Idled aluminium smelters totalling roughly 150,000 tonnes per year of Western capacity are being lined up for restart in 2026 to absorb a Middle East supply shock, with the Middle East still providing about 23% of ex-China primary output [S1][S3]. Goldman Sachs has flagged a price scenario of $3,600/tonne if regional production is lost for one month, against a 2026 intraday high near $3,418/tonne [S7].
The decisive variable is electricity. Industry guidance places competitive smelter power cost at roughly $40/MWh on a 10–20 year fixed contract, while AI data-centre buyers are currently signing 10–20 year deals above $115/MWh [S4]. Electricity alone accounts for 30–40% of smelter cash cost, so the restart question is really a power-contract question [S5].
Why power, not ore, gates every restart decision
An aluminium smelter is functionally an electrolysis plant: roughly 13–15 kWh of electricity is consumed for every kilogram of metal produced, which is why a long-dated power agreement is the first item on any restart checklist [S5]. ING's December 2025 note put the economic threshold at a 10–20 year power contract around $40/MWh, and noted that competing AI data-centre buyers are committing upwards of $115/MWh for equivalent long-term supply, an effective premium that crowds smelters out of the grid [S4].
Europe illustrates the inertia. More than 1 million tonnes per year of European aluminium capacity was taken offline after the 2022 energy shock, and around 800,000 tonnes per year remains offline entering 2026 with no announced restarts, because gas-indexed power prices have not returned to contract levels that cell-room economics can absorb [S4]. For comparison, restarting a 75,000 tonnes-per-year potline is the scale of the Western restarts now in motion, with restart lead-times measured in quarters, not weeks [S3].
Restart pipeline: confirmed, conditional, and at risk
Confirmed 2026 restart: Century Aluminium is bringing back about 50,000 tonnes per year at its Mount Holly smelter by Q2 2026 after securing power through 2031, the only concrete US restart on the public ledger [S4]. Conditional restarts: Reuters and Mining Weekly report a 75,000 tonnes-per-year potline being prepared for restart on the back of a supply shock traced to the Iran conflict, with the goal of displacing lost Middle East tonnage [S3][S6].
At-risk operations concentrate wherever power contracts expire. South32's 560,000 tonnes-per-year Mozal smelter in Mozambique could shut in March 2026 if affordable electricity is not secured beyond that date [S4]. Rio Tinto's 586,000 tonnes-per-year Tomago smelter in Australia is exposed from 2029 when its current agreement lapses in December 2028, and the 192,000 tonnes-per-year Bell Bay smelter has only secured a Tasmanian government-backed extension to December 2026 [S4]. On top of that, an October 2025 electrical-equipment failure at Century's Iceland Nordural 320,000 tonnes-per-year smelter has idled two-thirds of output, with an 11–12 month restart timeline that tightens the 2026 balance further [S4].
China's 45 million tonne ceiling shifts growth overseas
China's electrolytic aluminium capacity is administratively capped at 45 million tonnes, a policy set in 2017 to curb oversupply and emissions, and operating capacity is approaching 44 million tonnes by early 2026 at near-technical-limit utilisation [S4][S5]. Net Chinese exports are down roughly 700,000 tonnes year-to-date because domestic output is structurally flat, and any new exemption is being discussed only for renewable-powered smelters, of which a growing minority now qualify [S4].
With domestic expansion locked, Chinese firms are pivoting to overseas projects, and Indonesia is the standout destination: Indonesian aluminium exports are up 56% year-to-date on the ramp of coal- and bauxite-advantaged projects with low production cost [S4]. ING's framing is that ex-China supply is now driven by where 10–20 year power at ~$40/MWh can still be contracted, not by where bauxite is cheapest, which is why Indonesia is pulling capacity investment ahead of the Middle East on a forward basis [S4].
Restart economics: a four-criteria comparison
Across the restart candidates and at-risk assets, four criteria decide whether a potline runs in 2026: long-term power price relative to the $40/MWh benchmark, contract length against the 10–20 year requirement, grid reliability after recent electrical-equipment incidents, and the geopolitical exposure of the host country. The table below lines up the main categories of supply against those criteria. [S4]
US restart (Mount Holly) scores well on contract length (through 2031) and host-country stability, but is small at ~50,000 tonnes per year and exposed to grid-demand pressure from data-centre build-out [S2][S4]. European idle capacity (~800,000 tonnes per year) is large in theory but blocked on contract price, because European wholesale power remains indexed to gas rather than locked at long-dated fixed tariffs [S4]. Indonesian greenfield supply is the lowest-cost new tonnes, but is being built on coal-heavy power and is therefore exposed to future carbon-border adjustment, including the EU CBAM regime referenced across 2026 metals coverage [S4][S5]. At-risk incumbent supply (Mozal, Tomago, Bell Bay) is large in aggregate (over 1.3 million tonnes per year) but binary: it either renews a power contract or idles within the contract window [S4].
What this means for downstream buyers and engineering specs
For procurement and specification engineers, the 2026 takeaway is structural, not cyclical: aluminium billet, particularly 6xxx-series extrusion stock, is tightening as a strategic resource, and the premium for spot tonnes is widening against the LME benchmark because physical availability, not the futures curve, is the binding constraint [S5]. Scrap supply is also being reshaped, with XRT and LIBS sorting enabling alloy-specific aluminium scrap streams that partially offset primary shortages, a useful lever for downstream buyers who can qualify secondary material to their alloy spec [S5].
On the power side, the practical effect of competing with data-centre loads is that any new long-dated industrial power contract is now negotiated against hyperscaler benchmarks north of $100/MWh, which is why smelter restarts are clustering at sites where the host utility, state government, or an anchor industrial buyer has already absorbed the political cost of locking price for a decade or more [S4]. Watch the next two signals: whether Tomago secures a post-2028 contract in 2026, and whether Bell Bay's December 2026 extension is renewed, because together they cover nearly 800,000 tonnes per year of Pacific capacity and will set the regional benchmark for the next round of restart talks [S4].
For the relevant spec sheets and selection criteria, see power supply, dc power supply, and switching power supply.
Background reading: Cement Sector Signals: Demand-Side Brand Innovation, Eastern Europe M&A, and Multifunct....