The Strategic Mineral Reserve: facility brief & development roadmap.

The problem. American factories need small amounts of special metals to build almost everything important — jets, chips, cars, medical devices. Most of those metals come from China, and sometimes China says "no more exports." When that happens, factories stop: Ford lost a week of production in May 2025 over one of these metals. And here is the surprise — no insurance pays for this. Business insurance only pays when something physically breaks, like a fire or a flood. A shipment that never arrives breaks nothing, so the policy pays nothing.

The idea. AeX buys a large pile of those metals now, while they can still be bought, and keeps it locked in American warehouses. Thirty to fifty big companies each pay a yearly fee. If the metals ever stop coming, paying members get to buy from the pile — their fair share, at a fair pre-agreed price — while everyone else scrambles. It works like a fire-station membership: nobody hopes for the fire, but everyone sleeps better knowing the truck is fueled and three minutes away.

Why companies would pay. Keeping your own private pile ties up money, takes expertise, and tells the world you're worried. Sharing one pile is cheaper — and the yearly fee is pocket change next to what a stopped factory burns per week.

Why we don't go broke. We borrowed money to buy the metal, so every year the membership fees must cover the loan payments and the warehouse bills, with profit left over. That's the whole business. Two ironclad rules keep it alive: we never sell the pile to bet on metal prices (that exact bet destroyed the world's tin reserve in 1985), and if years pass with no crisis, members get some money back — so they don't quit. The piles that died in history all died of boredom, not crisis.

How we fit next to the government's pile. The government is building a giant version of this (Project Vault) with cheap federal money — but a government warehouse has government rules. Everything it buys is public, funded by Congress, and picked apart by politics. So there are ready-to-use, specialized forms of these metals — and material already trading quietly on the open market — that the government simply won't hold, because a public buyer can't touch them without a political fight. A private company can buy that same material calmly and legally, out of the headlines, and keep exactly those hard-to-get forms on the shelf. So we're not racing the government's truck; we carry the specialized gear it won't. We complement Vault — and promise what it can't: privacy, a contract instead of a political decision, and speed.

The catch. The bank that lent us money technically has first claim on the metal, so we agree with the bank in advance, in writing, exactly how members can still take their metal in a crisis. Solving that on the day of the crisis would be too late.

Standard business interruption and contingent BI policies pay only on physical damage to property; supply loss from an export restriction is excluded — Swiss Re and Marsh both describe non-damage supply disruption as largely or entirely uninsured, and the Source Library's BI overview confirms the physical-damage requirement is structural to CBI, not an underwriting choice. On the public side, GAO found NDS shortfall materials rose from 37 to 99 (167%) between FY2019 and FY2023; CRS puts current inventory at ~38% of projected military shortfalls. Baskaran and CFR add scale: the U.S. is 100% import-dependent for 12 minerals and >50% for 29 more, while China controls 65–90% of processing for key materials. The loss event now has a name for pricing conversations: Ford's one-week production shutdown in May 2025 over rare-earth availability. The addressable void — quantity-assured, U.S.-domiciled physical cover against non-damage disruption — exists exactly as the SMR thesis claims.

Finding 1's gap is not hypothetical. Each event below halted otherwise-healthy production through an input that simply stopped arriving — an export action, an invasion, a rationed component — so no physical damage occurred at the manufacturer and standard BI/CBI paid nothing. These are the named loss events a premium conversation is priced against.

Sources — Ford: this brief (Finding 1). Nexperia: CNBC, Oct 30 2025. Ukraine: Bloomberg / S&P Global Mobility, Mar 2022. Semiconductors: AlixPartners, Sep 23 2021. Magnesium: Supply Chain Dive / European Aluminium, Oct 2021 — a threatened halt averted, not a realized multi-week shutdown; use as exposure evidence, not a paid loss.

SOFR is 3.66% (NY Fed print, July 1, 2026). Observed metal-sector asset-based facilities price around SOFR + 150–275bp (Alpha Metallurgical; Ferroglobe; ABF Journal norms) — a first-of-kind facility holding off-exchange minor metals should assume the wide end or beyond, so this model uses 6.7% all-in (SOFR + ~300bp) as a deliberately conservative planning rate. The IEA notes financing is the single largest carrying cost for high-value stockpiled metals, with specialty warehousing able to triple storage costs; storage, insurance, assay, and rotation plausibly add 1–1.5%, plus ~0.35% operations. Total cash carry ≈ 8–9%; adding a ~20% AeX margin puts the required premium pool near 10% of facility value annually.

