is accelerating. Global military spending has climbed to approximately $2.6 trillion annually, all 32 NATO member nations met their 2% of GDP spending guideline for the first time in 2025, and the alliance has set a 5% of GDP target for 2035. The platform-level numbers are striking in their specificity: an F-35 Lightning II contains approximately 417 kilograms of rare earth materials spread across more than 20 individual actuator and motor assemblies, an Arleigh Burke-class destroyer requires 5,200 pounds, and a Virginia-class submarine needs 9,200 pounds. Abigail Hunter, Executive Director of the Center for Critical Minerals Strategy at SAFE, put the dependency plainly: "If it moves, sees, communicates in today's military, we usually say there's probably a rare earth in it."

The AI pillar is the newest and, in growth-rate terms, the fastest-moving. Annual data centre investment grew from approximately $100 billion in 2015 to $500 billion in 2024. Updated IEA projections see electricity consumption from data centres roughly doubling from 485 terawatt-hours in 2025 to 950 terawatt-hours in 2030, with AI-focused facilities tripling their consumption over the same period. When polled at a mid-April 2026 VettaFi/Sprott webcast on the primary driver of rare earth demand in client portfolios over the next three to five years, 42% of investor respondents selected AI, placing it ahead of all other options.

China's 94% Chokehold: Processing Power as Geopolitical Leverage

China's dominance over global rare earth supply chains is not principally a story about geology. The country holds only about 35% of global reserves, a large but not insurmountable share. What makes the situation strategically acute is China's near-total control over the processing stages that sit between mine output and finished magnets. Beijing controls roughly 90% of global rare earth processing, and its share of sintered NdFeB magnet production has risen from approximately 50% two decades ago to nearly 94% today. The USGS places China's share of separated heavy rare earth output, the fraction most critical for heat-resistant magnets used in defence and EVs, at 98 to 99%.

This concentration is not the product of market forces operating freely. It reflects decades of strategic investment in processing infrastructure, tolerance for environmental costs that other jurisdictions rejected, and disciplined use of scale to drive out competitors. IEA Executive Director Fatih Birol offered a concise diagnosis: "Rare earth elements are indispensable to many of the technologies shaping the Age of Electricity and our increasingly digitalised economies, yet their supply chains remain among the most concentrated of all critical minerals." His prescription was equally direct: "If we leave it to the market forces, we can never fix this."

The practical consequence of this concentration became impossible to ignore beginning in April 2025, when Beijing launched a strict licensing regime targeting rare earth exports to sectors feeding defence and high-technology industries. Chinese magnet exports declined by an estimated 75% on a temporary basis, sending shockwaves through supply chains in Europe, Japan, and North America. The controls were partially eased in July 2025 following US pressure but were expanded in October 2025 to cover internationally made products containing Chinese-sourced rare earths, a provision subsequently suspended for one year though the IEA notes that "the vulnerabilities remain."

Escalation continued into 2026. In June of this year, China added ten US entities to its export control list, including rare earth miners MP Materials and USA Rare Earth, and prohibited third parties anywhere in the world from transferring dual-use items of Chinese origin to those entities. The move represented a qualitative shift: controls that had previously targeted material flows were now targeting specific corporate actors in Western supply chain efforts. The weight of evidence, as one analysis framed it, suggests that Chinese rare earth export controls should now be understood as a permanent structural feature of global trade rather than a temporary negotiating instrument.

Schoffstall noted that rare earths processed outside China are now commanding premiums of four to six times domestic Chinese prices. As of March 10, 2026, neodymium oxide was trading at $113.05 per kilogram domestically in China but at $184 per kilogram FOB China, a 63% spread reflecting supply chain security premiums. An S&P Global analyst captured the buyer psychology: "Pricing premiums are most evident where performance, qualification and continuity matter for magnet materials like neodymium-praseodymium, because buyers aren't just purchasing a chemical, they're buying assured supply over long time frames."

