Chung Huynh
← Writing

The Scarce Inputs Thesis

September 18, 2026

Written with AI assistance (Claude). The thesis and reasoning are mine — the AI helped me draft, structure, and stress-test them.

An investment framework for hard assets in a multipolar, energy-constrained world.

Every plausible future runs on the same physical foundations. Whatever technologies come out on top, whoever gains the upper hand in a fragmenting world, it all still needs energy and raw materials — and the supply of both is structurally constrained. That constraint is the one durable thing I’m willing to build a portfolio around.

Most investors spend their effort trying to name the winners: the right companies, the right technologies, the right outcomes. I’ve taken a different job. Rather than bet on the uncertain outputs of the coming decades, I own the scarce inputs they all depend on. The market has lately put a name to the same instinct: Goldman Sachs and Morgan Stanley have been making the case for what they call HALO — Heavy Assets, Low Obsolescence — stocks. But the idea is older than the label, and I’ll come back to it.

Take AI, the loudest example. I think it’s one of the most powerful technologies of our time, and I expect it to disrupt one industry after another. That same power is what makes it hard to invest in directly. When change moves this fast, it’s difficult to say which companies will benefit and which will be swept aside. Today’s obvious winner can become tomorrow’s casualty, and it can happen fast. Meanwhile the enthusiasm is enormous, and enthusiasm inflates prices — so a company can be a real beneficiary of AI and still be badly overpriced. Whether the market has run ahead of itself is a different question from whether the technology is real, and a much harder one.

That is the whole idea. The rest is discipline.

This is mainly a scarcity-and-cycle bet: the direction of the return comes from physical supply staying tight while demand rises, not from any one company out-competing its rivals. That said, there is one moat I actively look for — a producer sitting in the lowest quartile of the cost curve on a tier-one deposit (large, long-lived, high-grade). In a commodity business that’s the only moat that lasts: price is set by the marginal producer, so the low-cost operator survives every trough and captures the most margin in every boom. Which companies I own decides how much of the rising tide I actually capture, so I analyze hard and buy selectively, never just “the sector.” My edge is therefore twofold. A ten-year time horizon lets me sit through the drawdowns that shake out more impatient money. And I fish deliberately in the market’s less-efficient corners — junior, small- and micro-cap miners — where careful research can still find mispricing, because they’re too small, too illiquid, and too technical to be picked over the way mega-cap tech is.

Underneath sit three macro currents: the slow diversification of central-bank reserves away from the US dollar, structurally rising energy demand in a fragmenting world, and AI as a bonus rather than a bet. I express them through a focused set of positions: physical gold and silver, the miners that produce them, copper companies, and the miners and refiners of critical minerals — with uranium a growing conviction I’m still building into. Each comes with a pre-defined set of conditions that would tell me it’s broken, because in this game a falling price is never by itself a reason to sell. Only evidence beats price.

Contents

The edge is patience, not prediction

“The real trouble with this world of ours is not that it is an unreasonable world, nor even that it is a reasonable one. The commonest kind of trouble is that it is nearly reasonable, but not quite. Life is not an illogicality; yet it is a trap for logicians. It looks just a little more mathematical and regular than it is; its exactitude is obvious, but its inexactitude is hidden; its wildness lies in wait.”

— G.K. Chesterton, Orthodoxy (1908)

I can’t outsmart the market over weeks or quarters, and I don’t try. The shorter the horizon, the more information you need to be consistently right, and against the desks and machines that dominate short-term trading I have no such advantage. What I do have is a game they can’t easily play: holding for a decade, long enough that fundamentals drown out noise. Patience here isn’t a virtue I claim for its own sake — it’s the entry ticket to an edge the impatient forfeit.

