Tearing apart the aluminium warehouse drain

Aluminium.

The headlines are busy with demand concerns. Bloomberg ran a piece last week about Chinese property weakness dragging base metals lower. CNBC had someone in a blazer explaining how "easing trade tensions could normalize premiums." Classic dashboard watching.

Meanwhile, LME aluminium warehouse stocks just fell to 259,400 tonnes. That is the lowest level since the exchange started keeping records in 1998. Less than one day of global consumption is sitting in the entire LME warehouse system. And here's the part nobody's putting on the chyron: 95% of what's left is Russian metal.

Half the Western market won't touch Russian-origin aluminium. It's technically eligible for trading — metal produced before the April 2024 cutoff still carries a valid warrant. But try convincing a European auto OEM's compliance department to sign off on Russian-branded ingots in 2026. Self-sanctioning has turned a 259,000-tonne headline into something closer to 13,000 tonnes of metal the market will actually use.

That's not a buffer. That's a rounding error.

Stories off. Let's open this up.

Three supply valves closed at roughly the same time.

First, the Strait of Hormuz. Gulf Cooperation Council smelters produce 6.16 million tonnes of primary aluminium a year — 8.35% of global output. When Hormuz effectively shut in late February, it didn't just block exports. It blocked inbound alumina shipments. Alumina imports into the Middle East fell 63% year-on-year. You can't make aluminium without alumina. That's not a trade disruption. That's feedstock starvation.

Second, the physical damage. Iranian missile and drone strikes hit EGA's Al Taweelah smelter on March 28. That facility produced 1.6 million tonnes in 2025. As of early July, only 89 of the smelter's 1,262 reduction cells had been restarted. Full recovery: up to 12 months. Reduction cells can't be flipped back on — they have to be relined, preheated, and brought to operating temperature one by one. Metallurgical surgery, not a light switch.

Alba in Bahrain declared force majeure and shut 19% of capacity — lines 1, 2, and 3 of a 1.62-million-tonne-per-year operation.

Third, China's ceiling. Beijing's 45-million-tonne production cap is now operationally binding. Chinese smelters ran at roughly 98.2% of nameplate capacity in 2025, producing 45.02 million tonnes. There's no room left. New output has to come from Indonesia, India, or the Middle East — and right now two of those three are either building or burning.

Output outside China fell 6.7% year-on-year in July.

Wood Mackenzie initially modeled a 0.4% decline in global output for 2026. After the direct strikes on Al Taweelah and Alba, they revised their deficit estimate to approximately 900,000 tonnes. That's not a revision. That's a different market.

Here's the part that keeps me staring at this screen past midnight.

The CFTC numbers show speculative net positions on aluminium went from +1.3K contracts on July 10 to −0.7K contracts by July 31. In three weeks, the paper market flipped from net long to net short. The machines read the EGA restart headlines — 89 pots restarted! progress! — and did what machines do. They sold.

But the Midwest Premium tells the real story. It hit $2,182 per tonne — a record. That premium now represents more than 40% of the total cost of aluminium in the United States. Historically, it was a rounding error on an LME invoice. Today it's the price of an entirely broken logistics chain.

The 50% Section 232 tariff on imports created the initial wedge. The Hormuz shutdown blew it wide open. American manufacturers buying physical aluminium are paying an all-in cost above $5,340 per tonne while the LME screen says $3,336. That's a $2,000 gap between what the terminal shows and what the loading dock charges.

And the war-risk insurance layer is the one nobody's modeling. Premiums for Hormuz transit went from 0.25% of hull value before February to 3–10% by mid-July. On a $100 million vessel, that's the difference between a $250,000 insurance bill and one exceeding $7 million. For a single crossing. That cost doesn't appear on any aluminium futures contract. It shows up as a wider spread, a longer lead time, and a counterparty who suddenly wants different payment terms.

The European supply picture isn't helping. Fifty percent of EU aluminium smelting capacity has been curtailed since the 2022 energy crisis — over 800,000 tonnes offline. Slovalco in Slovakia just announced a restart, but it's only 75,000 of 175,000 tonnes capacity, and production won't begin until Q4 2026.

That's the equivalent of bringing a garden hose to a refinery fire.

The sequence from here is mechanical.

If the Al Taweelah recovery timeline slips — and smelter restarts almost always slip — the deficit widens. Wood Mackenzie's 900,000-tonne estimate already assumes partial recovery. A slower restart pushes the LME toward what I'd call functional zero: stock levels where every tonne has an owner, a warrant, and a destination. No float. No buffer. No liquidity cushion.

When that happens, the LME stops being a market and starts being a queue.

The Midwest Premium decouples further. American manufacturers — auto, aerospace, packaging — face a choice: absorb the cost, pass it through, or substitute materials. Back in 2018, Ford estimated $200 to $300 in additional material cost per vehicle from aluminium tariffs alone — and that was before the Hormuz shutdown compounded the problem. At some point, engineers start looking at steel and composites for parts that have been aluminium for two decades. Material substitution is slow to start and very hard to reverse.

Where does the capital go? Not into the broad mining ETFs. Not into the Gulf producers — their assets are the ones that got hit. The edge, if there is one, sits with domestic smelters that have secured power contracts and unhedged spot exposure. Companies whose revenue rises dollar-for-dollar with the Midwest Premium. Alcoa's US operations. Century Aluminum's Oklahoma joint venture. These are the levered plays on a physical market already screaming while the paper market mutters about demand concerns.

Physical says scarcity. Paper says sell. I've watched this movie enough times to know which reel runs ahead. The physical layer wins. It just takes longer than your risk book wants it to.