Century Aluminum Company (NASDAQ: CENX) — Scarcity Rents Funding a Very Expensive Future

Report date: 30 August 2026
Reference price: $46.59 at 28 August 2026 close
Primary evidence cutoff: 30 August 2026
Scope: Century Aluminum Company, including its controlled Jamalco interest and proposed 40% interest in Oklahoma Primary Aluminum

The evidence-led analysis below is position-free and contains no recommendation or price target. The opening opinion is the sole exception.

⚡ Claude’s Take

The author’s independent subjective opinion; general information only, not investment advice. The evidence-led analysis below carries no position.

Verdict: AVOID-HERE / HOLD for existing owners; accumulate only on weakness in a $28–$36 zone. That zone corresponds broadly to 6–6.5× a $500–$600 million through-cycle EBITDA range, less current net debt and noncontrolling interest—not to capitalizing the Q3 guide as a perpetuity. CENX is not a short: the balance sheet is liquid, all three smelters are returning toward full output, and a still-unapproved U.S. onshoring plan could add roughly $244 million of annual cash economics at management’s illustrative metal price. But at $46.59, the equity asks the investor to pay before Commerce has allocated that benefit, before Oklahoma has a fixed power contract or final investment decision, and while the Midwest premium is at an exceptional, tariff-created level.

This is a policy-assisted commodity momentum setup, not a quality compounder. The All Factors model shows a 1.26 market beta, positive Materials/Mining/Oil exposures, negative Low Volatility and Liquidity loadings, and only 39.7% explained variance. The shares rose more than fourfold from April 2025 to June 2026, then fell 34% from their intraday high. The tape is no longer one-way, but history says CENX can still lose 75–85% when the aluminum/power spread turns. I would demand a price that works on normalized existing-asset cash flow and treat the import concession, data-center stake, and Oklahoma project as options rather than base value.

Conviction: medium. The bullish flip would be a binding Commerce approval plus a fully financed Oklahoma plan that caps Century’s equity contribution while preserving at least $500–$600 million of normalized existing-asset EBITDA. The bearish flip would be a Midwest-premium reversal or policy change while Century commits more than $1 billion of its own capital to Oklahoma.

The memorable version: current scarcity rents are real; their duration and destination are not.

📈 Stock Price Action — Five-Year Event Map

From 30 August 2021 through 28 August 2026, CENX rose from a $12.39 close to $46.59 (+276%), but the path included an 82% peak-to-trough collapse. The trailing 52-week intraday range was $20.91–$70.43, and the latest close was 33.9% below the 2 June 2026 high. Price moves below are facts from the AZI series; causes are labeled interpretations and cross-checked against company filings and public policy actions.

1. Century reported that favorable LME/premiums added $640.9 million to 2021 sales, partly offset by $251.1 million of worse power. Russia/Ukraine and European power fears likely extended an aluminum rally already visible in Century’s realization.
2. On 22 June 2022 Century announced Hawesville’s temporary idling because energy prices had risen to more than three times their historical average. The 82.6% collapse is direct evidence that replacement cost does not protect the equity when power and metal spreads invert.
3. On 19 December 2023 Century estimated $55–$60 million of initial annual 45X benefit and another potential $45–$50 million if material costs qualified; DOE then selected it for negotiations for up to $500 million. The price move repriced policy optionality before the eventual Jamalco accounting gain.
4. Q3 2024 adjusted EBITDA rose to $103.7 million from $34.2 million sequentially, including $34.9 million of expanded 45X benefits and stronger LME/premiums. The one-day move shows the equity’s sensitivity to changes in the perceived profit floor.
5. The U.S. announced that Section 232 aluminum duties would double from 25% to 50% effective 4 June 2025. For a domestic producer in an import-dependent market, the policy directly increased scarcity value.
6. Century’s Q3 2025 print guided Q4 adjusted EBITDA to $170–$180 million; days later Glencore sold nine million shares at $30.25. The rapid rise/reversal combined earnings acceleration with a large strategic-holder supply event.
7. The six-month surge incorporated higher LME/Midwest premium, the 750 kt EGA/Century venture, $200 million Hawesville sale and retained interest, Mt. Holly/Grundartangi restarts and Gulf supply disruption. No single event explains a 141% move.
8. Management said limited reopening of the Strait of Hormuz returned LME/premiums toward pre-conflict levels. The reversal despite $326.9 million of Q2 adjusted EBITDA is consistent with normalization overwhelming a strong historical print.

1. Executive Summary

Century is the largest U.S. primary-aluminum producer and a compact, unusually leveraged expression of three spreads: the LME aluminum price over alumina/carbon/power costs; the U.S. Midwest premium over offshore delivery economics; and public-policy benefits over the capital required to keep old smelters operating and build a new one. It owns three operating smelters—Grundartangi in Iceland, Sebree in Kentucky and Mt. Holly in South Carolina—with 770,000 tonnes of annual capacity. It also controls 55% of Jamalco, whose 1.4 million-tonne alumina refinery gives Century a 770,000-tonne attributable share, and owns the Vlissingen anode plant. Those assets make the value chain understandable. They do not make revenue recurring or price-setting.

The current earnings print is exceptional. Q2 2026 shipments were 130,632 tonnes, sales were $752.1 million and company-defined adjusted EBITDA attributable to Century was $326.9 million. Management guided Q3 to $325–$345 million. H1 operating cash flow was $236.0 million and capital expenditure $134.4 million, implying conventional free cash flow of $101.6 million before the July receipt of a $94.3 million 45X refund and additional insurance proceeds. Cash was $343.4 million at June; by July, management said cash exceeded total debt. The balance sheet therefore has more protection than it did in prior cycles.

The quality problem is composition. Average H1 2026 LME, Midwest premium and European duty-paid premium were approximately $3,386, $2,406 and $488 per tonne, versus $2,538, $855 and $241 in H1 2025. The U.S. premium increased much faster than underlying metal. H1 45X reduced cost of goods sold by $51.3 million, with an additional small SG&A benefit. Grundartangi results included transformer-failure costs and insurance recoveries. Q1 GAAP income included a $287.9 million gain on Hawesville’s sale. In other words, strong cash earnings are real, but GAAP income is not a clean proxy and current adjusted EBITDA is partly a policy-and-premium windfall.

The five-year record is sobering. Company-reported adjusted EBITDA was $174 million in 2021, $144 million in 2022, $120 million in 2023, $244 million in 2024 and $425 million in 2025—an average of about $221 million—before reaching $558 million in H1 2026. From 2020 through 2023, cumulative conventional free cash flow was approximately negative $168 million; 2024 was also negative because $24.6 million of operating cash use was followed by $82.3 million of capex. Only 2025 returned to positive conventional free cash flow, at about $85 million. This is not a business that historically converted accounting asset value into steady distributable cash.

Century has no durable pricing-power moat. Its smelters are difficult and expensive to replicate, electricity supply and permitting are formidable barriers, and Grundartangi’s renewable power plus U.S. local supply are useful positioning advantages. Yet aluminum is priced globally, product premiums are contestable, customers can multi-source, and Glencore both owns roughly 30% of Century and represented 45.5% of H1 2026 sales. High barriers protect the industry’s incumbents from new Western entry; they do not let Century set the LME or eliminate foreign, state-backed and recycled supply.

The capital cycle is favorable for existing ex-China smelting but ambiguous for the equity. U.S. primary production was only about 660,000 tonnes in 2025 against 5.7 million tonnes of apparent consumption, and roughly 60% net import reliance. A new 750,000-tonne smelter would more than double current domestic production. But the Aluminum Association estimates a new plant requires $4–$6 billion and roughly 11 TWh of annual electricity. EGA previously estimated around $4 billion for a smaller Oklahoma design. Century’s 40% share therefore points to $1.6–$2.4 billion before grants and incentives—large beside a $4.55 billion market capitalization and far larger than normalized annual free cash flow.

The July 2026 onshoring proclamation may bridge that funding gap. An approved plan can import an annual quantity commensurate with future output at half the otherwise applicable Section 232 rate. Century management illustrated approximately $244 million of annual economics for its 300,000-tonne share using $3,250/t LME and a 25-point tariff saving. That math is directionally correct. It is not yet an entitlement: Commerce determines approval, reasonable costs/output, benefit allocation and enforcement; construction must start by January 2029; benefits may be rescinded, including retroactively for fraud or deliberate misrepresentation.

At $46.59, ROIC.ai calculated a $4.55 billion market capitalization and $4.82 billion enterprise value using $343 million of cash, $480 million of debt and $131 million of noncontrolling interest. That equals only 3.6× the annualized midpoint of Q3 guidance but roughly 8.0× a $600 million normalized case and almost 22× the 2021–2025 average. The market is therefore not simply pricing peak EBITDA forever; it appears to price a substantial probability that tariffs, 45X, full-capacity production and Oklahoma-linked import economics sustain a much higher floor than history.

Central analytical verdict: Century is financially stronger, operationally fuller and more strategically relevant than it was at the 2022 trough. The industry still does not confer a moat, present earnings materially exceed an ordinary-cycle level, and Oklahoma converts today’s windfall into tomorrow’s financing/execution obligation. The decisive variable is not whether Q3 EBITDA is strong—it is whether Century can lock in the policy rent and power contract without turning shareholders into the residual funder of a $4–$6 billion project.

2. Business Overview

The asset system

Century is a holding company whose economic engine is a short chain from bauxite/alumina through carbon anodes and electric smelting to standard-grade and value-added primary aluminum. At 31 December 2025 the operating smelter system comprised:

The 2025 production shortfall versus capacity was not hidden slack of uniform quality. Sebree was near full. Mt. Holly was still being restarted. Grundartangi suffered a second transformer failure in October 2025 after a November 2024 failure; Line 2 stopped and output fell by roughly two-thirds until its April 2026 restart. By Q2 2026 the last 90 Mt. Holly pots were producing and Grundartangi Line 2 was near full output, though transformer replacement and production stability work continued.

How revenue and margin work

Century does not sell aluminum at a negotiated cost-plus margin. Contract prices typically combine the LME primary-aluminum benchmark, a regional delivery premium—Midwest in the U.S. and European duty-paid in Europe—and a product premium for billet, foundry alloy, molten metal or other value-added form. Alumina is sold at an alumina price index plus differentials. Revenue therefore changes with both shipment volume and benchmarks that Century cannot control.

The principal conversion costs are alumina, electricity, carbon products and labor. The 2026 Q2 filing says power, alumina, carbon and other operating inputs represented more than 82% of cost of goods sold. Smelting is continuous and power-intensive: a potline interruption can freeze production, damage equipment and require months of restart spending. This operating leverage explains why a $1,000/t movement in realized U.S. premium can matter far more to profit than a modest volume increase.