The inefficiency edge is where I point that patience. Not at mega-cap tech, where thousands of analysts price every rumour in minutes and any “insight” is usually the consensus arriving late — but at junior, small- and micro-cap miners, which are too small for the big funds, too illiquid for the fast money, and too technical for the generalist to bother with. Those markets genuinely aren’t efficient, so careful research can still find mispricing there. And the two edges feed each other: the long horizon is exactly what makes an illiquid, slow-to-correct mispricing survivable. I need the patience because I’m fishing in inefficient water. Howard Marks put the underlying constraint plainly: “Readily available quantitative information with regard to the present cannot be the source of superior performance.” Everyone has the same screens and the same filings. Whatever edge exists has to come from judgment applied where fewer people are looking — which is a fair description of where I look.

One clarification I owe the reader, though: less efficient is not the same as easy. The inefficiency is compensation for real difficulty — thin liquidity, promotion and outright fraud, geological complexity, financing risk — so the edge there is findable, not given, and it stays a hypothesis until the results confirm it. Neglect creates the chance to be right; it doesn’t make me the smart money by default.

I’ll admit this edge is a hypothesis about my own behavior, and every drawdown puts it back on trial. It’s worth only as much as the written framework behind it, because conviction without a framework doesn’t survive contact with a 40% decline. That’s what all the rules that follow are for: so that when a position is halved and instinct is screaming, I can answer the only question that matters — is this broken, or just painful? — with evidence I wrote down in calmer times.

What I’m actually aiming for

Let me be plain about the ambition, because it isn’t modest. On the positions where the theses prove right, I’m aiming for tenfold returns over roughly ten years. That’s compounding at about 26% a year — call it triple the long-run return of a broad stock index. By any conventional standard that’s aggressive, and I hold the goal consciously rather than pretending otherwise.

The ten years are a target, not a deadline. Commodity cycles don’t read calendars, and a thesis that’s intact but late gets held, not sold.

Where would a return like that even come from? From three growth levers that can stack on top of one another. The commodity itself re-prices higher. The company scales production into that higher price. And exploration expands the resource base beneath it. Those effects multiply rather than add — and that multiplication is the whole source of a tenfold outcome. I look for companies where all three are credible at once. The third only counts per share: if a miner grows its resource by printing new shares to pay for it, that isn’t growth, it’s dilution wearing a costume.

The consequence is a portfolio tilted toward small- and micro-cap companies. The macro tide lifts every boat in the harbor; the job is to find the ones it lifts highest.

I set a floor, though. I don’t buy pure grassroots explorers — companies drilling holes in the ground with no discovery yet. That’s a lottery ticket, not a thesis; none of the three levers exists until there’s something real in the ground. The developers I do own, I hold partly as acquisition bait: a decade of underinvestment in exploration has left the mining majors with depleting reserves and no cheap way to replace them except by buying someone who already found the deposit. A takeover is a welcome exit — but never the plan. Each position has to stand on its own economics, because a takeover premium alone won’t deliver a tenfold return. The re-rating on the road to being worth acquiring is what does.

And the caveats, because the levers cut both ways. Those same three can stack in reverse: the natural failure mode of a junior miner is an 80–90% loss or a total one, and most explorers never reach production at all. So the portfolio lives on a power law — a minority of winners must pay for a graveyard of losers — and the thing keeping me solvent is position sizing, not conviction. These companies are also illiquid when everyone heads for the exit at once, which means my selling has to be triggered early, by evidence, not by waiting for a buyer to materialize in a panic. None of this overturns the hierarchy: the macro tide sets the direction, and the sharpest selection won’t save a portfolio whose underlying scarcity thesis is wrong. But selection is more than pruning broken boats — within a rising tide it’s what separates a good outcome from a spectacular one, and it’s where my research is meant to pay for itself. I don’t claim brilliance; I claim diligence, aimed at a corner of the market inefficient enough to reward it.

Buying, and selling

Because patience is the edge, it has to be active. A drawdown in a position whose thesis is still intact isn’t a threat; it’s a sale. Two rules follow. First, I’m never fully invested — the dry powder isn’t lazy cash, it’s the ammunition the whole strategy depends on. Second, I think through my next moves before I need to make them. I don’t run a rigid ladder of fixed amounts at fixed levels — but I do know, in advance, which positions I’d want to add to on the way down and roughly what would justify it, so that a drawdown finds me with a plan rather than a reflex. I know from experience that at the moment of maximum opportunity, the courage to improvise simply won’t be there. So I do the thinking early, while it’s still cheap.

