|
Part I
The Mechanism
|
Aluminum.
Bloomberg ran the chyron last week: "Aluminum Premium Drops as US Plans to Cut Tariffs on Canadian Metal." CNBC had a guy explaining how the Bessent framework is "normalizing supply chains." The Midwest premium dropped 8.2% to 95 cents a pound and the market exhaled like the fire was out.
It's not out. They're staring at the price tag and ignoring the warehouse.
The United States has four operating aluminum smelters. Four. Down from 23 in 2000. Between them — two Century Aluminum, two Alcoa — they produce 683,500 tonnes a year. The country consumes roughly 5.5 million tonnes. That's 12% domestic coverage. The other 88% is imported — the majority overland from Canada — and the sources are running short.
The tariff cut on Canadian metal doesn't create a single new tonne of aluminum. Canada's smelters are already running near full capacity — 3.27 million tonnes of nameplate, almost entirely spoken for. What the tariff cut does is make existing Canadian flows marginally cheaper at the border. The market mistook a price discount for a supply fix. Those are different problems. One is arithmetic. The other requires a building that takes five to six years to construct.
I've watched the aluminum market celebrate tariff adjustments before. The relief tends to last about as long as it takes for the next warehouse report to land.
|
Part II
The Diagram
|
Story off. Numbers on.
LME aluminum warehouse stocks hit 244,100 tonnes on September 11. That's not just low — it's the 1st percentile of the exchange's entire recorded history. The drawdown has been relentless: stocks fell 35.8% in H1 2026 alone, from 498,670 tonnes in January to 320,314 by June, and kept falling through summer.
Three structural valves closed simultaneously this year.
First: Iranian missile and drone attacks hit EGA's Al Taweelah smelter in Abu Dhabi and Alba's facility in Bahrain on March 28. Combined annual capacity of the two smelters: 3.2 million tonnes. The Gulf region accounts for roughly 8–9% of total global aluminum supply. EGA's first cell restarted May 26. Three hundred fifteen cells are back online. Full production isn't expected until early 2027.
Second: US and UK sanctions barred Russian aluminum from Western warehouses. Rusal has capacity for 4.3 million tonnes a year but produces about 3.9 million. That metal now routes to China and Southeast Asia. It doesn't enter the LME system anymore.
Third: China hit its self-imposed 45-million-tonne capacity cap. National output ticked down from 11.1 million tonnes in Q4 2025 to 11.0 million in Q1 2026. There is no Chinese production surge coming. The ceiling is bolted shut.
Wood Mackenzie revised the 2026 global deficit to 900,000 tonnes. J.P. Morgan models a 1.7-million-tonne "invisible" deficit slowly bleeding through the system. Neither number includes a full-year accounting of Gulf disruption because nobody knows when Al Taweelah reaches pre-attack output.
Building a primary aluminum smelter takes five to six years and costs $4–6 billion. There is no quick fix here. There is only the math, and the math is ugly.
|
Part III
The Weak Link
|
The Midwest premium. That's the part of the machine everyone just stopped watching.
In February it broke $1 per pound for the first time in history — $2,182 per tonne. The Canadian tariff cut announcement in August dropped it 8.2% to 95 cents. October and November contracts fell more than 12%. The market declared the problem solved and moved on to the next chyron.
Ninety-five cents a pound is not relief. It is three times the European premium. The market moved from catastrophic to merely extreme and called it a win. I've seen this kind of celebration before — usually about two weeks before the next leg up.
The relief itself rests on a fragile assumption: that Canada can increase supply. It can't. US aluminum imports from Canada fell to 945,165 tonnes in January–May 2026, down from 1.23 million in the same period of 2025. Canadian smelters aren't adding capacity. They're already running flat out.
The EGA-Century joint venture announced in January will build the first new US smelter since 1980. In Oklahoma. Construction starts end of 2026. Production begins "by the end of the decade." I've watched enough smelter projects to know that "end of decade" means 2031 if everything goes right. It's like telling someone dying of thirst that you've started digging a well.
And the Strait of Hormuz is still effectively closed — traffic down 95% since Iran declared it shut in March. Even as EGA restarts cells at Al Taweelah, the metal has to get out of the Gulf. They've set up alternative export routes, but those routes are slower, more expensive, and capacity-constrained. Restarting the smelter without reopening the strait is like fixing the engine but leaving the wheels off.
Every tonne of Canadian aluminum already has a buyer. The tariff cut doesn't conjure new metal. It just changes who gets the invoice. And the invoice, even after the discount, is still three times what anyone budgeted for.
|
Part IV
The Chain Reaction
|
If physical tightness reasserts — and every drawdown trajectory on the LME says it will — here's the sequence.
LME stocks are draining at roughly 10,000–15,000 tonnes per month. At that rate, they breach 200,000 tonnes before year-end. Below that threshold, the exchange stops functioning as a liquid market and starts functioning as a rationing mechanism. Buyers can't wait for the premium to cool. They bid whatever it takes to secure physical delivery.
First move: backwardation hardens. Cash aluminum on the LME already trades above the three-month forward — $3,274 versus $3,250 as of September 11. That inversion deepens as available metal disappears. Second move: the MW premium, which just dropped on tariff optimism, snaps back above $1/lb as physical buyers realize cheaper Canadian tariffs don't solve a 900,000-tonne global deficit. Third move — and this is where the money is — domestic producers reprice.
Century Aluminum and Alcoa are the only primary aluminum producers in the United States. Between them, they control 100% of domestic smelter output. Century's Mt Holly reached full capacity in July. Century's Grundartangi in Iceland restarted in April, ahead of schedule. The Oklahoma JV won't produce metal this cycle. But the existing four smelters? Every tonne they pour sells into a market where the delivered premium alone is approaching the historical all-in price of the metal itself.
The edge isn't in broad metals ETFs. It's not in Canadian smelters — they're already maxed and their margins are capped by energy costs. It's in the companies pouring metal into a market where 88% of demand is met by imports and the import machine just lost three of its valves. Century and Alcoa are the toll booths on the only open road.
The physical market doesn't care about tariff sentiment. It doesn't care about chyrons. It cares about tonnes. And the tonnes aren't there.
