America's Aluminum Comeback Runs on Power It Doesn't Have

Aluminum closed at $3,287 a tonne on the LME this week — up 26% from a year ago, after touching roughly $3,520 during the worst of the Hormuz shipping crisis. LME warehouse stocks have drained toward their lowest levels this century. And that's the cheap price. An American manufacturer buying metal today pays the LME price plus a Midwest premium that hit a record $2,182 per tonne earlier this year — the first time in history the premium alone crossed $1 per pound. Buy a tonne of aluminum in Ohio and you're paying roughly two-thirds more than a buyer in Rotterdam for the identical metal.

Washington's answer has been tariffs: Section 232 duties on imported aluminum were hiked from 25% to 50% in June 2025, and the record premium is the direct result. The theory was that a wall of tariffs would make domestic smelting profitable again and the industry would rebuild itself behind it.

The theory is failing, and the reason is written into the physics of the metal itself.

Aluminum Is Congealed Electricity

Primary aluminum is made by running enormous electrical current through dissolved alumina — the Hall-Héroult process, essentially unchanged since 1886. Electricity is 30–40% of the cost of production. Producing one tonne takes roughly 14–15 megawatt-hours, which means a single world-scale smelter draws about as much power, continuously, as a mid-sized American city. Traders call the metal "congealed electricity" for a reason: an aluminum ingot is a battery you can stack on a pallet, and the country that smelts it is really exporting its surplus power.

That framing explains the entire history of the American industry. The U.S. built the world's largest smelting fleet — more than 30 plants at the peak — on cheap hydropower from the Pacific Northwest and the Tennessee Valley and cheap coal in the Ohio Valley. When American electricity stopped being cheap, the fleet died plant by plant. In February, Reuters reported the shutdown of yet another smelter, leaving the United States with just five primary aluminum plants — this in a country that consumes more aluminum than any other outside China.

The Reuters headline said it plainly: power trumps tariffs. A 50% duty cannot save a smelter that loses money on every megawatt-hour. The one bright spot proves the point — Century Aluminum's Mt. Holly plant in South Carolina is climbing back toward its 220,000-tonne capacity not because of tariffs, but because it extended a workable power deal with Santee Cooper, its local utility.

The Squeeze Has Three Jaws

What makes 2026 different from every previous aluminum cycle is that supply is capped almost everywhere at once:

• China is at its ceiling. Beijing's hard cap of 45 million tonnes of annual smelting capacity — imposed for energy and emissions reasons — has gone from theoretical to binding. The country that supplied virtually all of the world's production growth for two decades no longer can.
• The Gulf is disrupted. The Iran conflict and Strait of Hormuz shipping disruption have curtailed output and export flows from the Gulf smelters that were supposed to be the West's swing supply.
• The West can't add capacity. Europe's smelters never fully returned from the 2021–22 energy crisis, and CBAM carbon border fees are tightening the screws. America is down to five plants and a grid already straining under data center demand.

Demand, meanwhile, is not waiting. Every incremental gigawatt of grid buildout, every EV, every solar farm, every transmission line is aluminum-intensive. The metal is on the receiving end of the same electrification supercycle that is causing its own production shortage. That is the loop that matters: aluminum demand is being driven by the very force — electricity scarcity — that prevents new aluminum supply.

There is exactly one serious attempt underway to break the loop: in January, Emirates Global Aluminium and Century Aluminum signed a joint development agreement to build the first new U.S. primary smelter since 1980, in Inola, Oklahoma — a plant that would roughly double American primary production. Whether it gets built, what has to be true for it to work, and who makes money in every scenario along the way — that's where this gets actionable.

The rest of this briefing is for paid members: the data-center-versus-smelter power math that decides whether Inola ever gets built, the single contract milestone that is the real go/no-go signal for $CENX, the four names positioned to collect the record premium while supply stays broken, and the scenario zones — base case, squeeze extension, and the tariff-détente unwind — with the positioning framework for each.

AlphaBriefing Paid gets you every investment thesis, scenario framework, and catalyst brief we publish — the analysis private intel clients pay four figures for, at a fraction of that.

Unlock the full briefing →