The geographic mix creates different exposures. Sebree buys electricity at MISO market prices plus transmission and related charges; its agreements extend into 2027–2028 with automatic extensions. Mt. Holly buys from Santee Cooper under a cost-of-service arrangement through 2031, now sufficient for full production. Grundartangi receives roughly 545 MW under hydro/geothermal PPAs with several suppliers, combining fixed and LME-linked pricing and expirations extending into the 2030s. That makes Iceland structurally lower-carbon and partly margin-matched to aluminum, but it does not eliminate equipment or contract-renewal risk.

Jamalco adds partial upstream integration. Century owns 55%, the Government of Jamaica indirectly owns 45%, and Century’s subsidiary manages the venture. Century funds 55% of cash costs/capex, receives or sells the output, and consolidates the operation, with the government’s portion shown as noncontrolling interest. At full capacity, Century’s attributable 770,000 tonnes of alumina is only about half the roughly 1.5 million tonnes theoretically required to feed 770,000 tonnes of aluminum at an industry-standard near-2:1 alumina-to-metal ratio. The integration is meaningful but incomplete.

Customers and concentration

Glencore is simultaneously shareholder, customer, supplier, derivative counterparty and lender. It owned 36.4% at year-end 2025 after selling nine million shares in November and approximately 30% by June 2026. It represented 54% of FY2025 sales, 45.5% of H1 2026 sales and 44.0% of Q2 sales. H1 purchases from Glencore were $168.5 million. A Vlissingen credit facility supplied by Glencore offers up to $90 million through December 2026 at SOFR plus 3.687%, subject to a 7% floor and 9% cap; nothing was drawn at June.

The agreements use observable LME/premium or alumina-index pricing, which limits obvious transfer-pricing discretion. Concentration nevertheless has three costs. First, a counterparty disruption would affect revenue and inputs at once. Second, Glencore’s large vote creates governance influence even without majority ownership. Third, a commodity trader may optimize across its global book in ways that are commercially rational but not identical to minority shareholders’ preferred risk posture. The related-party committee and independent-board approval process reduce but do not erase the conflict.

Revenue quality

Revenue is repeat business in the ordinary commercial sense—industrial customers need recurring metal—but it is not recurring revenue in the software or regulated-utility sense. Volume is bounded by physical capacity and planned maintenance; price resets with public benchmarks. FY2020–FY2025 revenue moved from $1.61 billion to $2.21 billion, $2.78 billion, $2.19 billion, $2.22 billion and $2.53 billion. Those swings largely reflect price, premium, plant availability and the addition of Jamalco, not a steadily compounding customer franchise.

Value-added products help at the margin. Billet, foundry alloy and molten-metal delivery carry premiums and can deepen local relationships; Grundartangi’s Natur-Al product can satisfy low-carbon procurement requirements. But peers including Alcoa, Hydro, EGA and Rio Tinto offer comparable alloy, billet and low-carbon choices. Qualification and reliability create friction, not captivity.

Verdict: Century is an understandable but highly variable converter of commodity benchmarks and electricity into metal. The asset footprint is strategically scarce and improving toward full utilization. Revenue quality remains low because price, premium and plant availability dominate customer growth.

3. Industry Dynamics

A global price, local power and regional premiums

The aluminum industry is global in price discovery and local in cost. LME metal travels; electricity grids, tariffs, port costs and product form create regional wedges. This produces three overlapping cost curves:

1. Alumina: bauxite quality, caustic soda, energy, freight and refinery reliability determine the upstream spread.
2. Primary smelting: electricity—often 13–15 MWh per tonne—plus alumina, carbon, labor and anode efficiency determine the conversion spread.
3. Regional delivery/product: tariffs, freight, inventory, billet/foundry availability and customer location determine premiums above LME.

Century is exposed to all three. Jamalco offsets some alumina cost but also introduces mining/refining outages. Iceland offers renewable power and a European outlet. The U.S. assets capture Midwest scarcity but depend on policy and local power arrangements.

U.S. scarcity is real

USGS estimates that in 2025 three companies operated six U.S. smelters, with two at full capacity, two reduced and two idled. Domestic primary production was about 660,000 tonnes against 5.7 million tonnes of apparent consumption; net import reliance was 60%. Production had fallen from 889,000 tonnes in 2021, while 1980 capacity had been near five million tonnes across 33 sites. This is a genuine structural shortage, not a marketing claim.

Scarcity does not automatically mean attractive greenfield returns. The Aluminum Association estimates slightly less than half of installed U.S. primary capacity is operating; restarting all idled capacity would meet only 15–20% of imports. A new 750,000-tonne smelter takes roughly five to six years, $4–$6 billion and about 11 TWh of power annually. The bottleneck is not technical know-how alone. It is securing long-duration electricity at a cost compatible with global metal, then funding a plant whose returns can be destroyed by a commodity trough.

That is the Greenwald/Marathon distinction that matters: replacement cost and entry barriers can make existing assets scarce without granting the owner autonomous pricing. The U.S. premium can compensate local producers, but if policy is reduced or imports adapt, the rent shrinks faster than the physical plant can.

Global supply response

China dominates primary aluminum, and its approximately 45-million-tonne capacity ceiling is a policy constraint rather than a physical law of global supply. Output can rise through utilization, replacement and capacity migration; Chinese capital and technology can also appear offshore. India, Indonesia and the Middle East have access to lower-cost power or state-supported capital. EGA itself is a world-scale competitor and Century’s prospective partner.

Western smelter closures and limited restarts make the near-term ex-China supply curve favorable. Yet a favorable capital cycle can sow its reversal. High prices and tariff premiums justify restarts, recycling, substitution and new projects. Secondary aluminum already supplies far more U.S. tonnage than domestic primary metal. Scrap cannot replace every high-purity or alloy application, but it caps the addressable deficit and requires far less electricity/capital.

Alumina has a different cycle. Jamalco diversifies Century upstream and can benefit when refinery supply is tight. New Asian refinery capacity can instead compress alumina margins while lowering the input cost of Century’s smelters. Because the company is only partly integrated, it does not have a simple long or short alumina exposure; operating outages and transfer prices matter alongside the index.

Tariffs and manufactured scarcity rent

U.S. Section 232 aluminum duties began at 10% in 2018, rose to 25% in March 2025 and 50% in June 2025. The Midwest premium is not the tariff itself, but in an import-dependent market it tends to incorporate tariff and delivery economics. Century’s H1 2026 average Midwest premium of $2,406/t versus $855/t a year earlier illustrates the scale of policy transmission. That spread accrues to domestic production without an equivalent incremental production cost.

The July 2026 proclamation adds a second mechanism. Commerce may approve onshoring plans that commit to building, refurbishing or expanding U.S. primary capacity and starting construction by 20 January 2029. An approved company may import an annual quantity corresponding to the future facility’s reasonably anticipated output at half the otherwise applicable tariff. Commerce must assess timing, milestones, costs, output and benefit allocation, monitor performance and may rescind benefits.

For the 60/40 EGA/Century 750,000-tonne project, Century expects 300,000 tonnes of import allocation. At management’s $3,250/t illustrative LME and a 25-point tariff differential, the gross annual arithmetic is $243.75 million. It could be an elegant bridge: import savings finance domestic construction until output begins around the end of 2029. It could also produce moral-hazard and duration risk: the value depends on tariff rates, approvals, compliance, import sourcing and future domestic premium, all while capital spending is irreversible.

45X and the policy stack

Section 45X provides a 10% credit on qualifying U.S. critical-mineral production costs, including specified labor, energy, depreciation/amortization and overhead. Century recognized $59.3 million for 2023, additional catch-up benefits in 2024, and more than $50 million in H1 2026. The July 2026 receipt of a $94.3 million refund materially strengthened cash.

The Oklahoma project also has a DOE cooperative agreement of up to $500 million, paid as eligible milestones are completed on a cost-share basis. Only $10 million had been awarded for Phase 1 when DOE described the project in January 2025; later DOE communications characterized the full $500 million as support. The filing warns that availability remains subject to terms and performance. It is therefore wrong to count the grant as cash in hand or subtract the full amount from Century’s capital share without knowing how the EGA venture allocates it.

Demand and substitution

Transportation represented 36% of 2025 U.S. aluminum consumption, packaging 24%, building 13%, electrical 9%, consumer durables and machinery 8% each. Electrification, grid investment, lightweighting, packaging and defense support low-single-digit secular demand. Demand quality is mixed: vehicles and construction are cyclical; packaging is steadier; electrical/defense can be policy-supported. Aluminum can substitute for copper in some electrical uses, but steel, composites and scrap-based secondary aluminum also compete with primary metal.

Century’s Oklahoma high-purity ambition addresses defense, semiconductors and specialized applications where specifications matter more. Until contracts, qualification, mix and premiums are disclosed, those end markets are strategic rationale rather than modeled economics.

Verdict: structurally attractive near-term for incumbent U.S. smelters, structurally difficult over a full cycle. Scarce power/permitting and import reliance support current rents; global price-taking, recycling and state-backed supply prevent a durable industry-wide moat. Policy has raised the profit floor but also invited a capital response whose returns depend on policy duration.

4. Competitive Position

What the company genuinely has

Century owns assets that cannot be recreated quickly. Sebree and Mt. Holly sit inside the world’s most import-dependent large aluminum market. Grundartangi combines renewable electricity, long PPAs and low-carbon product qualification. Jamalco provides bauxite/alumina control. Vlissingen internalizes most Icelandic anode supply. The company also has operating experience, labor relationships, permits and customer qualifications that a greenfield entrant would need years to build.

Scale relative to the U.S. niche matters. With Sebree and Mt. Holly together at 450,000 tonnes of capacity, Century represents most currently operating U.S. primary production. That makes it a natural beneficiary of 45X and Section 232 and a credible partner for EGA. It also gives government a reason to keep the company solvent. These are real strategic advantages.

Why those advantages are not a moat

Greenwald’s test is whether an advantage protects demand and returns from imitation or competition. Century fails the pricing-power test. The LME and regional premiums set revenue. An equivalent tonne from Alcoa, Hydro, Rio Tinto, EGA or another qualified supplier is substitutable once freight, tariff, form and alloy are adjusted. Customers do not incur software-like switching costs, and Century does not own a network that gets stronger with use.

Economies of scale exist at the plant level but have limits. A potline spreads maintenance, labor and anode facilities over more metal, so full Mt. Holly/Grundartangi output should lower unit costs. Yet the corporate platform is smaller and less diversified than Alcoa or Hydro, lacks captive power ownership comparable to Hydro and has suffered multiple single-asset outages. Scale can reduce conversion cost; it cannot offset a $1,000/t market move indefinitely.

Low-carbon differentiation is useful but contestable. Grundartangi’s hydro/geothermal power supports Natur-Al, and Oklahoma is designed to avoid much of a traditional smelter’s emissions. Hydro REDUXA, Alcoa EcoLum, EGA CelestiAL and other certified products pursue the same buyers. The financial test is a sustained realized premium or protected volume. Century does not disclose a separate low-carbon premium large enough to establish that result.

