Century Aluminum Company CENX

Key data

The business

Century Aluminum makes primary aluminum, the raw metal poured into ingots and billets that other companies roll into sheet, extrude into shapes, or cast into parts. It runs three smelters that matter: Grundartangi in Iceland, which runs on cheap geothermal and hydro power and is the low-cost jewel, Sebree in Kentucky, and Mt. Holly in South Carolina. It closed and sold the Hawesville plant, once its largest, and keeps a carbon-anode operation in the Netherlands. Power is the whole game here, because smelting is electricity turned into metal, and Century's edge, such as it is, comes down to the Icelandic power contract and being the largest domestic pourer at a moment when the country wants aluminum made at home. There is no moat in the durable sense. Century is a price-taker: it sells at the exchange price plus a regional premium it does not set, and its profit is the thin gap between that price and its power-heavy cost.

The numbers

The story is a spread that blew open. A year ago Century barely cleared its costs; by the June quarter it was keeping thirty cents of every sales dollar as gross profit.

Revenue rose about 20% against a year ago, but revenue is the least of it. The inflection quarter is Q2 2026, where the metal price plus the Midwest premium plus the Section 45X credit inside cost of sales combined to triple the margin. One caution sits in Q1: most of that quarter's profit was a $287.9 million gain on the sold Hawesville plant, not aluminum, and Century's near-zero tax rate lets such a gain fall straight to the bottom line. The watch a careful reader would have set a week ago, that shipped tonnes climb back above 160,000 while the realised price sinks under $4,000, did not come to pass: the tonnes met a stronger price, not a weaker one, and the spread widened instead.

Stretch the lens across the cycle and the compounding is a flat line with spikes on it.

Revenue was about $2.2 billion in 2021 and about $2.5 billion in 2025, a rounding error of growth across a full cycle, while earnings ran from a large loss to a profit to almost nothing and now to a record. That is the signature of a price-taker, not a compounder. Diluted shares grew about 4% over three years, so the per-share line leans entirely on the metal price and the policy stacked on top of it, never on the business getting structurally better. Smelters are the classic value trap, cheapest at the top of the cycle, and this cycle's top has a policy stamp on it. The market's read, an 8 multiple on a peak, is close to right on the arithmetic and possibly wrong on the durability: if the tariff and the credit make this margin a plateau rather than a spike, the earnings are more repeatable than a cyclical multiple assumes.

Management

The record here is written in exits. Over twelve months insiders sold about $626 million of stock and bought none, and the bulk is Glencore walking out the door: two block sales of $272 million and $327 million that read as a large holder unwinding a position, not a bet on the year ahead. Chief executive Jesse Gary sold $8.3 million of his own in March, discretionary, not a scheduled plan disclosed as such. His pay ran $7.2 million in 2025, about 18% of net income, heavy for a company whose profit is set by the metal price. The one buyback, a token $4.7 million, was at least bought cheap, near $23. No dividend. The people who know the business best are selling into the record.

How it fails or surprises you

Policy is the margin. The 30% gross margin rests on two things Century does not control: the tariff-fed premium US metal earns over the exchange price, and the Section 45X cash credit buried in cost of sales, worth tens of millions a quarter. Either can change with a signature. The first print to watch is the Midwest premium, which rolls over before the margin does.

The gain that flattered the half. Half the first-half profit was not from making aluminum: a $287.9 million gain on Hawesville sits inside it and falls through nearly untaxed. Strip it and the record shrinks by more than a third. The fact that argues against a clean bull read is whether Q2's operating margin, gain-free, is a run rate or a top. The print: Q3 operating income holding near Q2's $211.6 million.

A walled home market with metal coming (right tail). If the tariff and credit hold for years, not quarters, Century's plan to build the first new US smelter in decades stops being a cost and becomes an option on a protected, premium-priced market rivals cannot cheaply enter. The market pays nothing for this today. The first sign is a funded decision on that smelter, or the premium holding above its five-year high through a demand soft patch.

Closing thoughts

This one gets settled by a number, and the number comes in November. The market pays about 8 times a record year and no more, which says it treats the profit as a peak that fades, and the sellers agree: Glencore has walked out of nearly $600 million of stock and the chief executive sold in March. The print that settles it is the third-quarter gross margin. If it holds near 30% without a one-time gain propping it, the spread is structural, made of tariff and credit that Washington shows no sign of pulling, and 8 times is too cheap. If it slides back toward the teens, the sellers were right and this was a spike. The fatter tail is down, because the thing holding the margin up is a policy choice and policy choices reverse faster than smelters retool, but the loss there is to the multiple and the peak earnings, not to solvency, which the company's light net debt protects.

The bet is still that the country keeps imports expensive and keeps paying Century to pour metal at home, and that its Icelandic power stays cheap. What breaks it is a pen: a softer tariff or a trimmed production credit, either of which erases the premium and the subsidy in the same stroke. Watch two numbers together, the Midwest premium and the gross margin, because when the first rolls over the second follows within a quarter, and the record earnings unwind as fast as they arrived.