Project Vault and the critical minerals paradox
Recent policy discussion around Project Vault — described in the briefing materials as a US-backed effort announced in February 2026 to assemble a critical minerals stockpile with private capital and Export-Import Bank financing — points to a hard reality many executives already sense. The United States cannot unwind its critical-mineral exposure on a political timetable. China still dominates large parts of the refining and processing chain behind rare earths, gallium, germanium, cobalt, and graphite that sit inside semiconductors, batteries, cooling systems, and advanced electronics. To build a stockpile quickly enough to matter, early purchases may still need to come from Chinese-linked supply chains.
For organisations that depend on AI infrastructure, data-centre expansion, battery backup, or specialised manufacturing equipment, that reframes minerals as an operating risk rather than a geopolitical headline. The near-term goal is continuity. The real cost of a constraint is rarely the material price; it is the delay. A GPU cluster that ships six months late can derail a roadmap, and a battery-backup project that misses a commissioning window can postpone a data-centre expansion.
Where the exposure shows up
- Rare earth magnets in motors, cooling systems, robotics, and manufacturing tools
- Lithium, cobalt, graphite, and nickel in batteries, backup power, and energy storage
- Gallium and germanium across semiconductor equipment, power electronics, optics, and secure communications
- Indirect dependencies that affect server lead times, networking availability, and rollout schedules
Why reducing China dependence still pulls through China in the short term
The explanation begins with process flow, not policy language. The paradox is midstream physics. Germanium is commonly recovered as a by-product from zinc-processing residues and fly ash; gallium is often recovered from Bayer liquor in alumina refining. Neither behaves like a primary mine product that moves cleanly from ore body to finished inventory. Reaching semiconductor- or optics-grade output requires leaching, chlorination, solvent extraction or ion exchange, precipitation, distillation, and in some cases zone refining to reach 5N purity (99.999%) and above. That sequence — together with clustered engineering know-how, reagent supply, effluent treatment, and customer qualification history — is why China retains leverage even where mining shifts to Africa, Australia, or North America.
Rare-earth magnets follow the same logic. Mine output or mixed carbonate is upstream success, but NdFeB magnet availability depends on solvent extraction of NdPr oxides, conversion to metal, strip casting, jet milling, alignment, sintering, machining, coating, and final integration. The Kipushi zinc-concentrate deposit illustrates the trap: contained germanium and gallium sitting at ppm levels within a concentrate still need smelting, residue recovery, purification, and qualification before they can support fibre optics or compound semiconductors. A stockpile of concentrate is geological contingency; a stockpile of qualified 5N gallium is operational continuity.
Hardware exposure: semiconductors, data centres, batteries, and defence-adjacent systems
The semiconductor link is often misunderstood. Gallium exposure in AI infrastructure rarely means the main accelerator die is a gallium compound; it more often sits in adjacent layers — power electronics, radio-frequency components, optoelectronics, and specialised compound-semiconductor devices. Germanium matters in infrared optics, fibre and photonics, secure communications, and sensing that overlaps government workloads. Once export controls tighten, the disruption does not stay in a narrow defence silo; it leaks into cloud, telecom, and advanced electronics.
The data-centre impact is most tangible in liquid cooling and electromechanical balance-of-plant systems. Neodymium magnets used in liquid-cooling assemblies, including those associated with suppliers such as Vertiv and Chilldyne, still draw on a heavily China-centred supply base. That surprises many infrastructure teams, because rare-earth exposure is usually framed around EVs and wind. Yet the same NdFeB chemistry sits in the pumps, motors, fans, and motion-control components that populate dense compute environments, and diversification delays can convert into component lead times of three to six months. Cobalt is more mixed — lithium iron phosphate reduces it in many stationary systems — but DRC cobalt offtakes still route through Asian processing before reaching battery-grade form. Tungsten sits quieter, in tooling, sputtering targets, shielding, and high-temperature contacts, where substitution is limited and disruption is disproportionate.
- Gallium: compound semiconductors, power electronics, RF devices, and high-performance networking.
- Germanium: infrared optics, fibre and photonics, secure communications, and sensing.
- NdPr / NdFeB: permanent magnets in cooling, pumping, fans, actuators, and high-efficiency assemblies.
- Cobalt: battery precursor chains, selected stationary chemistries, and superalloys.
