The Narrow Gate

Ask most people where the defence-industrial risk in U.S.–China relations sits, and they will point to Taiwan, semiconductors, or the South China Sea. Few will point to a small basket of grey, unglamorous metals with names most readers cannot spell: dysprosium, gallium, germanium, antimony, terbium. Yet a growing body of Pentagon and allied analysis suggests this basket, not the fleet or the fab, is where the next supply shock is most likely to originate.

The instinct is to treat this as a resource-scarcity story: whoever has the ore wins. That instinct is misleading, and acting on it will misallocate capital. These elements are not geologically scarce. What is scarce is the capacity to turn ore into usable metal, alloy and finished magnet, a capability one country built patiently over three decades while everyone else treated it as a low-margin, high-pollution business not worth keeping. The result is a form of leverage that has little to do with what is in the ground and everything to do with who owns the industrial process in between.

Start with the concentration numbers, because they are more extreme than most investors appreciate. One country accounts for something close to six in ten tonnes of global rare-earth separation and refining capacity, and around nine in ten finished permanent magnets, despite holding well under half of known reserves. That gap between share-of-reserves and share-of-output is the whole story. It did not happen by accident; it happened because one government treated midstream processing as a strategic industry to be built and subsidised for decades, while others treated it as a commodity to be sourced wherever it was cheapest.

The consequences for Western militaries are now well documented in open-source defence research. An annual assessment of the U.S. defence-industrial base, drawing on five years of Pentagon contracting data, found that firms based in the country in question made up 9.3% of so-called Tier-1 subcontractors across nine essential military capability areas as of 2024, rising above 11% in missile defence specifically. A related piece of that same research identified roughly 80,000 individual U.S. weapon components with a dependency on one of five controlled minerals, and concluded that on the order of three-quarters of American weapons systems carry some exposure to a disruption in their supply.

This is not primarily a story about mines. It is a story about the unglamorous middle of the supply chain: separation trains, alloying furnaces, sintering lines, metallurgical know-how, calibrated technical standards, and a trained workforce that takes years to build and cannot be conjured by a subsidy cheque alone. Reproducing that ecosystem outside the country that built it is achievable, but it is a matter of a decade of patient industrial effort, not a matter of finding a new deposit.

Anyone tracking this over the last three years has watched a deliberate, escalating campaign of licensing and export restriction. Gallium and germanium moved onto a licensing regime in 2023. Antimony followed in 2024, alongside an outright prohibition on gallium, germanium and antimony shipments specifically to American buyers. In the spring of 2025, seven medium and heavy rare earths, including the specific elements needed for magnets that must hold their properties at high operating temperatures, were added to the control list. Then, in October 2025, came the most sweeping step yet: an extraterritorial rule asserting licensing authority over any product anywhere in the world containing even a small fraction of content sourced or processed through the controlling country’s supply chain.

What happened next is the detail that gets underplayed in most commentary on this subject, and it matters more than anything that preceded it. Barely a month after the October escalation, the two governments reached an accommodation: the sweeping new measures were suspended for a year, and general licenses were issued covering the core group of restricted minerals for end users in the affected country. Flows have not fully normalised even so; magnet shipments remain meaningfully below where they stood before the restrictions began. But the point stands: the most aggressive posture was walked back within weeks of being announced, not because of a change of heart, but because a policy this powerful invites the very diversification it is meant to prevent if used too aggressively for too long. A supplier with this kind of leverage has every incentive to apply it in short, calibrated bursts rather than as a permanent blockade, easing off before the other side’s alternative investments become irreversible.

That single fact should reorder how any allocator thinks about this theme. The restrictions that make Western processing capacity valuable right now are exactly the restrictions the supplier holding the leverage has good reason to relax the moment they have extracted what they wanted from applying them. Keep that in view; it is the fulcrum on which any valuation of this sector has to turn.

Concentrated, weaponisable supply would matter less if demand were flat. It is not. Global defence outlays reached $2.887 trillion in 2025, an eleventh straight annual increase and, measured as a share of world output, the highest level in more than fifteen years. The composition of that growth is telling: outlays in the country that spends the most actually declined year over year, while European spending rose 14% and spending across Asia and Oceania rose over 8%. Alliance members have now committed, at least on paper, to a defence-spending target that would represent one of the largest peacetime rearmament commitments in modern history if honoured. A major European bloc has separately outlined plans to mobilise several hundred billion euros for defence investment by decade’s end.

Then a real conflict supplied the proof of concept nobody wanted. Over roughly six weeks earlier this year, American forces expended more than a thousand long-range cruise missiles in a single campaign against Iran, consuming close to a third of the estimated pre-conflict inventory. Independent defence-budget analysts now estimate it will take three years or more to rebuild stocks of several of the interceptor and strike-missile families used most heavily in that campaign, even with a proposed defence budget for the coming fiscal year that is the largest ever requested.

