Former Aluminum Smelters: Built for Power, Ready for Compute

Smelters typically draw power measured in the hundreds of megawatts — an order of magnitude beyond most industrial facilities. That means the substations, transmission ties, and often dedicated generation contracts serving the site were sized for genuinely enormous, continuous loads: precisely the load profile an AI training cluster needs.

Many smelter sites were also located near hydroelectric or other low-cost generation specifically because of the aluminum industry's power sensitivity — a siting logic that maps closely onto where compute buyers want to be today.

Beyond the power position, former smelter sites commonly offer:

Smelter sites almost always carry legacy environmental considerations tied to decades of industrial processing — this is well understood territory, not a dealbreaker, and it's factored directly into site diligence and deal structure rather than discovered as a surprise.

The buildings themselves may need significant retrofit for data hall use (insulation, containment, fire suppression), but the electrical backbone and structural shell are frequently the hardest and most expensive parts of a data center build — and here, they already exist.

The aluminum industry has historically sited primary smelting near the cheapest, most reliable power available — which in North America means a strong concentration around Pacific Northwest hydroelectric capacity (Columbia and Snake River systems) and Quebec's hydro grid, with older facilities scattered across other hydro- and coal-adjacent regions. That siting logic is decades old, well documented in industry and energy-policy history, and it's exactly the logic that makes these particular sites relevant again now.

Many of these plants curtailed or closed as global aluminum production shifted and power costs changed — the site and its power position often outlasted the economics that originally justified running a smelter there.

Smelter sites almost always carry legacy environmental considerations tied to decades of electrolytic processing — spent potlining and fluoride byproducts are well-understood contaminants in this industry, not a surprise finding. Expect environmental diligence to run longer here than on a typical industrial site; that's priced into how we structure a deal from the start, not discovered midway through.

Structural retrofit of the potroom buildings (insulation, containment, fire suppression for data hall use) is a real, budgetable workstream — but the electrical backbone and structural shell, historically the most expensive and time-consuming parts of any data center build, are the parts that already exist here.

If you own or represent a closed or closing smelter site, the power infrastructure that made it a smelter is likely the same infrastructure that makes it valuable for AI compute. Tell us about the site's location, remaining power contracts or interconnection rights, and current status, and we'll evaluate it against real buyer demand and financing.