Selling happens through exactly three doors. A thesis breaks — some pre-defined piece of evidence tells me the macro story was wrong. A company breaks — something specific to that business trips a wire. Or something better comes along — which I allow, but only with discipline: the replacement has to earn its place against the position it’s displacing, and that means real research, not a hunch. If a new idea can’t hold up when set beside the old one, it was an impulse, not an insight.

The three currents

The dollar loosens its grip

The dollar isn’t going anywhere as the world’s reserve currency — nothing else has the depth, the liquidity, or the legal plumbing to replace it. But it doesn’t need to collapse for this to work. It only needs to keep loosening at the margin, and three forces push that way: the weaponization of the dollar system through sanctions, a US fiscal path that keeps piling on debt and interest, and a slow erosion of institutional trust. Central banks, especially outside the West, are quietly trimming how many new dollars they accumulate. That’s a shift measured in years and decades, not a crash — and the thesis needs only the trend, not a crisis.

One wrinkle I measure carefully: I count my returns in euros. If the dollar weakens, some of my gain evaporates in home-currency terms. So the real bet isn’t “the dollar falls” — it’s “hard assets rise in every currency,” driven by fiscal excess everywhere, not just in Washington.

There’s a genuine counter-current worth naming, because it cuts against me: stablecoins — private dollar tokens like USDT and USDC, now roughly $320 billion outstanding, with the largest issuer holding more US Treasuries than Germany does. Every holder in Buenos Aires or Istanbul or Lagos is a brand-new dollar user, and US policy openly treats this as a way to defend dollar dominance. But it doesn’t sink the thesis, for a subtle reason: my bet is about official reserve behavior and fiscal erosion, while stablecoins are about private transaction demand. Both can be true at once. And there’s an irony that almost makes my point for me — all those stablecoin reserves buy Treasuries, which lets the government borrow more cheaply, which funds exactly the fiscal excess that makes gold attractive in the first place. The counter-force feeds the thing it opposes.

A multipolar, electrifying world

American primacy is giving way to a more multipolar world, even if the US stays the single strongest player in it. As living standards rise across that world, they bring electrification and consumption with them, and structurally higher energy demand behind both. That lifts every energy source — renewables fastest, while fossil fuels remain indispensable through the transition (what happens to oil demand past 2030 is an open question). The catch is that generating the energy is only half the problem: delivering it requires a vast buildout of grids and infrastructure, and that’s where a great deal of the material demand lives.

Layered on top is fragmentation. Export controls, trade wars, and economic weaponization aren’t a passing phase; they’re structural. And for a materials investor, fragmentation is a tailwind on the supply side: when the West sets out to duplicate supply chains China already controls, it manufactures demand that wouldn’t exist in a seamlessly globalized world, and it puts a strategic premium on any critical-minerals capacity sitting outside China. That premium is the foundation of my rare-earth position.

The politics underneath deserve a plain look, without partisanship. The rightward shift visible across many democracies is mostly national-populist rather than free-market — tariffs, industrial policy, reshoring, migration limits — which is structurally anti-globalization, the opposite of the Reagan–Thatcher right that once drove globalization forward. That accelerates the fragmentation trend. But the thesis doesn’t lean on it, because in the US de-globalization has become bipartisan: the China tariffs survived a change of administration, and industrial policy and export controls expanded under both parties. A change of government isn’t the thing that breaks this pillar. Two side effects are worth noting. Populist governments rarely balance budgets, which quietly feeds the debasement story behind gold; and where they cut support for renewables, the energy thesis barely flinches, because its core claim is about electricity demand and grid buildout — technology-agnostic — with the generation mix simply tilting toward gas and nuclear.