Asset-by-asset position

Grundartangi is probably the highest-quality smelter: modern relative to U.S. plants, renewable-powered and positioned for European foundry/billet demand. The LME-linked component of PPAs provides partial margin matching. The transformer failures expose concentrated infrastructure risk, and some power agreements require extension/renewal in the early-to-mid 2030s.

Sebree benefits from location, molten-metal/billet options and near-full utilization. Market-based MISO power gives upside in cheap-power periods and severe downside during spikes. It is an operating asset rather than a contractual-cost moat.

Mt. Holly captures U.S. premium and now has enough Santee Cooper power for full production through 2031. The restart improves fixed-cost absorption and adds more than 50,000 tonnes. Its economics depend on the cost-of-service tariff and successful steady-state operation; project costs were excluded from adjusted EBITDA during ramp-up.

Jamalco improves supply assurance and captures alumina exposure but has an uneven reliability record. A 2021 fire, 2023 power equipment failure, 2025 hurricane effects and continuing power-island work show the operational burden. The TG4 turbine and expected roughly $20/t improvement help, but the Government of Jamaica partnership and residue/mine obligations complicate capital decisions.

Glencore as advantage and constraint

Glencore reduces marketing and raw-material friction. It provides large offtake, alumina/raw materials, risk-management contracts and credit. Those relationships can lower transaction costs and ensure volume placement. They also concentrate bargaining power and make Century less independent. Because Glencore can source metal globally, its switching cost is likely lower than Century’s cost of replacing half its sales.

Governance protections matter: related-party transactions are intended to be arm’s length and subject to independent review. The structural conflict persists. A controlling-style shareholder with trading relationships can influence strategy, board composition and capital allocation even at 30% ownership. Minority investors should therefore require not only market-based contract formulas but transparent volumes, maturities and governance processes.

Competitor comparison

Moat scorecard

Verdict: Century has scarce assets and policy relevance, not a durable corporate moat. The defensible edge is relative—existing U.S. capacity and Icelandic power are better positioned than a new Western entrant—not absolute pricing power or customer captivity.

5. Growth History and Forward Opportunities

Historical growth: price before volume

Revenue growth has been cyclical rather than linear. FY2020–FY2025 sales were $1.61 billion, $2.21 billion, $2.78 billion, $2.19 billion, $2.22 billion and $2.53 billion. The 2021–2022 increase was mainly price/premium. The 2023 decline reflected lower pricing and lower shipments after Hawesville’s curtailment, partly offset by Jamalco. The 2025 increase of $308 million came primarily from $416 million of favorable realized LME and regional premiums, offset by lower alumina sales and unfavorable volume/mix. This is not historical evidence of steady unit demand growth.

Shipments tell the same story: approximately 784,000 tonnes in 2021, 769,000 in 2022, 701,000 in 2023, 678,000 in 2024 and 647,000 in 2025. Capacity was curtailed or disrupted even as price improved. Growth now depends on restoring output before building new output.

Near-term brownfield growth

Mt. Holly’s restart adds more than 50,000 tonnes of annual production versus its prior 75% operating level. Management said the last 90 pots had restarted in Q2 2026. At a hypothetical $1,000/t conversion margin, 50,000 tonnes would add $50 million of annual EBITDA before ramp inefficiency; at $2,000/t it would add $100 million. Those are sensitivities, not guidance. The economic merit rests on Santee power, pot stability and sustained U.S. premiums.

Grundartangi’s Line 2 return restores lost rather than new capacity. Its benefit is nevertheless material because 2025 production was 260,000 tonnes versus 320,000 tonnes of design capacity. Replacing failed transformers in Q4 2026 should reduce the risk of repeated forced outages. Insurance reimbursements soften capital cost but are not operating earnings.

Jamalco’s TG4 turbine targets reliability and about $20/t of cost reduction. Applied to 1.4 million tonnes gross capacity, that suggests roughly $28 million gross or $15 million at Century’s 55% interest if fully realized. This is a directional calculation; actual throughput, fuel efficiency and government sharing will determine value.

Oklahoma: transformative volume, disproportionate capital

Oklahoma Primary Aluminum is designed for 750,000 tonnes per year, with EGA 60% and Century 40%. Century’s economic share would be 300,000 tonnes—roughly 39% of current operating capacity—and the project would more than double U.S. production. Bechtel is conducting preparatory engineering. The partners target a final investment decision and start of construction by year-end 2026, with first hot metal near the end of 2029.

The growth case is attractive on four dimensions: new technology should lower energy intensity; the U.S. is deeply import-reliant; high-purity output can access defense/technology applications; and public support may subsidize both construction and bridge economics. It is also the most important risk because capital arrives years before stable production.

The broad funding envelope is visible even without a final budget. EGA described roughly $4 billion for the earlier 600,000-tonne concept; the Aluminum Association estimates $4–$6 billion for a 750,000-tonne plant. A pro-rata 40% Century share is $1.6–$2.4 billion. Against that are up to $500 million of DOE cost share, state/local incentives, project financing and potentially roughly $244 million per year of onshoring import economics. The allocation of each source is not yet disclosed. If Century can fund most equity from pre-completion policy cash and non-recourse debt, growth may be accretive. If the company guarantees project debt or issues equity at cyclical prices, absolute EBITDA growth can destroy per-share value.

Hawesville/data-center optionality

Century sold the idled Hawesville site in February 2026 for $200 million cash and a 6.8% interest in a Terawulf affiliate developing a data-center campus. The affiliate has a 20-year lease associated with Anthropic and phased 401 MW development targeted for late 2027/early 2028. Century has no funding obligation. The cash is realized; the equity stake is an option whose value depends on construction, power, tenant performance, dilution and monetization rights.

The transaction was sound asset recycling: an uneconomic, curtailed smelter became liquidity and upside without ongoing capital. The $287.9 million GAAP gain is not recurring EBITDA or cash beyond the $200 million proceeds. The 6.8% stake should remain outside base valuation until fair value, distributions or a credible liquidity path are disclosed.

Product and low-carbon mix

Value-added billet/foundry output and low-carbon Natur-Al can grow premium per tonne without adding equivalent smelting capacity. The Grundartangi billet casthouse and U.S. cast products support this strategy. Quality of growth depends on disclosed realized premium and customer commitments. Without that evidence, mix growth should be modeled as modest margin support, not a moat-like revenue stream.

Growth quality tests

Verdict: near-term growth from existing assets is relatively high quality because it improves utilization of sunk infrastructure. Oklahoma is strategically coherent but not yet demonstrably value-creating; financing, power and policy allocation—not demand rhetoric—will determine its per-share quality.

6. Financial Quality

Five-year record: improving margins, poor aggregate cash conversion

The FY2025 comprehensive 10-K restated FY2023–FY2024 for Jamalco’s consolidation. Those restated figures control over earlier reports. Conventional free cash flow below is operating cash flow less cash capital expenditure; it is not management’s adjusted measure.

The five-year FCF total was negative $219.6 million despite $154.3 million of cumulative operating income. Average annual operating cash flow was $45.4 million against $89.4 million of capex. Working capital was a recurring source of volatility: identifiable operating-asset/liability movements were approximately negative $75 million in 2021, negative $40 million in 2022, positive $27 million in 2023, negative $176 million in 2024 and negative $37 million in 2025. The 2024 drain included a $92.6 million 45X receivable build, $64.3 million inventory build and $50.6 million payables reduction.

The table does not prove that current assets are permanently cash-consumptive. It proves that the combination of commodity inventory, tax-credit receivables, restarts and outages has not generated a stable distributable record. Management’s identified non-growth plant capex was only $17–$28 million in 2021–2023 before Jamalco’s consolidation. But keeping an old smelter system reliable requires lumpy transformer, potline, environmental and pension cash that a single-year “sustaining” label can miss.

Earnings normalization

Company-adjusted EBITDA helps isolate derivative marks, impairments and outage/restart costs, but its trend should be read beside GAAP cash:

The 2021–2025 average is approximately $221 million. That is too low as a mechanical forecast because Hawesville has been sold, Mt. Holly is fuller, Jamalco is consolidated and 45X now exists. The Q3 guide is too high as a perpetuity because the Midwest premium, LME and Gulf-related scarcity are exceptional. A useful normalization therefore builds from plant volume and separate price/policy cases rather than averaging endpoints.

Section 45X is load-bearing. Credits reduced COGS by $56.5 million in 2023, $89.7 million in 2024 and $89.1 million in 2025, plus $2.8–$3.8 million annually in SG&A. Removing the credits changes operating income from $27.5/$108.4/$158.1 million to approximately negative $31.8/$15.8/$65.2 million. The 2025 benefit was nearly 59% of operating income. Under current law it begins a 25%-per-year phase-down in 2031 and reaches zero in 2034.

H1 2026 quality of earnings

H1 sales were $1.401 billion, gross profit $346.7 million, operating income $585.6 million, net income to Century $586.8 million, operating cash flow $236.0 million and capex $134.4 million. The $101.6 million of conventional FCF is real. The income statement requires three separations:

• Hawesville’s sale contributed a $287.9 million operating gain, versus $200 million of cash proceeds plus the data-center interest.
• Grundartangi insurance produced $73.1 million of gains in H1; it partly offsets actual outage cost and repair capital rather than representing a recurring margin.
• Derivatives generated a $58.2 million loss. Some contracts economically hedge metal, power and raw-material exposures, so removing all derivative results without examining the physical exposure would overstate cleanliness.

Excluding only Hawesville, H1 operating income was about $297.7 million. Excluding the roughly $52.5 million H1 45X operating benefit as well gives approximately $245 million. That remains a major improvement, consistent with the realized-price bridge, but far below reported GAAP operating income.

Management’s Q2 adjusted EBITDA attributable to Century was $326.9 million versus GAAP operating income of $211.6 million. The bridge added D&A, $76.6 million of Iceland failure cost, $10.7 million of Mt. Holly restart cost and other items, while excluding insurance within designated adjustments. This is a reasonable operating lens if the outages are truly exceptional. Two transformer failures and repeated Jamalco disruptions argue for retaining a recurring reliability reserve in through-cycle cash.

Unit economics and sensitivities

Primary shipments declined 17.4% from 783,647 tonnes in 2021 to 647,112 in 2025. Primary-aluminum sales per shipped tonne were approximately $2,755, $3,500, $2,806, $2,755 and $3,396 over those years. H1 2026 shipments fell to 253,497 tonnes from 344,413 because of Iceland, but primary-aluminum sales were approximately $5,030/t as U.S. premiums surged.

At 2025 volume, a $100/t change in realized aluminum price changes revenue by roughly $64.7 million before contractual offsets. In 2025, a $415.9 million metal-price benefit produced only an $84.4 million gross-profit increase after $115.7 million of higher raw-material cost, $93.0 million of power, $72.5 million of other costs and $42.1 million of volume/mix headwinds. Revenue beta is not EBITDA beta.