Price Movements and the Supply Deficit Signal

The pricing data for 2026 is among the clearest signals yet that structural supply-demand misalignment has moved from theoretical concern to market reality. NdPr alloy, the benchmark input for NdFeB magnet production, began 2026 at approximately $53 per kilogram and reached $126.16 per kilogram on April 1, 2026, a gain of 138% in three months. By July 2026, the benchmark had extended further, rising 21.4% month-on-month to $133.02 per kilogram, clearing the prior March peak and setting a new 2026 high. Neodymium metal rose 19.6% to $145.88 per kilogram and praseodymium metal rose 19.5% to $149.19 per kilogram over the same period.

The heavy rare earths critical for high-temperature magnet performance told a parallel story. Terbium posted its largest single-month gain since 2023 in April 2026, surging 20.7% to $970.18 per kilogram. By July 2026, terbium's FOB China price had reached $1,483 per kilogram. These are not commodity price fluctuations driven by inventory cycles; they reflect a second consecutive year of supply deficit against EV and wind turbine demand, compounded by precautionary stockpiling from buyers who saw the April 2025 export control episode and drew their own conclusions about supply reliability.

Building on my coverage of Lynas producing the first terbium oxide separated outside China in July 2026, the pricing context explains why that milestone commanded such immediate strategic significance. When terbium trades above $1,400 per kilogram and Chinese licensing approval rates for European firms have fallen below 25%, the commercial and security case for non-China separated heavy rare earth supply becomes self-reinforcing.

The broader price trajectory reflects what the IEA describes as an estimated $6.5 trillion per year of downstream production outside China at risk across the automotive, high-technology, defence, and energy sectors. That figure encompasses the full value of manufacturing that depends on reliable rare earth magnet supply, and it provides a rough sense of the economic stakes attached to Beijing's ability to constrict or redirect flows at will.

Western Policy Responses and the Investment Gap

Government responses have accelerated sharply, but the gap between policy ambition and actual supply chain capability remains large. Public finance commitments in advanced economies reached approximately $65 billion in 2025, more than four times the 2023 level, though the IEA notes a considerable gap between commitments and disbursements.

The most structurally significant US transaction to date is the Department of Defense's February 2026 commitment of $550 million to MP Materials, accompanied by a ten-year offtake agreement for 7,000 metric tons of magnet material per year at a price floor of $110 per kilogram and a pathway for the DoD to become the company's largest shareholder. The arrangement funds a new facility, designated 10X, designed to reach 10,000 metric tons of annual magnet production, which would address all of the DoD's rare earth magnet requirements. MP Materials is already commissioning a 1,000 metric ton magnet manufacturing plant in Texas. Together, the Texas facility and 10X would give the United States a vertically integrated rare earths-to-magnets supply chain for the first time in decades. Apple separately agreed to $200 million in prepayments for a long-term magnet supply agreement with MP Materials signed in July 2025, a commercial signal that technology sector buyers are no longer willing to treat Chinese magnet supply as a given.

Energy Fuels received a conditional commitment for up to $725 million in financing from the US Office of Strategic Capital to expand its White Mesa Mill in Utah and develop a rare earth metals and alloys manufacturing facility. The US also launched Project Vault in February 2026, a $12 billion public-private partnership to create a strategic reserve of critical minerals including rare earths. On the multilateral front, 55 nations participated in the inaugural Critical Minerals Ministerial convened in Washington on February 4, 2026, establishing a framework for a preferential trade zone covering strategic materials.

The European Union has layered the Critical Raw Materials Act of 2024 with the RESourceEU Action Plan in 2026 to push diversification. Europe is particularly well positioned in one supplementary supply lever: recycling. The IEA projects that Europe could generate half of global magnet scrap from wind turbines and a quarter from EVs by 2030, a secondary supply stream capable of reducing primary supply requirements by up to 35% by 2050 in an optimistic scenario. My earlier coverage of DEScycle's deep eutectic solvent e-waste plant at Teesside and the Lithios electrochemical lithium extraction milestone in July 2026 traced the technological frontier of this recycling buildout, and the rare earth dimension of that transition is following a similar trajectory.