Benchmark in three steps, in this order. Against commercial property programs ($0.30–1.50 per $100 of insured value; BI is added coverage) the SMR looks 7–30x more expensive per dollar of metal. Against observed parametric covers it is market-normal — Miami Beach paid $1.15m/yr for a $2–10m tiered limit (~11–57% rate-on-line; GFOA). Against a member's protected downstream revenue it is single-digit basis points ($3.3m on $5b of exposed revenue ≈ 7bp) for a peril BI and CBI exclude outright. Practical implication: 30–50 members at $1–3.5m realistically carry $500m–$1.25b; $2.5b needs the same membership at $5–8m each or a subsidized government anchor on the debt; $5b is not privately financeable from premiums alone. Lenders will advance perhaps 50–60% against off-exchange minor metals (no LME warrant, no forward curve), so the true capital stack needs 40–50% equity; the all-debt table overstates leverage but fairly approximates the total capital charge premiums must cover.

The International Tin Council defended a tin price with ~£900m of borrowed money secured on the metal itself; it collapsed in October 1985, halving tin prices and closing LME tin trading until 1989. The U.S. National Defense Stockpile peaked near $9.6b (1989) and was declared over 99% surplus by the late 1990s — liquidated by the calm, not the crisis. By contrast, OIL Insurance Ltd. (renamed Everen in 2022) — a Bermuda energy mutual formed in 1972 by 16 companies after commercial capacity collapsed — is still operating five decades later with 60+ members and ~$3.6b of shareholders' equity, because premiums monetize the quiet years and surplus returns to members. Japan's post-2010 response (JOGMEC stockpiles plus the Sojitz–Lynas offtake/equity deal) cut China rare-earth dependence from ~90% toward ~60%, showing the buyer-of-last-resort commitment mattered more than the warehouse. The SMR's premium-for-draw-rights design is, historically, the surviving architecture bolted onto the failing one — which is why the release rules (Finding 4) decide which gene dominates.

Peer-reviewed work on the closest analogue is sobering: Kilian & Zhou (2020) find SPR emergency releases moved real oil prices only modestly (~$2/barrel cumulatively for the 1990 Kuwait release), and Bai & Dahl (2018, Energy Policy) estimate the SPR's 1976–2014 real costs (~$219b) exceeded benefits (~$122b). Minor-metal markets are far thinner than oil — in March 2022 LME nickel nearly quadrupled in three trading days and ~$12b of trades were cancelled (FIA; OFR: $1.3b net P&L eliminated) — so a “volatility-capping” release either moves the price violently or is too small to matter. The structural problem is wrong-way risk: the same atoms are the lender's collateral and the member's claim, and collateral value spikes precisely when members want to draw. The Insurance folder now supplies the missing draw-mechanics parameters: 72-hour standard waiting periods (BI convention), tiered payouts (50/75/100% by severity), double-trigger structures, and 14–30-day payment standards (CCRIF pays within 14 days) — all of which port directly into the draw architecture in Section 04. Newbery & Stiglitz's classic result still governs: stabilization gains are mostly transfers, so the premium must be extracted from members' willingness to pay for continuity, not conjured from market smoothing.

The DoD–MP Materials transaction (July 10, 2025) — a 10-year $110/kg NdPr price floor, $400m convertible preferred, 100% magnet offtake — established that the U.S. government will directly backstop private stockpile economics. On February 2, 2026, EXIM's board approved a Direct Loan of up to $10b to launch “Project Vault” alongside nearly $2b of private capital. The new Vault record shows it is not just a subsidized warehouse: OEMs (Boeing, GE Vernova, Clarios, Western Digital, now GM and Google) pay commitment fees for annual pro-rata withdrawal-and-replenishment rights, gain full access on disruption triggers, and hold anonymized tradeable entitlements — on 10-year terms across ~60 minerals, supplied by Hartree, Mercuria, and Traxys. The premium-for-draw-rights model is being proven with public money, which validates the concept and captures the most obvious anchor demand. The strategic conclusion is complementarity, not competition — and the differentiator is commercial, not a claim to buy what Vault legally cannot. Vault has no explicit foreign-entity-of-concern sourcing restriction and may itself source Chinese-processed material where no alternative capacity exists (a gap BPC and CSIS both flag), so the private edge is not privileged access to forbidden supply. It is the transparency gap: a federal reserve's purchases are appropriated, disclosable, and politically owned, which leaves a large volume of openly-traded, already-exported refined and processed material — specific off-list grades, purities, and forms outside Vault's 60-mineral program — that a public buyer will not hold at the political cost, but a privately capitalized, commercially structured buyer can acquire quietly on world markets, within a hard compliance perimeter (origin-transparent, sanctions-screened, customs-clean). That confidentiality and off-list focus is the lane. Vault's weaknesses are on the record — no Inspector General during EXIM's largest-ever commitment, conflict-of-interest questions, no sunset provisions, a processing loophole letting released material flow to China-linked processors, and the agricultural-subsidy precedent of permanent political entrenchment — and each maps to a private differentiator: clean governance, contractual (non-political) draw certainty, confidentiality of member vulnerability, allied-processing conditionality, off-list refined and processed forms, and formula price protection on the draw. A private SMR that merely replicates Vault is dead on arrival; one that holds the off-list forms and openly-traded material a public program will not carry, with the commercial confidentiality members value, is complementary to it — filling the void the federal program leaves open rather than competing for the same shelf. The DPAS tail risk stands: in a true crisis, rated orders can supersede private contracts.