Here is where the metals and the war meet directly. None of those interceptors or cruise missiles function without permanent magnets built from the same heavy rare earths placed under export control the previous year. A magazine drawn down by a third in forty days has to be refilled through a supply chain whose critical inputs run, in large part, through the same government now deciding whether or not to keep those inputs flowing. Defence planners have effectively discovered that their capacity to sustain a real, sustained fight is gated by an export-licensing decision made in a ministry thousands of miles away.

For capital allocators, the implication is straightforward: this is now a large and structurally price-insensitive pool of demand for a narrow group of processed inputs. When an interceptor costs many millions of dollars and the magnet inside it costs a rounding error by comparison, the buyer will not blink at metal prices; the buyer cares only about whether the material is available at all. That is an unusual demand profile, and it is the foundation the rest of this thesis rests on.

The most interesting part of this story is not the threat itself but the policy response to it, because the response amounts to a rediscovery of state-directed credit allocation, a tool the West spent the better part of half a century insisting it had outgrown.

In mid-2026, the U.S. defence department’s capital-markets arm launched a new financing vehicle designed to funnel government-backed loans through private fund managers, who then blend that public capital with private money and deploy it across the mineral supply chain from extraction through advanced materials. Estimates of the eventual scale run as high as $100 billion. It builds on legislation passed the previous year that appropriated several billion dollars toward stockpile expansion, an industrial-base support fund, and direct-loan authority, alongside explicit permission for the government to take equity stakes in the companies it finances, something it has already done in at least two transactions.

Anyone who follows the credit-creation school of economic thought will recognise the mechanism immediately. The core claim of that school is that growth is driven less by the price of credit than by where newly created credit is actually directed, toward productive capacity or toward speculation. Every economy that industrialised rapidly and durably (nineteenth-century Germany, postwar Japan and Korea, China after its opening) did so by steering credit deliberately into strategic production, shielded from the discipline of open capital markets. What is happening now in Washington is a rough, improvised version of that same playbook: public money manufacturing credit and steering it into a chosen slice of the economy that ordinary capital markets, left to their own devices, would never fund, because the risk-adjusted returns are too thin and the payback period too long for a normal investor’s patience.

There is a genuine irony worth noting for anyone who has followed the argument that America’s own growth engine was hollowed out by decades of financialisation and the slow death of a decentralised, locally-rooted banking system that once funnelled credit into productive small business. The current minerals build-out is the state relearning that old developmental-credit instinct, but running it through a fund-of-funds structure administered by defence bureaucrats and private-credit managers rather than through anything resembling a network of local banks lending against real productive collateral. It is industrial policy grafted onto financial plumbing built for an entirely different purpose. Whether a credit fund can substitute for the deeper thing that actually built industrial ecosystems historically, patient local banking relationships plus decades of accumulated workforce and process knowledge, is a genuinely open question, and one worth being sceptical about.

A parallel version of this is playing out in Europe, where a major American and a major German defence manufacturer signed an agreement in mid-2026 to build, for the first time, a production line for a well-known American-designed missile system outside U.S. borders, backed explicitly by both governments and aimed at shortening supply lines and building sovereign capacity. Two U.S. treaty allies in Asia, both direct targets of mineral-related pressure in the past year, are pursuing similar strategies of their own. The common thread across all of it is a government actively choosing which industrial capabilities to build and finance, a posture that would have been unthinkable in Western economic policy circles a decade ago.

This is where most of the commentary on this theme goes wrong, so it is worth being deliberate about it: how should any of this actually be valued?

The natural reflex is to treat it as a commodity story: supply curve, demand curve, marginal cost, clearing price. That reflex will lose money here. What is actually being underwritten is not a commodity cycle; it is a policy bet dressed up in commodity language. Cash flows in this sector depend, to a degree that should make anyone uncomfortable, on subsidised financing terms, government offtake commitments, implicit price floors, and, above all, on the continued existence of the export restrictions that create the artificial scarcity in the first place.

Three things follow from that.

First: subsidised capital distorts every discounted-cash-flow model built on it. When a government lender extends credit below market terms and effectively backstops the buyer of last resort, a project’s realised cost of capital tells you almost nothing about its underlying commercial risk. That inflates valuations relative to what an honest, unsubsidised analysis would produce. The subsidy is real and probably durable for a few years, but it is not permanent and it is entirely subject to the next election cycle or the next budget fight. Treat it as a separate, explicit, probability-weighted item rather than folding it silently into a lower discount rate.

Second, and this is the risk that dwarfs everything else: the price umbrella that makes a Western processing plant economic exists only because the dominant supplier is currently choosing to restrict output. The moment that supplier decides, for its own reasons, to ease those restrictions and let global prices fall, as it has already partially done, the umbrella collapses and a meaningful share of Western capacity is revealed to be uneconomic at unrestricted market prices. This is not a low-probability tail event. It is the ordinary, rational behaviour of a dominant supplier managing a powerful but self-limiting tool. Any valuation that does not explicitly model a scenario where restrictions ease and prices fall by half is not really a valuation at all.