Power sensitivity is similarly large but nonuniform. Full-load consumption of approximately 11.63 TWh means a $1/MWh change equals around $11.6 million of annual system cost, before contractual structure. Filing sensitivity estimates based on plant arrangements are $3.4 million at Sebree, $3.5 million at Mt. Holly and $4.8 million at Grundartangi per $1/MWh, or $15.9 million total. Mt. Holly is cost-of-service, Iceland is largely LME-linked and Sebree retains the most direct market risk.

Returns on capital

Using 25% normalized tax on GAAP operating income and average invested capital defined as equity including NCI plus debt less cash, approximate ROIC was 6.1% in 2021, negative 17.5% in 2022, 2.3% in 2023, 7.4% in 2024 and 9.0% in 2025. Removing 45X lowers 2023–2025 to approximately negative 2.7%, 1.1% and 3.7%. These are reconstructed returns, not company KPIs, and purchase accounting/curtailments distort the base. They still fail to show sustained above-cost-of-capital performance.

ROE is less useful. The 2024 64% result was inflated by the $245.9 million Jamalco bargain-purchase gain; removing that item alone reduces it toward the high teens. 2025 ROE was only about 5.6%. A cheap acquisition can create accounting equity gain without recurring cash.

Balance sheet and liquidity

At 30 June 2026 unrestricted cash was $343.4 million, restricted cash $46.3 million, debt approximately $480.0 million and total equity including NCI $1.535 billion. Unrestricted net debt was only $136.6 million. Total liquidity was $784.9 million: cash, $44.8 million of restricted Hawesville proceeds and $396.7 million of borrowing availability. July’s $94.3 million 45X refund and further insurance recovery led management to say cash exceeded total debt.

Debt consists primarily of $394.4 million carrying value of 6.875% secured notes due 2032 and $85.6 million of 2.75% converts due 2028. The converts are deeply in the money at an $18.74 conversion price and represent roughly 4.6 million potential shares; fair value was $212.9 million at June. Treating them only as cheap debt understates dilution. The U.S. and Iceland revolvers were undrawn and covenant headroom was wide.

Retirement/environmental claims remain important. At year-end 2025 pensions were underfunded by $58.7 million and OPEB by $38.9 million. Expected 2026 pension contribution was $14.8 million, with annual benefits around $22–$24 million through 2030. ARO was $83.4 million at December and $74.3 million noncurrent after Hawesville’s transfer. These claims are not existential at current liquidity, but they belong in normalized cash and enterprise claims.

Accounting controls and working capital

In March 2026 Century concluded that its FY2023–FY2024 and Q1–Q3 2025 statements should not be relied upon because it proportionately consolidated certain Jamalco production assets rather than fully consolidating them. The restatement added $180.6 million to FY2024 assets, $34.3 million to noncurrent liabilities and $142.7 million to NCI; FY2024 operating and consolidated net income fell $13.0/$14.0 million, while income attributable to Century was unchanged. Deloitte gave an adverse ICFR opinion, and the material weakness remained unremediated at June 2026.

The error did not erase Century’s economic 55% share, but it matters. Oklahoma will introduce a second material joint venture with complex grant, tariff, financing and consolidation judgments. An unremediated consolidation weakness raises the execution premium minority holders should demand.

Working capital is procyclical. At June inventories were $582.9 million, accounts receivable $136.2 million, nontrade receivables $65.1 million and manufacturing-credit receivables $225.2 million. A commodity decline can release inventory cash, but it also reduces the receivables/inventory that support the borrowing base just as earnings weaken. The apparent liquidity cushion is therefore strongest near the top of the price cycle.

Verdict: financial quality has improved from fragile to liquid, not from cyclical to compounding. Full production and policy support can produce substantial cash; five-year FCF, subsidy-free ROIC, material control weakness and repeated outage capital show that scale has not yet delivered durable economics.

7. Capital Allocation

What the record says

Century’s capital allocation has been defensive and asset-centric. The 2021–2025 period combined debt refinancing, Mt. Holly/Grundartangi growth projects, Hawesville’s curtailment and impairment, Jamalco’s near-zero-price acquisition, and very limited shareholder distributions. This posture was rational for a business whose five-year FCF was negative, but it means the equity case depends almost entirely on better future operating assets rather than cash returned to owners.

The company refinanced maturities rather than maximizing leverage. It issued/refinanced secured notes, kept revolvers available and ended June 2026 with modest net debt. The 2028 convert has a low coupon but embeds approximately 4.6 million shares of potential dilution. Capital structure is now capable of absorbing normal volatility; it is not large enough to fund Oklahoma from balance-sheet cash alone.

Jamalco: asymmetric acquisition, asymmetric obligations

Century acquired 55% of Jamalco in May 2023 for $1 and assumed operating, environmental and partnership obligations. The later $245.9 million bargain-purchase gain indicates assets exceeded recognized consideration/liabilities, but the relevant return is future cash after sustaining capex, outage cost, government sharing and residue/mine closure—not the gain.

The acquisition was strategically coherent because it supplied roughly half of Century’s alumina needs and created upside to refinery recovery. It was also opportunistic at a distressed price, which is the kind of countercyclical behavior the capital-cycle framework favors. The record since purchase includes a generator failure, hurricane costs, power-island investment and an accounting restatement. It is too early to call the acquisition a value creation success, but the risk/reward was more favorable than a peak-cycle cash acquisition.

Brownfield capital

Mt. Holly and Grundartangi are easier to justify than a greenfield project. Returning curtailed pots spreads fixed costs over more tonnes using existing infrastructure and customer channels. Grundartangi transformer replacements restore high-quality capacity and are substantially insured. Jamalco TG4 should reduce operating cost and failure risk.

2026 capex guidance of $180–$190 million includes $70–$80 million of insured Iceland repair, about $50 million for Mt. Holly and $20–$25 million for Jamalco. That leaves roughly $40–$50 million of other spend. The classification is informative but not a promise that future sustaining needs stay that low; deferred maintenance, pot relining and environmental obligations are lumpy.

Hawesville: good recycling, careful accounting

The February 2026 Hawesville sale converted an impaired/idled smelter into $200 million of cash, transferred $8.8 million of net assets/ARO and added a 6.8% data-center interest. The stake includes a contingent put that may require the majority holder to buy Century’s interest for $300 million after the campus delivers at least 375 MW and specified timing conditions are met. The initial accounting implied roughly $96.7 million of value for the interest/put, using Level 3 assumptions.

The transaction is a positive allocation decision because Century has no funding obligation and retained upside. Two cautions matter. First, $44.8 million of proceeds was restricted at June for capital/business investment. Second, the contingent $300 million is not cash today; it depends on power delivery, construction, tenancy and enforceability. The $287.9 million gain must never be capitalized.

Shareholder returns and issuance

Century paid no regular dividend and has executed no repurchases since April 2015. Approximately $43.7 million remained under the stale authorization, but management deferred returns while defining Oklahoma financing. That choice preserves optionality and is sensible before a large FID; it also signals that current windfall cash is likely project equity, not distributable capital.

Common shares increased from roughly 90.1 million in late 2021 to 99.0 million by June 2026. Much of the 2025 jump reflected automatic conversion of Glencore preferred stock, whose economic participation already affected EPS, rather than a fresh cash raise. SBC nevertheless rose to $47.0 million in 2025—1.9% of sales and nearly 30% of operating income. The in-the-money convert creates additional dilution.

The five-year Form 4 corpus shows only one open-market purchase, a stale May 2023 purchase of about $130,500. Code-S sales totaled $633.6 million, but Glencore’s two strategic sell-downs accounted for $599.1 million. Excluding Glencore, officers sold roughly 1.02 million shares for $34.6 million. CEO Jesse Gary sold 300,000 shares for about $15.6 million in early 2026 under disclosed trading plans. The evidence is not a clean bearish signal, but there is no recent open-market insider buying to validate value at current prices.

Incentives and governance

The 2026 proxy should be read in light of unusual 2025 earnings and project ambitions. Annual/long-term awards use safety, operations, adjusted EBITDA/cash and relative shareholder-return measures. The CEO’s 2025 annual incentive paid at 136.4% of target and the completed PSU tranche at 200%. Those outcomes can be mechanically consistent with strong share price and adjusted earnings, but the scorecard lacks a simple, externally verifiable through-cycle ROIC, absolute FCF or FCF-per-diluted-share gate for Oklahoma. Adjusted EBITDA can rise when a tariff raises premiums or when management excludes project/outage costs; absolute growth does not ensure per-share value.

Glencore owned 29.99% after its March 2026 sale and remains the dominant commercial counterparty. It retains a board-nomination right and pre-emption rights under governance arrangements, although Century is not disclosed as a NASDAQ controlled company. Related-party governance and independent director review are therefore essential. The Jamalco material weakness further raises the burden on audit/control oversight just as the EGA venture adds complexity.

Oklahoma capital-allocation test

The project should pass all of the following before Century commits uncapped parent capital:

1. a long-duration power contract with transparent escalators and curtailment protections;
2. a fixed or bounded construction budget, contingency and EGA completion support;
3. binding Commerce approval and disclosed allocation of import benefit;
4. clear ownership of the $500 million DOE grant and state/local incentives;
5. project debt that is non-recourse or tightly capped at Century;
6. per-share return above WACC under normalized LME/Midwest premium and after 45X phase-down;
7. governance that prevents a 40% owner from funding overruns without proportionate control.

Without those gates, today’s peak cash could finance new capacity that compresses tomorrow’s premium—the classic capital-cycle error. With them, Century can use temporary policy rent to acquire a modern, lower-cost share of U.S. capacity without balance-sheet distress.

Verdict: recent allocation is mixed-to-improving. Jamalco and Hawesville were opportunistic, liquidity management was prudent, and brownfield spend is economically sensible. Rising SBC, no per-share return hurdle and the unresolved Oklahoma funding envelope prevent a stronger verdict.

8. Changes and Headwinds — Last Two Years

The company entering H2 2026 is materially different from the 2024 version. The sequence matters because reported earnings combine operating improvement, policy, asset sales, outages and accounting repair.

Operational change

The positive case is that 2025 understated physical earnings power. Grundartangi lost much of Line 2, Mt. Holly had not completed its restart, Jamalco still needed power work and Hawesville consumed attention. By Q3 2026 management expects all operating smelters nearer full output and TG4 online. If reliability holds, volume and fixed-cost absorption should improve even without another premium increase.

The counterpoint is that repeated failures are part of the evidence set, not arbitrary adjustments. Grundartangi’s first failed transformer was followed by a second. Jamalco has experienced fire/power/hurricane interruptions. Mt. Holly’s restart incurred excluded costs and must demonstrate stable amperage. A normalized case should retain a reliability allowance rather than assume every plant runs at nameplate forever.