1. HoldCo. Delaware LLC (AeX SMR Holdings) — governance, member agreements, IP; Delaware per the Latham incorporation comparison in the Source Library.

2. Metal OwnerCo(s). Delaware series LLC per mineral family holding title. Store in U.S. Foreign-Trade Zones where sourcing is imported (defers customs duties and Section 301 tariffs until withdrawal; FTZ status plus state freeport exemptions can eliminate ad valorem inventory tax) and in no-sales-tax states (e.g., Delaware) otherwise. Title moves by endorsement of negotiable warehouse receipts under UCC Article 7 — metal never physically moves on a draw, minimizing transfer/sales-tax friction and enabling in-place settlement.

3. CapacityCo — the two-contract stack. (a) A Reserve Capacity Agreement: the annual standby fee (legally a fee for contractual capacity, not indemnity); (b) a pre-priced, physically settled forward/call on the member's pro-rata entitlement, exercisable only on defined triggers. Physically settled forwards in nonfinancial commodities are excluded from the CEA swap definition (7 U.S.C. §1a(47)(B)(ii)); the commercial-end-user framing follows CFTC guidance (77 FR 48208) rather than statute — it needs an opinion, not an assumption.

4. Insurance wrapper (now a co-equal track, not a fallback). For members electing insurance accounting: a Vermont/South Carolina captive, or a Bermuda Class 3 vehicle with an IRC §953(d) election (avoiding the 4% federal excise tax on premiums to foreign insurers). Per Swiss Re and Wharton, the parametric cover qualifies as insurance where insurable interest is documented (the member's declared supply-chain exposure) and proof-of-loss is embedded in the trigger (third-party supply data). Bermuda's 15% CIT (2025) applies only to MNE groups with ≥€750m revenue. For a facility with U.S.-domiciled inventory and U.S. members, cross-border optimization mostly collapses back onshore — a feature: it simplifies the national-security narrative.

5. Offshore trading arm (thin, deferred). A DMCC (Dubai) entity only when non-U.S. sourcing volume justifies it. Income tied to U.S. inventory and U.S. members is U.S. effectively connected income regardless of booking; keep this layer thin and transfer-pricing clean.

6. Credit stack. Borrowing-base revolver at OwnerCo level; intercreditor with pre-negotiated automatic lien release against cash collateral; premium waterfall: debt service → carry/ops → replenishment reserve → AeX margin — published to members.

1. Dual triggers, physically settled. A draw window opens on either (a) an objective event trigger — export-control action by a named jurisdiction on a reserve mineral, or force majeure/allocation notices from two or more qualified suppliers — or (b) a sustained physical-availability trigger (U.S. delivered premium over benchmark exceeding a set threshold for a set number of days). Physical delivery, not cash settlement: basis risk in a cash parametric payout is exactly what collapses demand (Clarke 2016); with physical settlement the trigger data doubles as the regulator's proof-of-loss (Wharton test).

2. Verification window and delivery standard. 72-hour to 5-day verification after trigger (BI waiting-period convention, buyable-down); 14–30-day delivery standard (CCRIF pays in 14 days — match it; the speed differential vs indemnity claims is half the product).

3. Tiered, capped, pro-rata draws. Tiered release 50/75/100% by disruption severity (parametric convention); entitlements pro-rata to premium share; monthly velocity caps (e.g., ≤25% of entitlement) so no single event empties the reserve; internal secondary market for entitlements — mirroring Vault's anonymized tradeable entitlements — reveals true willingness-to-pay and prices renewals.

4. Formula pricing on draw. Trailing 12-month benchmark, not spot. Quantity certainty plus squeeze-price protection is the actual insurance content.

5. Never defend a price. Releases go only to members against draw rights; buyer-of-last-resort purchases run through a separately funded, capped, sunset sidecar — never through the members' reserve (ITC lesson; Breakthrough principle). Allied-processing conditionality on any released material closes the loophole BPC flagged in Vault.

6. Replenishment & governance. Post-draw: temporary premium surcharge, right of first refusal on member excess, defined rebuild window. Independent release committee applies triggers mechanically; no discretionary sales; annual third-party assay and existence audit (Qingdao standard); DPAS acknowledgment clause in every member contract.