Third, there is a market-structure dimension worth taking seriously on its own terms. Themes like this one attract flow-driven capital, ETF allocations and momentum mandates that are not underwriting discounted cash flow at all; they are buying a story and a ticker. In a corner of the market with only a handful of genuinely pure-play, liquid names, that kind of flow can push valuations far above anything a fundamentals-based model would support, and the process is reflexive: rising prices reinforce the narrative, which draws in more flow, which pushes prices higher still, in a loop that has nothing to do with underlying cash generation. The same concentration and fragility that defines the physical supply chain has an unnerving mirror image in the trading structure of the securities built on top of it. When that flow reverses, it will reverse hardest in the names that led the rally, and fundamentals will offer little protection at the top.

Put together, the discipline required is fairly simple to state even if it is hard to execute: strip out the durable underlying economics from the subsidy-and-scarcity premium sitting on top of them, treat the policy support as contingent rather than permanent, and decline to pay momentum-driven prices for cash flows that depend entirely on a foreign government’s continued restraint.

If that framework holds, it has fairly specific implications for where along the chain an investor is being paid for something real versus simply renting a policy position.

• Processing capacity, not mine ownership, is the scarce asset. Owning reserves in the ground is close to the least attractive way to express this view. Deposits are relatively plentiful, new supply is coming from several directions, and a mine sells an undifferentiated commodity into a market whose price is effectively set by one government’s policy choices. Refining, separation and magnet-manufacturing capacity, particularly when paired with feedstock that does not run through the dominant supplier’s system and offtake agreements backed by a government buyer, is where genuine and durable pricing power lives. If you are going to take on the policy-reversal risk at all, take it on at the point in the chain where you are also buying a real, physical asset that would take years for anyone to replicate.
• Locked-in supply and locked-in demand separate the survivors from the rest. Within processing, the strongest businesses have secured both ends of their supply chain: feedstock that bypasses the dominant supplier, and offtake at contracted, often government-backed prices. A magnet producer with a decade-long defence-department contract and non-dependent feedstock is underwriting a fundamentally different risk than a merchant processor selling into spot markets. One survives a normalisation in global supply; the other does not.
• Defence primes and dual-use manufacturers offer a lower-beta way in. The large contractors and dual-use industrial firms being pulled onto expanded production schedules, missile co-production agreements, munitions capacity expansions, automotive manufacturers being converted toward defence output, capture much of this theme’s upside with far less exposure to swings in metal prices than the pure materials names carry. Their risks look different (programme delays, budget politics, execution) but they largely sidestep the specific policy-reversal risk that defines the raw-materials complex, which for many portfolios makes them the more comfortable way to hold this exposure.
• Diversifying across countries reduces one risk but not the main one. The build-out spans several jurisdictions and several allied “friend-shored” supply corridors. That geographic spread genuinely reduces exposure to any single government’s domestic politics, but it does nothing to reduce the systemic risk, which is a single dominant supplier’s pricing behaviour. A basket of processors spread across five countries is still, underneath the diversification, one large bet on how long a particular set of export restrictions persists.

The plain-language summary: durable value sits in policy-backed, vertically-integrated processing capacity; the fragile money is in unhedged raw-materials exposure bought at momentum-driven prices; and the steadier way to own the broader rearmament trend is through the demand side of the equation rather than the supply side.

Three things hold simultaneously, and the discipline is in holding all three at once rather than picking the comfortable one.

The dependency is genuine, structural, and will take most of a decade to meaningfully unwind. Even under an aggressive buildout scenario for capacity outside the dominant supplier’s borders, independent energy-agency modelling suggests the rest of the world will cover only around a quarter of global refining needs and less than a fifth of magnet demand by the mid-2030s. On any realistic timeframe, full decoupling is not achievable; diversification is, slowly.

The Western policy response represents a genuine change in regime: the return of deliberately directed industrial credit, governments acting as strategic capital allocators, and the rediscovery of a developmental playbook the West had largely abandoned. That is a real and probably durable tailwind for the sector as a whole, and it is why this deserves to be treated as more than a passing trade.

And yet the individual securities riding this wave remain policy-dependent, distorted by thematic capital flows, and exposed to exactly the kind of concession the dominant supplier has good reason to eventually grant. The imbalance hanging over Western defence readiness in the near term has a mirror image hanging over anyone holding Western processing capacity at inflated, subsidy-dependent valuations: whoever created the scarcity that supports today’s prices can, at a moment of their choosing, take it away again.

The genuine edge here is not in recognising the theme; everyone already sees the theme. The edge lies in separating durable midstream economics from the temporary premium created by subsidy and artificial scarcity, in treating policy support as conditional rather than assured, and in refusing to pay momentum prices for cash flows that ultimately depend on a foreign government’s continued forbearance. In a supply chain, and a market, both defined by concentrated points of failure, the job is the same for the strategist and the investor alike: identify the piece that genuinely cannot be replaced quickly, and pay only for the portion of its value that would survive the other side deciding to let go.