Policy change

The 2025–2026 U.S. tariff stack fundamentally changed realized economics. Domestic producers receive the Midwest scarcity premium; 45X reimburses qualifying cost; DOE supports the greenfield project; the new import program may pre-fund construction. This is better than a simple one-off stimulus because separate mechanisms reinforce each other.

It is also concentrated sovereign risk. Section 45X phases out after 2030 under current law. Section 232 and import-program rates can change with executive policy. Commerce can modify rules and rescind benefits. A project whose construction extends to 2029 cannot instantly adapt if the rent disappears.

Oklahoma legal/power headwind

Oklahoma’s Attorney General has challenged arrangements linked to generation/storage investment for large loads including the smelter and sought remand/preliminary relief. The Department of Justice filed in support of allowing federal Clean Air Act/Clean Water Act processes to proceed. The dispute does not prove the project will fail. It shows that the most important operating input—power—also carries political, regulatory and ratepayer risk.

Accounting change

The Jamalco consolidation restatement did not change net income attributable to Century, but it exposed inadequate controls over a complex JV. Until remediated and tested, reported working capital, liabilities and noncontrolling-interest presentation warrant extra scrutiny. Oklahoma will make this weakness more—not less—relevant.

Verdict: changes strengthen near-term cash generation and strategic relevance but increase duration, policy and execution risk. The operating asset base is better; the future capital commitment and reporting complexity are much larger.

9. Risk Analysis

Downside architecture

The most plausible downside is not immediate insolvency. It is multiple and earnings compression: Midwest premium falls, annual EBITDA normalizes below $500 million, working capital/borrowing base weaken and Century continues Oklahoma spending. With current cash and long-dated notes, management can survive that scenario, but the equity can reprice severely—as the 82% 2022 drawdown demonstrates.

The severe downside combines three variables: tariff/import benefit is reduced, Oklahoma capital has already been committed and metal/power spreads turn adverse. In that state, project debt or new equity could subordinate/dilute shareholders while existing plants generate little cash. The current balance sheet buys time; it does not make a $1.6–$2.4 billion gross capital share small.

Catastrophic physical risks are lower probability. Potline freezes, transformer loss, tailings/residue failure or environmental orders can close a plant and create uninsured liabilities. Geographic diversification across the U.S., Iceland and Jamaica limits a single-event total loss, but the company is still small enough that one major asset matters.

Mitigants

The strongest mitigants are $785 million of June liquidity, July cash receipts, low unrestricted net debt, insured Iceland repairs, full U.S. capacity, benchmark-linked Iceland power and EGA’s 60%/technology role. The Hawesville stake provides contingent value without funding obligation. A binding import award can materially de-risk project funding. None of these mitigants eliminates commodity duration or project-cost risk.

Verdict: balance-sheet risk is currently low-to-medium; earnings and capital-cycle risk are high. The equity’s main danger is paying/committing capital on peak policy-assisted cash flow, not near-term debt maturity.

10. Valuation Discussion — Embedded Expectations

Current enterprise bridge

At the 28 August $46.59 close, ROIC.ai calculated market capitalization of approximately $4.55 billion and enterprise value of $4.82 billion using $343.4 million cash, $479.9 million debt and $131.4 million noncontrolling interest. The diluted market capitalization was $4.68 billion, reflecting in-the-money securities. The filing’s 99.0 million common shares and roughly 4.6 million convertible shares make diluted treatment important.

No single denominator is “correct.” The current quarterly guide captures tariff/premium and full-restart upside but not through-cycle mean reversion. The five-year average captures adversity but understates Jamalco, full Mt. Holly and current policy. A decision should scenario LME, Midwest/European premium, 45X, volume, power and project funding separately.

Comparable-company context

ROIC.ai TTM enterprise values/multiples are a date-consistent cross-check, not primary filings. Different integration and accounting make them directional.

CENX’s lower current EV/EBITDA than some peers is not evidence of undervaluation by itself. Its denominator is more exposed to tariff-created Midwest premium and 45X. Hydro’s low multiple comes with different currency, power ownership and integrated earnings. Constellium/Kaiser deserve different margins and working-capital treatment. Alcoa is the best cross-read, but its alumina and broader portfolio dilute the U.S.-premium beta.

Scenario framework

The following scenarios value no equity and set no price target. They ask what the current $4.82 billion EV is paying per dollar of operating/cash output. Revenue is an assumption derived from volume and realized price; EBITDA is after an explicit policy/premium view. Oklahoma construction capital is shown separately because it is not existing-business EBITDA.

The bear is a 2028 downside rather than a liquidation case: 620,000 tonnes, sharply normalized Midwest premium, recurring outage/power cost and no Oklahoma earnings produce about $190 million of EBITDA. A true potline/environmental crisis could be worse. The base uses $480–$500 million, above FY2025 but less than 40% of the Q3 annualized guide, recognizing fuller plants and 45X while fading exceptional Midwest premium. The bull retains a high premium and excellent uptime but remains below current annualized guide to allow normal outage/cost variability. None of the three includes Oklahoma operating earnings before first hot metal; project financing/dilution is evaluated separately.

Embedded expectations

At current EV, a 7× through-cycle multiple requires about $689 million of sustainable EBITDA; an 8× multiple requires about $603 million. Alternatively, at $500 million sustainable EBITDA the current EV is 9.6× before Oklahoma capital. Those relationships suggest the market assumes several of the following:

• operating output returns toward 700–770 kt and stays reliable;
• the Midwest premium remains structurally above 2024 even after Gulf supply and domestic capacity respond;
• 45X survives to its scheduled phase-down and is converted to cash;
• Commerce approves a meaningful import allocation whose cash helps rather than merely offsets construction cost;
• Century funds its $1.6–$2.4 billion gross project share without major parent dilution or leverage;
• Oklahoma produces at an attractive power cost by 2029–2030;
• the Hawesville interest has positive value, though not necessarily the full contingent put.

The current price does not require Q3 EBITDA to last forever: capitalizing $1.34 billion at even 5× would exceed current EV substantially. It does require a much higher floor than the historical record. The fragile load-bearing assumption is that policy-raised Midwest economics persist long enough to finance Oklahoma and do not collapse once new supply arrives.

Project and option treatment

Oklahoma should not be added to current EBITDA before FID. Nor should its full capital cost simply be subtracted today: grants, EGA, project debt and import benefits can fund it. The correct approach is milestone probability:

1. Pre-FID: zero operating value; expense Century development cash; recognize no import benefit until approved.
2. Approved/power-contracted/financed: recognize the net present value of Century’s 40% cash flows after its disclosed equity contribution, not gross capacity × spot margin.
3. Construction: discount for schedule/overrun and policy recapture risk.
4. Stable hot metal: value through-cycle EBITDA with full sustaining capital, working capital and 45X phase-down.

The Hawesville interest is also probability-weighted. The $200 million cash is already in the balance sheet. The 6.8% stake/put has positive optionality but no reliable public mark; the contingent $300 million put cannot be treated as near-cash before the 375 MW milestone.

AZI’s own-history valuation percentile was unavailable because the environment lacked a valid AZI API token. No percentile is inferred. That limitation is especially relevant for a cyclical company whose P/E is distorted by sale gains and derivative marks; EV/normalized EBITDA and FCF scenarios are more informative anyway.

Verdict: the security is inexpensive on current quarterly earnings and demanding on demonstrated through-cycle cash. Current EV embeds a durable step-up to roughly $600–$700 million of EBITDA or substantial project/policy option value. The valuation debate is therefore duration and funding, not whether Q2 was strong.

11. Variant Perception

What consensus appears to believe

The price has more than tripled over five years and remains more than double its September 2025 low despite a 33.9% retreat from the June 2026 intraday high. The current 3.6× annualized Q3 guide suggests the market knows current earnings are elevated and discounts them. The higher price versus prior cycles suggests it also assigns value to a raised U.S. profit floor, full production, Oklahoma import benefits and Hawesville optionality.

The dominant narrative is coherent: U.S. import dependence is strategic, Section 232/45X have bipartisan/national-security logic, Century is the largest domestic operator, EGA/DOE reduce greenfield risk and near-term cash can fund growth. The variant question is not whether those facts exist. It is whether benefits accrue to existing shareholders after the capital cycle responds.

Strongest bull case

The bull starts with current cash, not distant promises. Q3 guide midpoint is $335 million, Mt. Holly/Grundartangi are only now reaching better volume and July receipts put cash above debt. If the U.S. Midwest premium remains structurally high, existing plants can generate several hundred million dollars of annual FCF. Commerce approval could add approximately $244 million gross annual import economics from 2027 until Oklahoma starts. Three years would approach $730 million before tax/working capital. Combined with a $500 million DOE grant, state incentives and project debt, that could finance much of Century’s 40% contribution without equity issuance.

Oklahoma then gives Century 300,000 tonnes of modern, efficient capacity in a market still dependent on imports even after commissioning. EGA supplies technology/operating expertise and bears 60% of capital. High-purity output can earn better mix. Hawesville’s data-center interest may monetize for meaningful additional capital. Under that path, current EV captures neither the duration of existing scarcity rent nor the new plant’s cash.

Bull load-bearing assumptions are: tariff/import benefits survive changes in administration/rules; Midwest premium does not fully arbitrage away; power cost is globally competitive; project capex stays near budget; and Century’s parent funding is capped. The bull is falsified if FID requires large recourse leverage/equity or if Commerce benefit is materially smaller/shorter than management expects.

Strongest bear case

The bear observes that subsidy-free ROIC was only about 4% in 2025, five-year FCF was negative $220 million and shipments fell 17%. H1 2026 pricing—not volume—drove the earnings step-up. A 50% tariff created a Midwest premium of more than $2,400/t, while 45X supplied more than $50 million and insurance/sale gains distorted GAAP. CENX has repeatedly lost 75–85% when spreads reversed.

The same rents invite supply. Mt. Holly restarted, Inola alone would add 114% of 2025 U.S. primary output, global Gulf/Chinese/Indian/Indonesian capacity can recover or expand, and secondary aluminum meets many applications with a fraction of the power. When premiums normalize, Century could already be committed to $1.6–$2.4 billion of gross project capital. Shareholders then fund a price-taking asset whose output helps compress the premium used to justify it.

Bear load-bearing assumptions are: policy duration is short, supply/recycling responds, project funding reaches the parent and operational reliability remains mediocre. The bear is falsified if Century sustains more than $700 million of existing-asset EBITDA at ordinary metal/premium conditions while funding Oklahoma without net dilution and earning above-WACC returns.

Factor and positioning evidence

FactorsToday’s 31 July 2026 All Factors model had 39.7% R² and 54.0% stock-specific volatility. CENX loaded +1.256 to Market, +0.878 Materials, +0.613 Mining, +0.862 Oil, +0.792 Gold, +0.687 Infrastructure and +0.238 Value; Low Volatility was −0.625 and Liquidity −0.675. Quality was slightly negative and Momentum was absent—zeroed as negligible by the sparse model, not missing data.

