Gavin McCracken (Pt. 2): Robotics, AI, and A Commodities Supercycle

Silver is unusual among precious metals: about half of it gets used up in industry rather than stored in vaults. It conducts electricity better than anything else that is affordable, so it ends up in solar panels, EV batteries, and every electrical contact inside a machine.

McCracken's argument is that the market is pricing silver as a monetary metal that happens to have industrial uses, and has not yet done the arithmetic on robots. His unit of measurement is blunt: roughly an ounce of silver per Tesla Optimus humanoid. If you take Musk's stated ambition of one in every household on the planet even half-seriously, the ounces required run into the hundreds of millions per year against an annual mine supply of well under a billion. He thinks the recent price inflection is the market beginning to sense this — "I think actually it's front running robotics."

The practical part is how he expresses it. He puts a number on the metal — "$100 is about to be the floor price" — but immediately caps the upside: "it may not go much higher. It might go there and sit there." A metal that goes to a level and parks there is a poor place for new bullion, so he is not buying physical (what he holds is old stock from $40–50, kept as an emergency fund). Instead he wants miners: if silver settles at $100 and a miner's cost of production is, say, $20, the profit per ounce multiplies far faster than the metal price does. That is operating leverage, and it is the reason to own the company rather than the commodity when you expect a price step-change followed by a plateau.

There are two headline oil prices. WTI is the American benchmark, priced inland in the US; Brent is the international one, priced on seaborne cargoes. Normally they trade within about $3 of each other, because oil can be shipped between the two markets.

McCracken's position is not really a bet on oil going up. It is insurance against a specific political event. He runs his portfolio on margin — borrowed money — which means a sharp adverse move can force his broker to sell his holdings at the worst possible moment (a "margin call"). If Washington banned the export of US crude, American oil would be trapped at home and cheap while the rest of the world's oil went vertical. Every North American producer he owns would fall; Brent would spike. So he keeps roughly 5% of his net worth in long-dated, out-of-the-money Brent call options.

A call option is the right, not the obligation, to buy at a fixed price. "Out of the money" means that fixed price is far above today's, so the option is cheap and worthless unless something dramatic happens. "Long dated" means it has years to run, so he is not forced to be right about timing. His framing of the payoff is deliberately extreme — if both the US and Canada stopped exporting, "there's no way Brent stays below 500… I think Brent would go to 1000."

He also explains why this insurance is not obviously mispriced, and why it might still work: the Brent premium over WTI had doubled to about $6, which tells him others are hedging the same way. His warning about crowded option hedges is worth keeping: "they don't matter until they matter. And then the gamma blows someone's head off" — i.e. the seller of those options is fine until the price starts moving, at which point their losses accelerate uncontrollably.

Tungsten is the densest common metal with the highest melting point, which makes it the material of choice for armour-piercing penetrators, missile components and industrial cutting tools. China controls the large majority of supply.

The demand case McCracken makes is the simplest kind: the material is being destroyed, not stockpiled. "Russia, Ukraine was constantly blowing shit up made of tungsten. And now in the Middle East, they're blowing shit up made of tungsten. Just vaporizing it." Every munition fired is tungsten that has to be mined again. Layer on Western rearmament and a US administration talking about a "military economy," and demand goes up while supply sits behind a geopolitical wall.

He is unusually direct that he got the sizing wrong — "tungsten I think is my biggest mistake… I should have had more" — and equally direct that the miss does not invalidate the thesis: "I feel like tungsten is still great here." That distinction (a missed entry versus a broken thesis) is the whole point of the segment.

It also illustrates why he prefers raw materials to the companies that use them: as Beylo notes, there is almost no way to buy pure exposure to missile rearmament through a defence contractor, because every prime is a conglomerate of unrelated programmes. The metal is the pure play.

EQ Resources is an Australian-listed tungsten producer — a small company operating actual tungsten mines, which is a short list globally outside China.