Fragmentation doesn’t make China disappear as a risk, though — it just relocates it from supply to demand. Copper has one global price, and China drinks up about half the world’s supply. If Chinese construction contracts faster than electrification and emerging-market growth can fill the gap, the price falls no matter how the supply chains are drawn. That risk stays on my watch list, permanently.

AI, as a bonus and not a bet

I have no special insight into who wins artificial intelligence, and I’ve stopped pretending otherwise. The model layer is commoditizing. The application-layer moats are anyone’s guess. The chip-layer advantages — ASML, TSMC, Nvidia — are real but come loaded with concentrated geopolitical and competitive risk. Judging the durability of the mega-cap tech leaders’ growth is not an edge I possess. That’s humility, not a valuation call.

But every version of the AI future runs on energy and materials. So AI, for me, is upside optionality rather than a dependency. Data centers are around 4% of copper demand today, perhaps 6% by 2040 — a garnish, not the meal. The load-bearing demand comes from grids, electrification, and emerging-market development. The thesis survives an AI winter; my entry prices might even enjoy one.

Is AI a bubble right now? My framework doesn’t have to answer that, which is precisely why I like it. Valuations may be stretched — maybe, maybe not. I don’t know how large the ultimate impact will be, and I don’t know who captures it. But underneath all the uncertainty, one thing isn’t in doubt: AI is happening, its footprint on the world keeps growing, and it needs energy and materials. That narrow claim — more demand for physical inputs, not more earnings for any particular company — is the only AI exposure I take. The fair caveat is that this slice of my demand story is quietly correlated with the very tech valuations I avoid; hyperscaler spending sits upstream of my copper. An AI capex bust would dent the thesis at the edge — maybe a tenth of it — not at the core.

This is the HALO idea I flagged at the start. The case Goldman and Morgan Stanley have been making is that businesses built on real, hard-to-replicate physical assets are structurally insulated from AI while asset-light software gets repriced downward — and Goldman’s phrase for it, “a repricing of scarcity,” is close to the language of this document. I take it as confirmation more than instruction. One distinction I keep in view: HALO’s moat is the irreplaceable network — own the grid because no one can build a second — whereas the moat I look for is cost position on a world-class deposit, inside a bet that is mainly about scarcity; different logic, heavily overlapping holdings.

The holdings

Physical gold and silver anchor the book; everything else I own through companies — the miners that produce those metals, copper producers, and the miners and refiners of critical minerals. The split is deliberate: metal where the case is monetary, equities where the scarcity and the operating leverage live in the mining and the processing.

Gold

Gold is the first beneficiary of a diversifying reserve system: the only reserve asset with no counterparty and no sanctions risk, accepted everywhere, legally and culturally, in a way nothing else — crypto included — can match. Central-bank and institutional demand for it is structural and likely to persist. Today’s price may already reflect much of that, so my edge here isn’t timing the entry; it’s holding through the volatility, accumulating steadily over a long horizon.

The part that excites me is an asymmetry in ownership. Gold is drastically under-owned in Western portfolios by historical standards, and because the investable supply is small and slow to grow, even modest shifts in allocation move the price out of proportion. Roughly speaking, investor-held bullion — bars, coins, ETFs, and the rest — comes to something like $9 trillion against perhaps $320 trillion in global financial assets. That’s about 3%, versus around 14% four decades ago. If private allocations moved even a point or two back toward gold, that flow would collide with a tiny float of available metal, and it’s the price, not the quantity, that would give.

I’ll temper that: it’s an asymmetry, not a guarantee. Some of it is already in the price, the old highs came from gold-standard and mania eras that make poor anchors, and every allocation statistic shifts with how you define it. Treat the numbers as illustration, not a target.

Gold miners

A good gold miner is leverage on the metal: costs stay roughly fixed while the gold price rises, and the margin expands underneath. The catch is that the sector as a whole has historically underperformed gold itself, undone by cost inflation, dilution, and poor capital allocation. So this is emphatically a stock-picking exercise, not a buy-the-sector one. (My specific selection criteria are being rebuilt as I write this, and I’d rather leave them out than publish a version I’ve already outgrown.)