That mix says CENX trades as a high-beta commodity/materials special situation, not a quality or pure momentum factor. One-year return through the factor snapshot was about 108% with 65.7% volatility and a 39.7% maximum drawdown; five-year annualized return was 32% with 71% volatility and an 82% drawdown. The trailing three-month raw move was approximately negative 27% and six-month roughly negative 9%, consistent with the AZI series. Strong 12-month relative strength coexists with a broken short-term one-way trade.

Factor-similar names include Alcoa, XME, SLX, GNR, Teck and Freeport-McMoRan—confirmation that the tape prices commodity/capital-cycle exposure. Low R² and high residual volatility mean company events—tariffs, Oklahoma, outages and Hawesville—still dominate much of the move.

Nasdaq/FINRA-derived secondary data reported 9.18 million shares short at 14 August 2026, down 7.4% from 31 July, with 3.86 days to cover. Float definitions vary because Glencore still owns nearly 30%, so the absolute shares and days-to-cover are more reliable than a quoted short-float percentage. Positioning is material but not an unambiguous squeeze. Glencore’s completed 15.3 million-share sell-down is known strategic supply and should not be conflated with short positioning.

Variant synthesis

The most defensible variant is neither “tariffs are fake” nor “Q3 annualized EBITDA is permanent.” It is that the market may correctly discount current earnings yet underappreciate the destination of the cash. If temporary rent finances a bounded, high-return minority project, value compounds. If it finances uncapped capacity at a cycle peak, current cheapness is optical. The next decisive evidence is contractual—Commerce, power, capex and financing—not another quarter of high premium.

Verdict: consensus is probably right on near-term earnings strength and strategic relevance. The under-debated variable is per-share capital allocation through the supply response. The tape signals elevated event risk rather than clean trend confirmation.

12. Fact vs. Interpretation

13. Open Questions

1. What final gross budget, contingency and completion-guarantee structure will Oklahoma use after the 750,000-tonne redesign?
2. What exact amount of parent equity, recourse debt or guarantees can Century be required to provide under the final JV agreement?
3. Will Commerce approve 300,000 tonnes for Century, for how many years, from which import sources and with what benefit-sharing/audit conditions?
4. How will EGA and Century allocate the up-to-$500 million DOE grant and Oklahoma incentives?
5. What fixed/escalating $/MWh formula, duration, curtailment rights and infrastructure cost will the Oklahoma power contract contain?
6. How do project returns look at ordinary LME and a Midwest premium below $1,000/t, after 45X phases out?
7. What are the stable monthly production and cash cost of Mt. Holly’s last 90 pots after restart exclusions end?
8. What permanent transformer redundancy and insurance recoverability will Grundartangi have after Q4 replacements?
9. Does Jamalco TG4 actually deliver the ~$20/t saving, and what is its normalized sustaining/closure cash burden?
10. What proportion of each smelter’s output is value-added, and what realized premium does it earn after casting cost?
11. What is Natur-Al’s independently measured premium and retention benefit versus Hydro/Alcoa/EGA low-carbon products?
12. When will the Jamalco consolidation material weakness be fully remediated and tested by Deloitte?
13. What protections do minority shareholders have over Glencore-related sales, purchases, derivatives and financing?
14. What is the fair value, dilution protection and earliest realistic monetization date of the 6.8% Hawesville data-center interest?
15. Is the $300 million put funded/guaranteed, and exactly what happens if 375 MW is delivered late or only partially?
16. When do shareholder returns resume, and will management use a through-cycle ROIC/FCF-per-share hurdle before buybacks or Oklahoma funding?
17. How much of current derivative exposure hedges physical risk versus introduces basis/mark volatility?
18. What portion of June inventory and receivables remains in the borrowing base under a 30% aluminum-price decline?
19. Will state litigation alter Oklahoma timing, rate recovery or required Century/EGA commitments?
20. What production/premium response follows Gulf restoration, Chinese output growth and U.S. restarts over 2027–2030?

14. What Must Be True

Bull case requirements

• Existing smelters must sustain at least roughly 700,000 tonnes with no repeat transformer/potline outage large enough to erase brownfield gains.
• Normalized Midwest premium must remain materially above 2024 even after Gulf recovery and supply response.
• Commerce must award a commercially usable import benefit close to Century’s expected 300,000 tonnes, and rules must persist long enough to finance construction.
• DOE/state incentives and project debt must limit Century’s parent equity/guarantee burden; dilution must remain modest.
• Oklahoma’s all-in power and construction cost must support above-WACC returns at ordinary metal prices, not only H1 2026 premiums.
• Jamalco must improve reliability/cost and the accounting weakness must be remediated.

Bull falsification test: before or at FID, the bull is false if the Commerce allocation is absent/materially smaller, Century must provide more than roughly $1 billion of uncovered parent capital or recourse guarantees, or the disclosed project return requires a Midwest premium near H1 2026. After FID, two consecutive years of existing-asset EBITDA below $500 million at ordinary operations would independently falsify the raised-floor thesis.

Bear case requirements

• The Midwest premium must materially mean-revert as tariffs change, import flows adjust, Gulf output returns, recycling grows or domestic capacity restarts.
• 45X/onshoring cash must be temporary or consumed by capital rather than distributable.
• Century’s asset-specific advantages must remain insufficient to earn above-WACC returns through a full cycle.
• Oklahoma must experience cost, power, permitting or financing friction that pushes risk onto Century’s parent balance sheet/share count.
• Recurrent outages, working capital and retirement/environmental cash must continue to absorb reported EBITDA.

Bear falsification test: the low-quality/peak-rent thesis is false if Century produces at least two consecutive post-restart years of more than $700 million existing-business EBITDA and at least $350 million FCF at ordinary LME/premium conditions, while independently calculated after-tax ROIC exceeds WACC, controls are remediated, and Oklahoma is funded without material net dilution or recourse leverage. High EBITDA at today’s premium alone does not falsify it.

Monitoring dashboard

15. Source Appendix

The complete source list appears as Appendix B in the combined report. Primary evidence includes the FY2025 comprehensive 10-K/restatement, Q1 and Q2 2026 10-Qs and earnings exhibits, 2026 proxy, 60-month Form 4 corpus, Oklahoma/Hawesville 8-Ks, White House tariff proclamations, DOE project documents, USGS 2026 aluminum summary, Chinese official output and EGA project releases. ROIC.ai and FactorsToday are identified as third-party quantitative sources and reconciled to filings where material. AZI supplied price history; its valuation-percentile endpoint was unavailable because authentication was absent.

APPENDIX A — Standard Diligence Questionnaire

Answers are current through 30 August 2026. “Fact,” “Interpretation,” “Assumption” and “Open question” distinguish the evidence from judgment.

General

What thoughtful questions have other investors asked? The Q2 2026 call focused on the issues that matter: when Mt. Holly and Grundartangi reach stable output; whether Q3’s $325–$345 million adjusted-EBITDA guide includes restart instability; how Century calculates the prospective onshoring benefit; whether Commerce has approved it; how Oklahoma will be financed; and when shareholders can receive cash. Management’s illustrative arithmetic was roughly $3,250/t LME × 25% tariff differential × 300,000 tonnes, or about $244 million annually, but it repeatedly acknowledged that rules/approval remained pending.

The deeper diligence questions are contractual rather than promotional. What is Century’s maximum parent capital/guarantee? Who owns the DOE benefit inside the 60/40 venture? What power price/escalator makes a $4–$6 billion greenfield smelter competitive after 45X fades? Can a project designed to increase U.S. supply preserve the scarcity premium that funds it? Is the Hawesville put genuinely realizable? Will Jamalco’s accounting weakness be remediated before a second complex JV reaches construction?

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: high. Q2 adjusted EBITDA attributable to Century was $326.9 million and the Q3 guide midpoint annualizes to $1.34 billion, versus FY2021–FY2025 adjusted EBITDA averaging approximately $221 million. H1 2026 LME/Midwest/European premiums were $3,386/$2,406/$488 per tonne versus $2,538/$855/$241 a year earlier. Full production can keep earnings above the old average, but current U.S. pricing is exceptional.

External or internal drivers? External drivers dominate: LME, regional premiums, tariff policy, 45X, alumina/carbon and power. Internal execution determines how many tonnes capture the spread. The 2025–2026 transformer failures, Mt. Holly restart and Jamalco turbine are therefore material, but they did not create the $1,551/t year-over-year increase in H1 Midwest premium.

How stable are revenues? Low stability. FY2020–FY2025 revenue was $1.61B/$2.21B/$2.78B/$2.19B/$2.22B/$2.53B. Shipments fell from 784 kt in 2021 to 647 kt in 2025 while revenue rose and fell with price. Customer orders recur, but prices reset with benchmarks.

Outlook and market size? USGS estimates 2025 U.S. apparent aluminum consumption of 5.7 Mt, primary production of 660 kt and 60% net import reliance. Transportation, packaging, building and electrical are the largest end markets. Long-run electrification and lightweighting support demand, while recycling supplies a large and growing share at much lower energy use. Global primary supply remains competitive and state-influenced; U.S. scarcity is stronger than global industry structure.

Domestic or international? Both. Sebree and Mt. Holly capture U.S. Midwest economics; Grundartangi sells into Europe and uses Icelandic power; Jamalco is Jamaican; Vlissingen is Dutch. Aluminum pricing and raw materials are global even when premiums/power are local.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Global competition remains intense, and high U.S. rents are beginning to induce supply response. Existing Western smelter closures create near-term tightness, but China continues to grow through utilization/capacity transfer, Gulf disruptions can reverse, recycled supply expands and Oklahoma itself would add 750 kt.

How profitable is Century? Reported 2025 operating margin was 6.3% and reconstructed ROIC about 9.0%. Removing Section 45X lowers 2025 ROIC toward 3.7%. The five-year FCF total was negative $219.6 million. H1 2026 is far more profitable, but sale/insurance/derivative/policy items require separation. Interpretation: present profitability is high; demonstrated through-cycle quality is low-to-middling.

How profitable is the industry? Highly variable. Smelters can earn large scarcity rents when metal/premiums rise faster than power/alumina, then operate at losses because continuous high-fixed-cost plants resist shutdown. State-supported capacity and global benchmark pricing limit sustained returns.

What are the barriers to entry? A 750 kt U.S. plant may need $4–$6B, five-to-six years, 11 TWh/year, permits, labor, anodes/alumina, grid investment and customer qualification. These are high barriers. They protect incumbent assets from rapid local duplication; they do not grant CENX global pricing power.

Can foreign low-cost production undermine it? Yes, subject to tariffs/freight. Canadian, Gulf, Russian, Indian, Chinese/offshore and other power-advantaged metal set the marginal global curve. Policy currently makes imported metal expensive in the U.S.; that is a government barrier rather than a proprietary one.