It appears here as the equity behind the tungsten call, and specifically as the one McCracken says he should have owned more of. Both he and Beylo describe the chart as having already run hard, so this is not presented as a fresh entry at a good price. It is presented as evidence for a rule: when you can identify a commodity whose demand is structurally rising and whose supply is politically constrained, the handful of listed producers are the leveraged expression — and hesitating on valuation while the thesis plays out is itself a cost.

Worth noting the honest asymmetry: he made the tungsten call and did not size it, which is why the stance here is a positive view on a name he under-owns rather than a recommendation to chase.

Micron makes memory chips — DRAM and the high-bandwidth memory that sits next to AI accelerators. Memory is the part of a computer that holds data the processor is actively working on, as opposed to the processor that does the arithmetic.

McCracken arrives at Micron from first principles rather than from the AI narrative. His argument earlier in the episode is that digital computers are inherently inefficient — they win on generality, and pay for it by having to shuffle enormous numbers of bits around. That shuffling is a memory problem, not a compute problem, which is why he says "you end up needing all this memory and Micron stock's going up."

The demand evidence is personal and checkable: he priced a machine to run his own large language model locally and the quote was about $4 million, "and most of the cost is memory." If one individual faces that bill, so does every small business or independent lab that wants a private model. That is Jevons paradox — the observation that when a resource becomes cheaper or easier to use, total consumption of it usually goes up, not down, because far more people start using it.

The framing that makes it fit his book: he thinks memory and compute are "evolving into commodities." He does not want to pick which model wins (Gemini, GPT, Claude are "all doing the same thing"); he wants to own the standardised input they all consume. His warning is that this leg is not over — it will "break people who aren't long memory when they watch it go off." Note the sizing though: he calls it a "very small position," bought on a dip.

ROK Resources is a small oil producer drilling in southeast Saskatchewan. The thesis has nothing to do with the oil price forecast and everything to do with tax.

A royalty is the share of a well's production the provincial government takes off the top. In Saskatchewan that is normally around 25%. McCracken discovered that the province has granted a holiday: the first 38,000 barrels from each new well pay only 2.5%.

Why that is a bigger deal than it sounds: the main risk in drilling is that a well costs a fixed, large amount up front and then may not produce enough to pay that back. Cutting the government's take by roughly 22 percentage points over the well's most productive early barrels means, in his words, it "basically guarantees that these wells will pay off their own cost as long as they're not a dud." The economics stop depending on a bullish oil price and start depending only on the well not being a total failure.

He is equally clear about what could break it, and it is not geology: this is a Canadian barrel sitting inside a live cross-border trade fight. If Washington bans crude exports and Ottawa retaliates by withholding Alberta and Saskatchewan barrels, the netback for a Saskatchewan producer changes overnight. That risk is precisely what his Brent calls are there to offset — the two positions are designed as a pair.

The second material to fall out of the robotics thesis, and it comes from asking a deliberately childlike question: what will the robots physically be made of?

His answer is aluminium, for a mechanical reason. A robot that moves has to accelerate its own limbs; every kilogram of structure costs battery life, motor torque and stability. So designers optimise hard for weight, and aluminium is the cheap, abundant, machinable structural metal with the best strength-to-weight profile. "You're going to want them to be light."

This is earlier-stage than the silver call — he has not sized it and says so ("we're going to have to keep our eyes open for other things"). It is included because the method is the transferable part: take a projected unit volume of a new physical product, decompose it into its bill of materials, and go find whichever input is already supply-constrained.

USO is an exchange-traded fund that tracks the price of crude oil by holding futures contracts. For a retail account it is the simplest way to own oil exposure — and, importantly here, to buy options on oil exposure.

This entry documents one specific decision. A headline about a Middle East "Memorandum of Understanding" knocked crude down to around $68 and took roughly 60% off his net worth from the peak. He believed the deal would not hold. Rather than simply sitting in his existing positions, he sold shares and used the proceeds to buy call options, mostly on USO — about 5% of the portfolio.

