A Tariff Can Raise a Price. It Cannot Write a Power Contract.

The United States now taxes imported aluminium at fifty percent. The rate went from ten percent to twenty five in February of last year and to fifty in June, all country exemptions were removed, and talks that would have given Canadian metal a lower rate collapsed in August. The Midwest premium, which is what an American buyer pays above the exchange price to have metal delivered, passed a dollar a pound for the first time in January.

The policy worked exactly as designed on the price. It has done almost nothing to the productive base it was meant to protect. Three companies operated six primary smelters in five states last year, of which four were actually producing, two of those at reduced rates. Combined capacity is 1.31 million tonnes a year and output was about 660,000 tonnes. In 2000 the country had twenty three reduction plants and 4.27 million tonnes of capacity. In 1980 it had roughly thirty three sites producing 4.65 million tonnes. Net import reliance has gone from a third to sixty percent across that span.

The Historical Echo

The industry's own history contains the cleanest case anywhere of smelting capacity leaving a country on a power shock, and of what happens afterwards.

The Pacific Northwest once hosted ten aluminium smelters with about 1.64 million tonnes a year of capacity, roughly forty percent of all American smelting, employing some eleven thousand people and drawing 3,150 average megawatts of hydroelectricity. During the Western energy crisis of 2000 and 2001, wholesale power prices rose by factors of ten and twenty and more. The Bonneville Power Administration needed load relief and paid the smelters around twenty dollars a megawatt hour for electricity they agreed not to use. Selling the power became worth more than turning it into metal.

The region shut in weeks. Columbia Falls in Montana, 84,000 tonnes, at the end of January 2001. Wenatchee, 80,000 tonnes. Intalco at Ferndale, 140,000 tonnes, in stages through the spring. Longview, 204,000 tonnes, at the end of February. By the middle of 2001 something like 25,000 tonnes a year of the region's 1.64 million remained open.

The crisis passed within two years. The smelters did not come back. The last two were idled in 2016 and Intalco was closed permanently in March 2023. A frozen potline costs close to new construction money to restart, and nobody commits that capital without a power price they can see for a decade.

Where Patient Capital Is Positioning

The 2026 evidence is that the constraint has not changed, and it is visible in what does and does not get signed.

Electricity is roughly forty percent of a smelter's production cost. The industry's own stated viability threshold is around forty dollars a megawatt hour locked for ten years. American industrial rates run from the mid sixties to the low eighties. Canadian smelters pay between the high twenties and the low forties. That spread, not the tariff, is the whole economic story, and it is why Canada has supplied well over half of American aluminium imports for years.

Where a power contract exists, capacity moves. Century restarted expanded production at Mt. Holly in South Carolina in April, adding about a tenth to American primary output. Alcoa secured a ten year, 240 megawatt contract at Massena in New York beginning in April and committed about sixty million dollars of capital alongside it.

Where it does not, nothing moves. The proposed Oklahoma smelter, a four billion dollar joint venture that would more than double American primary capacity and be the first new plant in roughly half a century, has been negotiating its electricity supply for close to a year without signing. It needs over eleven terawatt hours a year, the consumption of a mid sized city, at a moment when data centres are bidding for the same electrons.

The lesson for long horizon capital is a distinction worth holding onto whenever policy moves a market. A tariff changes a price, and a price is a transfer between parties. It does not change a cost, and cost is what determines whether a physical thing gets built. The Midwest premium can go to a dollar and a half and it will still not produce a ten year contract at forty dollars a megawatt hour. Assets that already own their energy input, whether that is generation, a mineral interest or productive land with water on it, are positioned for a world where the electricity is the scarce part. Assets that must buy it at whatever the grid charges in 2031 are not.

The Capital Memo