Yet the IEA's own accounting makes clear that the investment response, however historically unprecedented, remains insufficient relative to need. Meeting demand for magnet rare earths outside China requires approximately $60 billion of investment over the next decade, with refining accounting for nearly 50% of that total and magnet manufacturing representing around 33%. By 2035, existing and announced capacities are expected to cover only about half of mining requirements, a quarter of refining needs, and less than a fifth of magnet demand outside China. The pipeline of downstream magnet projects is particularly thin compared to upstream mining developments, a structural imbalance that mirrors the chokepoint China has exploited so effectively.

The 16-Year Clock and the AI Timeline Problem

The convergence of the AI demand vector with rare earth supply chain timelines creates a specific and uncomfortable arithmetic problem. Schoffstall was direct on the point: "When you look at mining of new metals, it takes about 16 years on average to go from a new discovery to production." The IEA's Fatih Birol framed the same constraint in even starker terms, saying the gap between China and the rest of the world in critical minerals processing is "at least eight years."

The IEA forecasts that AI data centres will account for 3% of magnet rare earth consumption by 2030, roughly four years from the date of this analysis. That 3% figure, layered on top of EV, wind, and defence demand that is itself growing by more than 30% over the same period, lands well before most greenfield mining projects currently in development could reach commercial production. The implication is that the near-term supply response must come primarily from brownfield expansions at existing facilities, non-Chinese processing upgrades, strategic stockpiling, and recycling rather than from newly discovered deposits. It is precisely this logic that explains the policy emphasis on assets like MP Materials' Mountain Pass, Lynas's Kuantan heavy rare earth separation facility, USA Rare Earth's hydrometallurgical demonstration plant in Colorado, and the Carester-Malaco joint venture in Malaysia's Perak state, all of which are attempting to compress the timeline by working within existing infrastructure rather than building from scratch.

Schoffstall's observation that market forces alone are unlikely to rebuild supply chains after decades of Chinese dominance is not a counsel of despair so much as a description of the coordination problem governments are now attempting to solve with direct investment, offtake agreements, stockpiling mandates, and multilateral frameworks. The question is whether the policy response, for all its historical scale, is moving at a pace commensurate with the convergence of three structural demand drivers in a sector where the dominant supplier has demonstrated both the willingness and the capability to use supply as a geopolitical instrument.

Conclusion: A Strategic Inflection Point with an Uncertain Resolution

The Sprott analysis of July 17, 2026, and the IEA data it draws upon, represent the clearest articulation yet of how profoundly the rare earth demand landscape has shifted in a short period. What was once a market driven primarily by specialty industrial applications and niche defence procurement has become a central node in the geopolitical contest over technology and economic security. The three pillars of demand, AI infrastructure, defence modernisation, and the energy transition, are each growing independently and reinforcing each other in ways that compress the tolerance for supply disruption.

China's 94% share of global permanent magnet production is not merely a market statistic. It represents a structural vulnerability that becomes more acute as each new demand vector materialises. When the five largest hyperscale operators are committing $400 billion in annual capex and AI data centre electricity consumption is on course to triple by 2030, the 3% of magnet rare earth demand that the IEA attributes to data centres by that date is not a ceiling; it is a floor for a demand category that did not exist in rare earth forecasting models a decade ago.

The pricing data reinforces the urgency. NdPr alloy's rise from approximately $53 per kilogram at the start of 2026 to $133 per kilogram by July represents not just a commodity price cycle but the market's attempt to signal that supply-demand equilibrium has broken down in ways that cannot be resolved quickly. Terbium at $1,483 per kilogram FOB China sends an equally clear message about the scarcity premium attached to heavy rare earth separation capacity located anywhere outside Beijing's direct influence.

The $60 billion investment requirement identified by the IEA, the 16-year mine development timeline cited by Schoffstall, and the IEA's projection that non-Chinese capacity will cover less than a fifth of magnet demand by 2035 together describe a gap that no single company, government programme, or technological breakthrough can close on its own. What they can collectively do is reduce dependence to a level where the next round of Chinese export controls, whatever form it takes, does not translate immediately into cascading production delays across the defence, technology, and clean-energy sectors simultaneously. That, at the current moment, appears to be the realistic near-term ambition for the Western supply chain response.