Do brands matter? Only modestly. Natur-Al’s low-carbon attributes, alloy quality and delivery reliability can qualify Century for business and modest premiums. Peers offer competing low-carbon/billet products. Century does not disclose a brand premium sufficient to demonstrate customer captivity.

Nature of competition and switching cost? Competition is benchmark price plus regional/product premium, reliability, form and carbon. Qualification creates friction, especially for molten/high-purity/special-alloy uses. Glencore and industrial buyers can source globally; switching costs are far below those of a proprietary platform.

Is the business understandable? Yes. It converts bauxite/alumina, carbon and electricity into benchmark-priced metal. Complexity lies in derivatives, joint ventures, policy credits and continuous-process outages—not in the product.

Financial Condition & Balance Sheet

Assets not fully recognized? Grundartangi’s renewable-power position, U.S. permitted sites, the contingent 6.8% Hawesville data-center interest and tax NOLs may exceed accounting value. The $1.545B federal NOL has a valuation allowance and depends on future taxable income. Hawesville’s interest was Level 3 and should not be assumed equal to the $300M contingent put.

Off-balance-sheet or senior claims? At year-end pensions/OPEB were underfunded by approximately $97.6M; June noncurrent ARO was about $74.3M; operating leases, power commitments and conditional environmental obligations also matter. Oklahoma may create future funding/guarantee claims, but final contracts are not disclosed. Do not count hypothetical project obligations until signed; do not ignore them in scenario analysis.

How conservative is accounting? Mixed. The Hawesville and Jamalco gains follow GAAP but distort net income. Management adjusted EBITDA excludes SBC and numerous outage/restart items, which can overstate normalized cash if disruptions recur. More seriously, prior Jamalco consolidation was incorrect, leading to restatement, an adverse ICFR opinion and an unremediated material weakness at June 2026.

How capex-hungry? Structurally high and lumpy. Cash capex was $82–$100M annually in 2021–2025, then $134.4M in H1 2026. 2026 guidance is $180–$190M, largely insured Iceland repair, Mt. Holly restart and Jamalco. Historical “ongoing” capex could be $20–$50M, but pot relining, transformers, environmental work and project spend make a smooth maintenance assumption unsafe. Oklahoma’s gross cost may be $4–$6B.

Liquidity and refinancing? June liquidity was $784.9M, cash $343.4M and debt $480M. July receipts put cash above debt by management’s statement. The main notes mature in 2032; converts mature in 2028. Near-term refinancing/covenant risk is low, though the convert is dilutive.

Capital Allocation & Management

How much FCF, and how is it used? Five-year FCF was negative $219.6M; 2025 was positive $84.8M and H1 2026 $101.6M. Cash funded restarts, casthouse/power projects, debt management, Jamalco and working capital. No dividend or active buyback returned cash. Current windfall appears reserved for Oklahoma and existing-asset reliability.

Significant acquisitions? Jamalco was acquired in 2023 for $1 plus obligations, producing a $245.9M bargain gain after accounting finalization. It was countercyclical and strategically useful but added operating/control risk. The EGA venture is not yet a completed producing acquisition; it is a proposed 40% greenfield interest.

Asset sales? Hawesville was excellent recycling of an uneconomic asset: $200M cash plus a contingent data-center interest with no funding obligation. The $287.9M gain and $300M put should not be treated as recurring or unconditional.

Buybacks/dividends? No repurchases since April 2015 despite $43.7M nominal authorization remaining; no current dividend and none expected in the foreseeable future under company disclosure. The authorization is stale.

Issuance and dilution? Shares increased roughly 10% from 2020 to 2025, partly because Glencore’s economically participating preferred converted. 2025 SBC was $47M. The 2028 convert can add roughly 4.6M shares at an $18.74 conversion price, subject to capped-call terms. Oklahoma could add further dilution if external funding is insufficient.

Compensation? The CEO’s 2025 annual incentive paid at 136.4% of target and PSUs at 200%. Metrics include safety/operations/adjusted financial and TSR measures but no clean through-cycle ROIC, absolute FCF or per-share FCF hurdle. The design can reward commodity/policy beta and adjusted EBITDA even if a future project earns below WACC.

Management motivations? CEO Jesse Gary has industry/legal/operating experience and is pursuing domestic scale at a rare policy window. That is strategically understandable. Management also receives higher absolute compensation/status from a larger asset base; only disclosed per-share return gates would neutralize that ordinary principal-agent risk.

Insider evidence? The 119-file Form 4 corpus contains one stale open-market purchase: $130,500 in May 2023. Ex-Glencore officer sales totaled about $34.6M over five years. CEO sales of 300,000 shares/$15.6M in 2026 were under a 10b5-1 plan; several other executive sales were discretionary. Glencore sold $599.1M in two strategic blocks and remains at 29.99%. Overall signal: not a bullish insider-confirmation set, but planned/strategic sales must be separated from conviction sales.

Governance? Glencore is not disclosed as formally controlling but has a board-nomination right, pre-emption rights, roughly 30% voting ownership and large commercial relationships. Independent related-party review is crucial. The accounting weakness and EGA venture raise the audit committee’s workload.

Valuation & Market Data

Security type? CENX is a NASDAQ-listed U.S. corporation, not an ADR, MLP or K-1 issuer. It pays no dividend.

Current statistics? At the 28 August $46.59 close, ROIC.ai market capitalization/EV were approximately $4.55B/$4.82B and diluted EV $4.95B. Vendor TTM EBITDA was $761M, giving 6.34× EV/EBITDA; filing-derived TTM levered FCF was about $151M, a 3.3% equity yield. Q3 guide annualizes to far more than vendor TTM EBITDA, but represents current spot/policy conditions.

Net income versus cash? Divergence is substantial. FY2024 net income to Century of $336.8M included a $245.9M bargain gain while OCF was negative $24.6M. H1 2026 net income to Century was $586.8M and OCF $236.0M, with the $287.9M Hawesville gain, insurance and derivatives in the bridge. Cash flow and normalized operating income are better anchors than EPS.

What is embedded? A $4.82B EV at 8× implies about $603M of sustainable EBITDA; at 7×, about $689M. Current market capitalization requires around $228M annual FCF for a 5% yield, above filing-derived TTM $151M. The market embeds a much higher floor than the five-year record and/or value for onshoring/Oklahoma/Hawesville options.

Comp set? Alcoa is the closest public upstream benchmark. Norsk Hydro adds integrated power/upstream/downstream quality. Constellium and Kaiser are aluminum names but downstream and not apples-to-apples. EGA is strategically relevant but private/state-linked. A single peer multiple is unreliable.

Risks & Downside

What causes a decline? Lower LME/Midwest premium; tariff/45X/import-program change; Gulf/China/domestic supply response; power spike; smelter outage; Oklahoma cost/power/permit/financing failure; Glencore conflict; accounting weakness; recession; dilution; or failure of the Hawesville milestone.

Catastrophic loss? Potline freeze, transformer/fire, residue/tailings failure, major environmental order or prolonged grid loss can create material uninsured cost. Geographic diversification and insurance reduce companywide impact. A project overrun combined with a commodity trough is a financial catastrophe even without physical disaster.

Chance of total loss? Low at current liquidity and maturity profile, but not zero over a full project/commodity cycle. A total loss would likely require simultaneous prolonged trough, major physical/environmental event and recourse Oklahoma obligations. The more realistic downside is severe equity impairment and dilution, consistent with historical 80%+ drawdowns.

Recent News & Events

Has the environment changed? Yes. Section 232 rose to 50%, Midwest premiums surged, 45X became material, the onshoring program was created and Gulf disruption tightened metal. Existing U.S. smelters are more profitable and strategic than two years ago.

Significant transactions/projects? Hawesville was sold in February 2026; Century joined EGA’s 750 kt Inola project at 40%; Bechtel is engineering it. Final power, definitive venture documents, financing, permits and FID remain outstanding. DOE’s primary disclosure supports $10M awarded for Phase 1 within up to $500M federal cost share—not $500M received.

Accounting changes? The Jamalco consolidation was restated in March 2026, and the material weakness remained unremediated in Q2.

Facilities/management? Mt. Holly returned its last pots, Grundartangi Line 2 returned near full and Jamalco TG4 came online. No CEO/CFO change alters the thesis. Oklahoma litigation and power arrangements remain active headwinds.

APPENDIX B — Source Appendix

Company: Century Aluminum Company
Ticker / CIK: CENX / 0000949157
Research date: 2026-08-30
Evidence standard: Primary public sources are preferred. SEC filings were read from the local verbatim corpus and linked below to the corresponding public filing. Company and government releases are used for later events or program terms. “Up to” awards, targets, fair-value estimates, and management expectations are not treated as cash received or finalized commitments.

Core SEC filings

1. Century Aluminum Company, 2025 Form 10-K. SEC filing, filed 2026-03-03. Covers the business and plant portfolio; Glencore ownership, customer concentration and related-party transactions; debt, convertible notes, share repurchases, dividends, stock compensation, Jamalco acquisition accounting, capital expenditures, Section 45X credits, and risk factors. SEC filing (accessed 2026-08-30).

2. Century Aluminum Company, Q2 2026 Form 10-Q. SEC filing, filed 2026-08-06. Covers Glencore’s 30.0% ownership and 2026 related-party sales/purchases; the Hawesville sale and contingent equity/put valuation; H1 2026 capex and stock compensation; Section 45X; Inola project status; and current debt and liquidity. SEC filing (accessed 2026-08-30).

3. Century Aluminum Company, 2026 Definitive Proxy Statement. SEC filing, filed 2026-04-24. Covers beneficial ownership, Board independence, Glencore governance and related-party review, 2025 compensation metrics and payouts, ownership/hedging policies, restatement and clawback analysis, and executive/board biographies. SEC filing (accessed 2026-08-30).

4. Glencore plc, Glencore AG and Glencore International AG, Schedule 13D/A. SEC filing, filed 2026-03-04. Reports the March 2026 Rule 144 block sale and subsequent 29,690,702-share / 29.99% beneficial ownership; shared voting and dispositive power; governance agreement, irrevocable proxy, pre-emptive rights and pledged shares. SEC filing (accessed 2026-08-30).

5. Glencore plc, Glencore AG and Glencore International AG, Form 4. SEC filing, filed 2026-03-04. Reports sale of 6,315,245 CENX shares at $51.75 per share and post-transaction holdings. SEC filing (accessed 2026-08-30).

6. Glencore plc, Glencore AG and Glencore International AG, Schedule 13D/A. SEC filing, filed 2025-11-12. Reports the November 2025 Rule 144 block sale, conversion of the remaining Series A preferred stock, and subsequent ownership/governance arrangements. SEC filing (accessed 2026-08-30).