The reasoning is a piece of arithmetic worth internalising: on a rebound, shares recover roughly in line with the underlying, but out-of-the-money options can multiply many times over. So converting equity into options at the point of maximum despair concentrates all your remaining risk budget into the recovery. It worked — "that's what slingshotted me back and brought me back to life" — and he is honest that it was "probably the ballsiest thing I've ever done in my life."

The obvious caveat, which the episode does not soften: this is a maximally aggressive, single-outcome trade run on top of margin. It is a description of what he did, not a template for a diversified account.

Kuya Silver is a small silver developer/producer. Beylo calls it "the cheapest silver miner out there," which is exactly the kind of statement McCracken treats as a question rather than an answer.

His response is the whole reason this is a Neutral and not a Positive: "I think if they execute, that's interesting… but I barely own any because there's still this 'if they execute'." In junior mining, "execution" means the specific, unglamorous work of getting a mine from a resource estimate into steady production — permits, financing, construction, and then actually hitting the tonnage and grade you promised. Most of the reason a small miner trades cheap is that the market is discounting the chance it never gets there.

So the useful takeaway is about position sizing under uncertainty: he likes the silver thesis a great deal and still owns almost none of this particular vehicle, because cheapness that is entirely conditional on future operational success is not the same thing as value. He explicitly says he is "still unsure which ones are the best to play with."

Andean Precious Metals is one of the two positions (with Valeura Energy) behind the roughly 2,000x personal return that made McCracken's name. It appears in this episode not as a current holding but as a benchmark.

The number he quotes is the point: he bought it when it traded at "1.5 times enterprise value over cash flow." Enterprise value is the market value of the equity plus debt minus cash — what it would cost to buy the whole business outright. At 1.5x cash flow, the entire company was priced at about eighteen months of the cash it generated. That is a level at which you are being paid to take the risk rather than paying for the story.

His observation now is that this setup has largely gone: "the market is starting to behave a bit and price things more correctly, at least in miners." That is why the mining sleeve of his book is under-sized despite a bullish commodity view — not because he dislikes the metals, but because he has an anchor for what a genuinely cheap miner looks like and is refusing to lower it. The one place he still sees it is gold miners after a pullback, which he is waiting for.

Calumet is a US specialty-products and renewable-fuels refiner. It comes up only in the accounting of how his portfolio recovered to new highs: he held call options on it and the stock ran "from like $12.00 all the way to 27 or something."

There is no forward view attached, and he credits the idea to someone else. It is recorded here for two reasons: it shows the shape of what actually generates his returns (a small number of leveraged, high-conviction expressions rather than a broad book), and it is a rare case of a pundit naming which position drove a headline P&L number rather than leaving it vague.

This is the position change to pay attention to, because he cut it while agreeing the fundamentals are good.

He halved his gold-miner exposure for two separate reasons. The first is a claim about how gold behaves in a genuine liquidity crunch — a moment when investors need cash urgently and sell whatever they can, regardless of quality. Historically gold falls in those episodes, because it is the easiest thing to sell. His twist is to add the retail supply channel: "imagine the price of everything doubles. People are gonna pawn their jewelry… especially with the current gold price, people are going to sell it so fast." That is physical supply arriving exactly when financial demand is being liquidated.

The second reason is positioning, and it is the same logic he applies to every crowded trade: "literally everyone on Twitter called the bottom. Everyone." When a view becomes unanimous there is nobody left to buy it, so "I think gold pulls back a bit here."

The important nuance is what he intends to do about it. He is not bearish on gold as an asset — he is trying to be liquid before the crunch so he can buy into it. His stated dream sequence is oil spiking, liquidity cracking, gold crashing, and rotating the oil proceeds into cheap gold miners: "it'd be like a legendary play." His complaint is that the market is refusing to cooperate — "both gold and oil want to go up at the same time."