Silver

Silver rides gold’s monetary tailwind, but in a weaker and more jittery form — central banks don’t hold silver as a reserve asset (Russia’s state fund began adding a little from 2025, the lone and small exception), so the monetary bid comes almost entirely from retail and investors. Its stronger leg is industrial: silver’s conductivity makes it hard to remove from solar cells, electronics, and electrification generally. The risk to watch is thrifting — industrial users engineering out some of the metal when the price climbs. Like gold, I hold it two ways: as metal, and through silver miners, where the same fixed-cost leverage that lifts a gold producer’s margin applies to the silver price.

Copper

Everything structural points copper’s way: mines that take fifteen years or more to build, ore grades that keep declining, relentless grid and electrification demand, and buyers — data centers among them — who don’t much care what it costs. A serious deficit looks likely. But the timing is uncertain, because high prices summon their own antidotes: scrap supply, substitution (aluminum in transmission lines), and demand that quietly retreats. Those forces have pushed predicted copper deficits further into the future before, more than once.

It helps to know what actually carries this thesis. Construction and industry are about half of copper demand; grids and electrification about a third; AI and data centers a mid-single-digit garnish. I size the position for the structural trend, not for a particular deficit arriving on a particular date. And I take it through copper companies, not the metal — copper isn’t something I hold physically the way I do gold and silver; the equity is where the leverage to the price sits.

Critical minerals

Rare earths are the flagship of this position, so start there — and start with the name, which misleads. Rare earths aren’t geologically rare — they’re reasonably abundant, in deposits scattered across many countries. That fact does more than puncture a misconception; it sets the bar for the mining side of the bet. When a material is abundant, a merely decent deposit is worth little, because there’s no scarcity in the ground to reward it. Only a world-class deposit earns its place — exceptional in grade, in size, and above all in economics — the kind of project whose costs keep it standing wherever the price sits in its cycle. So here I’m not hunting for rarity; I’m hunting for quality.

The real scarcity lives one step downstream, in separating and refining the ore: a difficult, capital-hungry, environmentally punishing process that China dominates. By the IEA’s reckoning, China holds an average share above 70% across nineteen of the twenty most strategic minerals, and its grip is tightest precisely at that refining stage. The US Geological Survey’s methodology for the 2025 critical-minerals list draws its risk boundary at refining for the same reason — that’s where the vulnerability lives — and documents both China’s recent export controls on medium and heavy rare-earth items and the American response, such as the Department of Defense taking a stake in MP Materials. What’s being priced, in other words, is processing capacity outside China, not ore in the ground.

That combination — abundant rock, scarce refining — is what draws me to the whole rare-earth complex, light and heavy alike. These sit on the USGS critical-minerals list for good reason: high economic impact, and genuinely hard to substitute in the industries that lean on them — the magnets inside electric motors, wind turbines, and defense systems, and the electronics threaded through modern life. When something is both essential and supply-insecure, the West has to build non-Chinese capacity almost regardless of cost, which makes this a policy-driven thesis rather than a pure demand-growth one. That strategic bid — government funding, price floors, guaranteed offtake — is largely deaf to the commercial forces that would normally cap the upside. The heavier rare earths, dysprosium and terbium, are the scarcer and more strategic corner, but the light workhorses behind neodymium magnets sit on the same fault line. There’s no drop-in substitute for those magnets, though designers can sometimes engineer around them, and clever process improvements have already cut how much dysprosium and terbium each magnet needs — even as removing them entirely from high-temperature uses stays hard.

The things that would break it: China relaxing its export controls and crushing the price, or Western political support quietly stalling out.

Rare earths are the flagship, but I treat the wider critical-minerals complex the same way. The 2025 USGS list runs to dozens of minerals flagged for high supply risk and economic importance, and a familiar pattern repeats across the sharpest of them — tungsten, gallium, germanium, antimony: unremarkable geology, a refining or processing step concentrated in China, and a fresh round of export controls that hands any capacity outside China a strategic premium. So I look in two places at once — world-class deposits in sound jurisdictions, and the miners and refiners building the processing capacity the West can no longer do without. The metal is rarely the point; the chokepoint is.

