|
Part I
The Mechanism
|
Palladium.
The consensus narrative writes itself: electric vehicles are killing palladium demand, the metal is a relic of the combustion age, move along. CNBC runs a segment about the EV transition every third day and palladium sits in the background like a prop from a cancelled show. The sell-side has been writing the obituary since 2021.
The obituary is wrong. Not because EVs aren't coming — they are — but because the people writing it haven't checked what's happening on the supply side of the machine. The engine is seizing up faster than demand is fading, and nobody in the mainstream press has bothered to open the hood.
Here's what the EV-kills-palladium crowd hasn't processed: battery electric vehicle market share actually fell to 12.5% of new car sales in the European Union in May 2024, down from 13.8% a year earlier. Hybrids — which still need a gasoline engine and a catalytic converter loaded with palladium — are the segment that's accelerating. Range-extender EV sales rose 83% in 2024 alone. Every one of those cars uses 110% to 120% of the PGM loading of a pure ICE vehicle.
The demand floor didn't collapse. It shifted shape. And the supply ceiling is collapsing on top of it.
|
Part II
The Diagram
|
Story off. Wiring diagram on.
Three countries — Russia, South Africa, and Zimbabwe — produce roughly 90% of the world's palladium. That's not a supply chain. That's a single hallway with three doors, and two of them are jammed.
Norilsk's decline isn't sanctions-driven — it's geological. Ore grades are falling at the Talnakh deposits, and the Chernogorskoye replacement project won't contribute meaningfully until 2027. Russia's output is simply getting harder to pull out of the ground. BMI forecasts Russian mine production contracting 7% to 2.9 million ounces this year. The revenue is still there — Norilsk doubled its half-year profit to $2 billion on higher prices — but the tonnes are not.
South Africa, the second-largest source, is running older and deeper shafts every year. Sibanye-Stillwater is restructuring its loss-making Kwezi shaft. Costs keep climbing. Producers are returning cash, not building new capacity. Valterra's H1 metal-in-concentrate production was up 4%, but the industry-wide trajectory is flat to declining on falling grades and rising power costs.
The broader PGM inventory picture tells the story. Platinum's cumulative above-ground stock drawdown since 2023 equals roughly 42% of beginning inventory; what's left covers approximately four months of global demand. Palladium's own above-ground stocks have been eroding along the same trajectory — five consecutive annual deficits will do that. That's not a buffer. That's a rounding error with a fancy name.
COMEX registered palladium inventory dropped below 200,000 ounces in early August. As of September 25, it sat at 196,800 ounces. For context, US palladium imports from Russia alone were 27.6 million grams in 2024 — roughly 887,000 ounces. The exchange warehouse holds less than a quarter of one year's imports from a single country.
This isn't a disruption. It's a structural erosion that's been masked by drawing down the same stockpiles for five years running. The buffer is almost gone. When it runs out, the market doesn't gradually adjust — it re-prices in a day.
|
Part III
The Weak Link
|
Stillwater, Montana. The only primary palladium mine in the United States. The largest primary source outside Russia and South Africa. Four hundred and twenty United Steelworkers members walked off the job on September 3. They haven't gone back.
The dispute is over healthcare costs and incentive pay. The workers rejected three offers. Sibanye-Stillwater's CEO has warned about shutting down operations entirely. There have been no formal negotiations between management and the roughly 400 Nye-section miners since September 17. East Boulder, the smaller adjacent mine, reached a separate deal. The Stillwater East mine and the Columbus smelter — the complex that actually refines the metal — remain idle.
Sibanye-Stillwater's combined US PGM operations — the Stillwater and East Boulder mines — produce approximately 284,000 ounces a year. In a market running a 370,000-ounce deficit, that is not a footnote. It's a load-bearing wall. Every week this strike runs adds to a hole that recycling cannot fill and Norilsk's declining ore body cannot cover.
And here's where the gap widens into something genuinely dangerous. While the physical metal is being choked off — mine output declining, a primary producer on strike, above-ground stocks steadily draining — the managed money crowd on NYMEX has done the opposite. They read the EV narrative, they read the price weakness, and they shorted more. Bearish positioning in palladium futures and options has been heavily concentrated, though still below the record levels reached in 2024–2025. The last time shorts were squeezed from such extremes, palladium spiked 18% in a single session.
I've been on the wrong side of a PGM trade before. You wake up, the position is fine, you get coffee, and by the time you sit back down the market has moved three months' worth of range in forty minutes. Palladium's market is small. Illiquid. And when the shorts try to cover in a thin market against declining physical supply, the exit door gets very narrow very fast.
The specs see a metal that EVs will eventually kill. They're probably right — on a ten-year horizon. The physical market sees a metal that is running out right now. Those two timeframes are the whole trade.
Financial layer says surplus is coming. Physical layer says the deficit is deepening. I know how this resolves. It just takes longer than the position sizing usually allows.
|
Part IV
The Chain Reaction
|
If you've watched PGMs long enough, you know how the choreography works. Palladium did it in 2020. Rhodium did it more violently. The trigger changes. The mechanics never do.
Something cracks. Maybe the strike passes day 30, day 45, and the Stillwater smelter goes cold in a way that takes months to restart. Maybe an automaker — Toyota, Stellantis, someone with hybrid-heavy production — hits the spot market because their contract allocations come up short against a Norilsk that's delivering 14% fewer ounces. First wave: the managed money shorts, sitting at extreme levels, start covering. Price pushes up. Their stop-losses trigger. The algorithms that were aggressively short yesterday flip to buy.
Second wave — bigger, messier. Autocatalyst fabricators who assumed palladium was going to stay cheap suddenly realize the spot market has moved $200 while they were waiting for the dip. They stop waiting. Third wave: that's the one Bank of America's $2,200 target is built on. Whether it gets there depends on the strike and on Norilsk's Q3 numbers. But the direction of travel in a market with shrinking above-ground cover and heavily concentrated speculative shorts is not ambiguous.
Where does the capital go? Not into the broad PGM ETFs — those are weighted toward platinum, which has its own dynamics, or diluted with recyclers who benefit from volume, not price. The edge, if there is one, sits with mid-cap South African producers running unhedged palladium exposure — Valterra, whose refined output jumped 25% in H1 and whose profits track spot directly. Implats, where annual profit rose 31-fold on price and throughput gains. These are the levered plays on a physical market that is already telling you the deficit is real while the paper market pretends the EV transition has already happened.
On the other side, Sibanye-Stillwater itself is the highest-beta name. If the strike resolves and production restarts, the stock reprices on the restart plus the palladium price recovery. If the strike drags on, the mine's economics deteriorate further — but the palladium price goes up and every other producer benefits. It's a rare setup where the worst outcome for one company is the best outcome for the physical market.
Financial layer says palladium is dying. Physical layer says it's running a fifth consecutive deficit with thinning above-ground stocks and the only US mine on strike. In my experience, when the paper market and the physical market disagree this loudly, the physical market wins. It just takes longer than the quarterly earnings cycle wants it to. The loading dock doesn't care about your model.