- Tungsten: tooling, shielding, sputtering targets, and high-temperature contacts.
What form of stockpile actually matters
The decisive question is not only which country supplies the material, but which stage of the value chain gets buffered. Four forms appear in practice, each with a different resilience profile.
| Stockpile form | Continuity value | Main weakness |
|---|---|---|
| Raw ore / concentrate | Broadest geological exposure, sometimes securable from non-Chinese mines earliest | Exposed to smelter availability, recovery chemistry, tolling slots, and quality variability — deferred feedstock, not immediate buffer |
| Intermediate chemicals (oxides, hydroxides, salts) | Closer to manufacturable value, easier to assay and warehouse | Still requires conversion capacity; ppm impurity tolerances limit substitution |
| Refined metal / battery-grade salt | Strong continuity — the hardest purification stage is already complete | Shelf-life, packaging compatibility, and requalification rules still apply |
| Finished components | Shortest path to deployment continuity | Highest obsolescence risk against changing form factors and qualification |
A country can report impressive tonnage and still fail to protect end-use manufacturing if the inventory sits too far upstream from qualified hardware demand. The form of stockpile determines the form of resilience.
Implementation realities: traceability, compliance, environment, and logistics
The operational burden sits in four places at once. Traceability has moved beyond mine origin to the last transformative step: refining jurisdiction, toll-processing relationships, and whether transshipment masks actual processing exposure. A non-Chinese certificate of origin does not settle the question if the critical purity upgrade occurred in China. Compliance reinforces the shift, as export controls, customs enforcement, and forced-labour screening push buyers to document legal provenance alongside chemistry — a shipment that meets specification but fails traceability can be unusable. Environmental and safety burdens are routinely underestimated: gallium and germanium recovery involves corrosive acids and hazardous waste, and rare-earth separation produces significant effluent. Logistics complete the picture; corridors such as the Lobito route can support diversification, but concentrates, hydroxides, and refined metals each travel, insure, and qualify differently once they arrive.
The honest trade-off: resilience improves before sovereignty does
Project Vault exposes the limits of American mineral sovereignty. Even with political urgency and private participation, domestic and allied processing will not scale overnight, so early stockpile builds may still depend on Chinese inputs while the broader strategy aims to reduce future dependence. That is sequencing, not failure. Step one is a buffer against disruption; step two uses that buffer to give non-Chinese refiners enough committed demand to invest. Short-term acquisition costs may rise when paying for guaranteed access, qualifying new suppliers, or carrying strategic inventory, but total cost of ownership often improves as missed deadlines, expedited logistics, and interrupted programmes decline. The hardest part is not the premium; it is acting before a shortage is visible to everyone and the market has already repriced the risk.
A disciplined path forward
- Map exposure (30–60 days): identify where critical minerals affect servers, batteries, networking, cooling, and vendor lead times.
- Prioritise high-impact categories (60–90 days): separate business-critical components from easily replaceable items.
- Pilot resilience measures (90–180 days): test dual sourcing, reserved inventory, and longer-term commitments — and decide which value-chain form to buffer.
- Integrate sourcing risk into planning (6–12 months): connect procurement, ERP, infrastructure planning, and finance.
- Scale diversification (12–24 months): align future buying with domestic and allied supply where volume and economics support it.
The organisations that benefit most treat mineral exposure the way they treat power availability or cloud concentration: a board-level dependency that deserves active management. Project Vault is important precisely because it is imperfect — it does not solve US–China mineral dependency, it reveals it, and that clarity is useful.
Note on Procyon methodology. Procyon crosses policy-text monitoring, including export-control signals from bodies such as BIS and, where relevant, MOFCOM, with market and logistics indicators, then tests that evidence against the technical specifications of end uses — purity class, qualification status, component architecture, and substitution limits in semiconductors, data centres, batteries, and defence-adjacent systems.
Selected sources referenced in the briefing materials
- U.S. Geological Survey, Mineral Commodity Summaries 2026 and Germanium Statistics.
- U.S. Bureau of Industry and Security, gallium and germanium export-control materials.
- Ivanhoe Mines, Kipushi technical materials.
- Semiconductor Industry Association supply-chain materials.
- Vertiv and related data-centre minerals references.
- Cobalt Institute logistics references.
- EXIM Project Vault terms; IEA Critical Minerals Market Review 2026.