Uranium

One area I’m increasingly convinced by, and still building into rather than fully positioned in. The setup rhymes with the rest of the book: demand is turning up structurally as nuclear returns to favour for clean baseload power — now including the data-center operators chasing firm, around-the-clock electricity — while supply spent a decade underinvested after Fukushima. And a critical stretch of the fuel cycle, enrichment and conversion, is concentrated in Russia in a way that mirrors the refining chokepoint in critical minerals. I’m treating it as a conviction to scale into deliberately, not a position to complete in a hurry — more to come as the thesis firms up.

What I accept, and what higher rates would do

“It ain’t what you don’t know that gets you into trouble. It’s what you know for sure that just ain’t so.”

— commonly attributed to Mark Twain (though likely apocryphal)

Every strategy hides its risks in the fine print; I’d rather put mine in the open.

Three risks I take by choice. These are deliberate exposures, not blind spots:

  • A deflationary bust. In a real credit event, copper, silver, and the miners fall hard, and only gold has historically held its footing (after an initial scramble for liquidity). I accept that exposure on purpose.
  • Chinese demand. China consumes roughly half the world’s copper, so a sharp, sustained contraction there would drag the price down regardless of how the supply chains are drawn. I handle this through my sell rules, not through the fragmentation story, since the two sit on different sides of the ledger.
  • A single correlated bet. Every thesis points the same way — dollar down, hard assets up. That internal coherence is a strength when I’m right and a concentration risk always. I owe myself a hard ceiling on how much of my net worth rides on it, and setting that number precisely is unfinished work.

The one macro variable that reaches inside: higher-for-longer interest rates. It’s the question I get asked most, and one of the few forces that truly touches this framework — so it’s worth walking through by thesis rather than waving away.

  • Gold — a real headwind, not a fatal one. Higher real rates raise the opportunity cost of holding something that pays no yield, and history bears that out. But the thesis was never built on low rates; it rests on fiscal stress and central-bank diversification. With US interest costs compounding and no serious move toward a balanced budget, higher rates without fiscal discipline only bring forward the eventual choice between inflation, financial repression, and restructuring — and all three are kinder to gold than to paper money. The one combination that would truly falsify it — high real rates and a credible, sustained US surplus at once — is exactly the tripwire in my notes.
  • Energy and materials — barely touched. Grids, electrification, and defense procurement are policy-driven and largely non-negotiable; they don’t get postponed because capital grew expensive, and defense spending ignores the rate cycle almost entirely. Rate-sensitive corners like rooftop solar and some renewable project finance slow at the margin, but the core demand holds.
  • AI — narrative, not copper. The exposure was always to physical inputs rather than tech valuations, so higher rates pressure the story more than the metal underneath it.
  • Developer financing — the real pressure point. This is where higher rates bite hardest, and it’s in my own backyard. Small developers fund themselves by issuing shares; when capital turns expensive and nervous, those raises get smaller, more dilutive, and slower — stretching out the time a developer needs to become worth acquiring, and quietly eroding the per-share value the whole tenfold math depends on. The 2022–23 rate shock showed this vividly across junior mining. I treat it as a pace-of-execution risk, not a reason to abandon the thesis: a developer that stays on schedule despite expensive money is proving its quality, and one that stalls or dilutes recklessly trips my position rules anyway. So I manage it company by company, not by drawing a line at some rate level.

The whole thing in one breath

I don’t know what the next decade holds, and I’ve built a strategy that tries not to need to know. Own the scarce physical inputs that every plausible future consumes; buy them through the drawdowns that scare off shorter-horizon money; write down in advance what would prove me wrong; and then be patient enough to let a slow thesis become an obvious one. The edge isn’t cleverness. It’s the willingness to hold.

This document describes a personal investment framework. It is not investment advice, and nothing in it constitutes a recommendation to buy or sell any security or asset.