This is not a forecast that American oil will fall. It is a rule for screening out otherwise-attractive companies.

The risk is a US ban on exporting crude. If that happened, US barrels would be stranded inside the country, WTI would collapse relative to the world price, and every producer selling into WTI would see its revenue fall regardless of how cheap the stock looked beforehand. Beylo describes the practical effect on his own research: he finds a cheap exploration and production company, notices its output is all WTI-linked, and hears "this little Gavin on my shoulder" telling him not to.

McCracken's reason for raising the probability now is the trade fight with Canada. Eastern Canada depends on American refineries for its fuel, so a US products ban would be devastating there; Canada's obvious retaliation is to withhold Alberta and Saskatchewan crude, at which point "America is also going to run out of gas… there's no winning here. It's just shooting ourselves in the foot while people like China benefit."

The reason this belongs in a table at all: it dictates the structure of his entire energy book. He still wants North American assets, because he has ruled out Africa, South America and the North Sea on expropriation and tax grounds. So he owns the barrels and buys Brent optionality to neutralise the one policy outcome that would otherwise ruin them. Hedging is not free — "hedges cost us money… I'm paying a premium for options and I have no choice."

A commodity chemical company buys an oil-derived feedstock, processes it, and sells the output. Its profit is the spread between the two, so the input price matters as much as the selling price.

His objection is that the input side is currently unreadable. Naphtha — a light petroleum fraction that is the primary feedstock for plastics and much of the chemical chain in Asia and Europe — has been distorted by the Strait of Hormuz disruption. When your feedstock price is being set by shipping risk rather than supply and demand, you cannot forecast a chemical producer's margin with any confidence: "the chems are super hard because their input costs are distorted."

The generalisable lesson is that a commodity shock is not automatically good for everyone touching that commodity. It splits the chain into winners (whoever owns the scarce input) and losers (whoever has to buy it), and if you cannot tell which side a company is on, it is not an investment, it is a coin flip.

The most instructive negative in the episode, because he believes the bullish story and still refuses the trade.

Sulfur is a required input for phosphate fertilizer production, and much of the world's sulfur is a by-product of refining sour crude. A sulfur shortage therefore genuinely does threaten fertilizer supply, and would push fertilizer prices up. Plenty of investors reasoned from there straight to buying fertilizer producers — and lost money.

McCracken explains why: "unless the fertilizer company had sulfur on hand, that's a problem. That's gonna hurt their inputs." The producer is a buyer of the scarce thing. A shortage raises its costs before it raises its selling price, and if it cannot pass that through fast enough, the crisis compresses its margin instead of expanding it. Hence: "there's probably gonna be a fertilizer crisis and it's hard to make money off it."

This is the same test he applies to chemicals, and it is the most portable idea on the page: when you identify a shortage, do not buy the industry that consumes the scarce input — find whoever owns it. He could not identify a clean Canadian sulfur owner, so he passed rather than substituting a worse vehicle.

Unitree is the Chinese humanoid and quadruped robot maker that recently listed publicly. It is the most direct listed expression of exactly the theme McCracken spent the first half of the episode calling — and he passed.

His reason was price and nothing else: "I was looking at the price, it was IPO, and I was like, yeah, no." He had even met the company at a machine-learning conference, so this is not unfamiliarity.

That decision is the practical hinge of the whole episode. He believes robotics is the next era; he also believes the listed robotics equities have already priced in that belief, while the raw materials those robots require have not. As he puts it, the question is "which ones, and are they already priced correctly?" — and his answer for now is to own the picks-and-shovels (silver, aluminium) and wait for a moment when the equity leg is buyable, "basically like buying NVIDIA in 2022."

Beylo adds a second, non-financial objection to Chinese robotics specifically: in a bifurcating world, a Chinese-made machine inside a Western home is a policy risk as much as an investment one — either it gets banned, or, as McCracken puts it, "they just turn off and you lose a pile of money."