7. Glencore plc, Glencore AG and Glencore International AG, Form 4. SEC filing, filed 2025-11-13. Reports sale of 9,000,000 CENX shares at $30.25 per share and the preferred-stock conversion. SEC filing (accessed 2026-08-30).

8. Century Aluminum Company, Hawesville transaction Form 8-K. SEC filing, filed 2026-02-02. Announces closing of the Hawesville property sale for $200 million cash plus a non-dilutive interest and contractual put right. SEC filing (accessed 2026-08-30). The Q2 2026 Form 10-Q is the controlling source for subsequent accounting and fair-value details.

9. Century Aluminum Company, Section 16 Forms 3/4/5 corpus. SEC ownership filings through 2026-08-30. Used to classify open-market sales, equity grants and tax withholdings and to identify Rule 10b5-1 disclosures. The locally preserved SEC corpus contains 119 Form 4 files for the five-year collection window, matching the filing enumeration; every file parsed without failure. Century ownership filings index (accessed 2026-08-30).

Project and policy sources

10. U.S. Department of Energy, “Award Wednesday: January 15, 2025.” DOE Office of Clean Energy Demonstrations, published 2025-01-15. States that DOE awarded $10 million for Phase 1 of Project ARAL, within a total federal cost share of up to $500 million; Phase 1 includes site selection, initial engineering, environmental review and community/labor engagement. DOE (accessed 2026-08-30).

11. U.S. Department of Energy, “Energy Department Awardee to Build First American Aluminum Smelter Since 1980.” DOE, published 2026-02-10. Describes the EGA/Century Oklahoma project and federal support. The headline award amount is a program ceiling and is read together with DOE’s phase-specific $10 million disbursement disclosure. DOE (accessed 2026-08-30).

12. U.S. Department of Energy, “CX-033530: Project Horizon Green Aluminum Smelter Project – Phase 1.” DOE NEPA determination. Documents the limited Phase 1 activities and environmental-review scope. DOE NEPA (accessed 2026-08-30).

13. Century Aluminum / Emirates Global Aluminium, “Century Joins EGA in Project to Build Smelter in Oklahoma.” Company release, published 2026-01-26. Announces proposed 60% EGA / 40% Century ownership, 750,000 tonnes/year capacity, Inola site, target construction and production timing, and outstanding engineering/power work. Century Aluminum (accessed 2026-08-30).

14. Century Aluminum Company, Q2 2026 results and conference-call materials. Company materials, published 2026-08-06. Current operating context for 2026 capex guidance, Mt. Holly restart, Grundartangi repairs, Jamalco programs, and Inola milestones. Century results and call transcript (accessed 2026-08-30). SEC filings control where figures differ.

15. White House, “Further Strengthening Actions Taken to Adjust Imports of Aluminum into the United States.” Presidential proclamation, issued 2026-07-20. Establishes the project-linked onshoring import framework relevant to the proposed Oklahoma smelter. White House (accessed 2026-08-30).

Additional operating, industry and quantitative sources

17. Century Aluminum Company, Q1 2026 Form 10-Q. SEC filing, filed 2026-05-07. Establishes the first-quarter statements, Hawesville accounting, liquidity, derivatives, capex and Oklahoma project status used in the TTM/H1 reconciliation. SEC filing (accessed 2026-08-30).

18. Century Aluminum Company, Form 8-K non-reliance disclosure. SEC filing, filed 2026-03-03. Describes the Audit Committee’s conclusion that specified historical statements should not be relied upon because of Jamalco consolidation accounting. SEC filing (accessed 2026-08-30).

19. Century Aluminum Company, 2022 Form 10-K. SEC filing, filed 2023-02-27. Primary source for the earlier five-year financial record, Hawesville curtailment/impairment, power and debt before Jamalco. SEC filing (accessed 2026-08-30).

20. Century Aluminum Company, FY2025 results. Company release, published 2026-02-19. Reconciles FY2025/FY2024 company-adjusted EBITDA, earnings, shipments and Q1 outlook. Century Aluminum (accessed 2026-08-30).

21. Century Aluminum Company, FY2023 results. Company release, published 2024-02-21. Reconciles FY2023/FY2022 adjusted EBITDA, shipments, 45X recognition and exceptional items. Century Aluminum (accessed 2026-08-30).

22. Century Aluminum Company, FY2021 results. Company release, published 2022-02-24. Reconciles FY2021/FY2020 adjusted EBITDA and provides realized-price/premium and shipment context. Century Aluminum (accessed 2026-08-30).

23. U.S. Geological Survey, Mineral Commodity Summaries 2026 — Aluminum. Published February 2026. U.S. production, capacity, imports, apparent consumption, end markets, recycling, net import reliance and world production. USGS aluminum data sheet (accessed 2026-08-30).

24. The Aluminum Association, “Powering Up American Aluminum.” Current industry roadmap. Provides U.S. restart potential and estimates of $4–$6 billion, five-to-six years and approximately 11 TWh/year for a new 750,000-tonne smelter. Aluminum Association (accessed 2026-08-30). It is an industry source, so its policy framing is treated as advocacy; engineering magnitudes are used as directional cross-checks.

25. National Bureau of Statistics of China, July 2026 industrial production. Official release, published 2026-08-18. Supplies January–July primary-aluminum production and growth used to test the 45 Mt capacity-cap shorthand. NBS China (accessed 2026-08-30).

26. PRC Ministry of Industry and Information Technology, aluminum-industry implementation plan. Published 2025-03-28. Capacity transfer, energy-efficiency, clean-energy and recycling policy. MIIT (accessed 2026-08-30).

27. International Aluminium Institute, Gulf production disruption update. Published 2026. Current regional output disruption and trade exposure used to contextualize exceptional U.S./European premiums. IAI (accessed 2026-08-30).

28. European Aluminium, “Aluminium Industry.” Current industry page. Directional European primary demand/import dependence; geography/scope limitations are stated in the memo. European Aluminium (accessed 2026-08-30).

29. White House, “Adjusting Imports of Aluminum and Steel into the United States.” Presidential proclamation, issued 2025-06-03. Primary source for the increase of applicable Section 232 aluminum duties to 50%. White House (accessed 2026-08-30).

30. White House, “Strengthening Actions Taken to Adjust Imports of Aluminum, Steel, and Copper.” Presidential proclamation, issued 2026-04-02. Primary source for subsequent coverage/value-base changes in the tariff regime. White House (accessed 2026-08-30).

31. U.S. Treasury/Internal Revenue Service, Section 45X final regulations. Published 2024. Primary legal/administrative context for qualifying advanced-manufacturing production costs. IRS final regulations release (accessed 2026-08-30). Century’s recognized/received amounts come from SEC filings.

32. Emirates Global Aluminium, initial Oklahoma project release. Published 2025-05-16. The earlier 600,000-tonne concept carried an approximately $4 billion investment estimate; later Century/EGA disclosures increased planned capacity to 750,000 tonnes. EGA (accessed 2026-08-30). The older estimate is a scale reference, not a final current budget.

33. Emirates Global Aluminium / Century, Bechtel engineering appointment. Published 2026-02-10. Describes preparatory engineering, value improvement and the target for FID/construction start. EGA (accessed 2026-08-30).

34. Oklahoma Attorney General, Inola case updates. State primary releases dated June and August 2026 regarding the challenge and remand/preliminary-relief posture. Initial release and August update (accessed 2026-08-30).

35. U.S. Department of Justice, statement of interest supporting the Inola project. Published 2026-08-28. Federal position on the litigation/permitting dispute. DOJ (accessed 2026-08-30).

36. SEC Companyfacts, CIK 0000949157. Machine-readable XBRL used only to cross-check statement line items and cover dates against the filings. SEC companyfacts (accessed 2026-08-30).

37. ROIC.ai. Third-party company profile, statement/ratio, enterprise-value, price, news and earnings-call tools, retrieved 2026-08-30. Current CENX EV/market cap and peer EV tables are identified as vendor calculations and reconciled to SEC cash/debt/share counts. The company-news endpoint returned no articles in the requested 2026 window, so material events were built from filings, company releases and government sources.

38. FactorsToday. All Factors model, stock information, leaderboard, stock-specific volatility and related-stocks endpoints; model date 2026-07-31, retrieved 2026-08-30. Loadings are sparse ElasticNet estimates and descriptive, not forecasts. R², factor loadings, returns, drawdowns and similar-stock results in the memo use this source.

39. AZI adjusted daily price CSV for CENX. Downloaded 2026-08-30 through the 2026-08-28 close. Used for the five-year event map, moving averages, drawdowns and raw trailing returns. AZI’s valuation-index endpoint could not be accessed because a valid API token was unavailable; no percentile was inferred.

40. MarketBeat compilation of Nasdaq/FINRA short-interest data. Record date 2026-08-14, published 2026-08-25. Used only for absolute shares short and days-to-cover because direct exchange presentation did not render current values and float denominators vary with Glencore’s stake. MarketBeat (accessed 2026-08-30).

41. Century Aluminum, Hawesville curtailment release. Exhibit 99.1 filed 2022-06-22. Energy-cost rationale and timing used in the five-year event map. SEC exhibit (accessed 2026-08-30).

42. Century Aluminum, initial Section 45X investor disclosure. Filed 2023-12-19. Initial quantified annual-benefit range and qualifying-cost uncertainty. SEC exhibit (accessed 2026-08-30).

43. Century Aluminum, Q3 2024 earnings release. Filed 2024-11-04. Adjusted EBITDA, expanded 45X benefit and pricing bridge used in the event map. SEC exhibit (accessed 2026-08-30).

44. Century Aluminum, Q3 2025 earnings release. Filed 2025-11-06. Q3 result and Q4 adjusted-EBITDA guidance used in the event map. SEC exhibit (accessed 2026-08-30).

Evidence cautions and unresolved items

• The proposed Oklahoma smelter does not yet have a publicly disclosed definitive total cost, Century cash commitment, final joint-venture agreement, final long-term power agreement, financing package, or final investment decision. The $500 million DOE figure is an “up to” federal cost-share ceiling; the primary DOE phase disclosure identifies $10 million awarded for Phase 1.
• Glencore is not disclosed as controlling Century under NASDAQ’s controlled-company framework, but 29.99% voting ownership, a Board nomination right, pre-emptive rights, and sales/purchase relationships make it economically influential. Related-party terms are benchmark-linked, but that does not eliminate counterparty, conflict, route-to-market or concentration risk.
• Form 4 code F transactions are tax withholding and code A transactions are grants; neither is treated as a discretionary market signal. The largest CEO sales in 2026 were disclosed under a Rule 10b5-1 plan adopted 2025-08-20.
• Hawesville’s $300 million put is contingent on the data center meeting the contractual operating threshold and is not equivalent to cash at closing. The retained interest and put are Level 3 fair-value estimates.
• Management’s target dates, expected policy benefits, grant availability, insurance recoveries and project economics are forward-looking unless an executed agreement or cash receipt is specifically disclosed.