# Century Aluminum: $550.7 million of adjusted EBITDA, $236.0 million of operating cash CENX, Nasdaq. Published October 11, 2026. Price as of the October 9, 2026 regular session close. All amounts in US dollars. ## Verdict: HOLD at $36.76 Hold, do not add. At $36.76 the shares trade 12.2% above my Base value of $32.77, and the Bear case sits 87.8% below the price. The market is already paying for a Midwest premium of about $0.73 per pound at an aluminum price of $2,650, against the $0.60 and $2,650 in my Base. The buy zone starts at $14.75. - Bear value: $4.49 - Base value: $32.77 - Bull value: $71.44 - Upside to base value: -10.9% - Maximum buy price: $14.75 - Required margin of safety: 55% - EV / mid-cycle EBITDA after SBC implied by the price: 7.35x (Base uses 6.0x) - EV / annualized Q3 2026 guidance: 2.95x - Mid-cycle free cash flow yield (Base): 8.3% - Business quality: 49/100 - Valuation score: 26/100 - Confidence: 11/25, grade D Century Aluminum reported $550.7 million of consolidated adjusted EBITDA (earnings before interest, taxes, depreciation and amortization, after the company's own adjustments) for the first half of 2026: $227.9 million in the first quarter and $322.8 million in the second. Over the same six months its operations produced $236.0 million of cash, and $134.4 million of that went into capital spending. The gap is not a mystery. The company adds back $151.6 million of costs from the failure of equipment at its Iceland smelter, plus $31.7 million of Mt. Holly restart and emergency power charges, and those were cash costs. Adjusted EBITDA is a ceiling here, not a cash number. At the October 9, 2026 close of $36.76, the shares trade 12.2% above my triangulated Base value of $32.77. The Bear value is $4.49 and the Bull value is $71.44. The required margin of safety (MOS), the discount to fair value I need before buying, is 55%, so the maximum buy price is **$14.75**, about 60% below today's price. The business is earning very large profits at today's prices. The question is how much of that is the metal and how much is a premium that policy can change. In the second quarter, Century's realized US Midwest premium, the extra amount a US buyer pays for delivered metal, was $2,480 per tonne. The aluminum price itself on the London Metal Exchange (LME) was $3,250. The premium was 76% of the metal price (both figures from management's second quarter call). That premium exists because of a 50% US tariff on most imported aluminum, the rate management cited on the call. It is the single most valuable input in this model and the one nobody at Century sets. Route: SF-13 industrial and manufacturing. Subtype: primary aluminum smelter with a consolidated alumina refinery stake, commodity cyclical tier (5x to 8x mid-cycle EV/EBITDA). SF-12 (mining, net asset value of reserves) does not fit because the business transforms purchased inputs and owns no reserves, but I borrow its rule to run three commodity-price scenarios. Economic classification: Standard (2/12). Primary method: mid-cycle EBITDA after stock compensation at 6.0x, built from an aluminum price deck. Secondary method: a scenario cash flow model (DCF) that starts from today's prices and fades to mid-cycle. Cross-check: price to free cash flow (P/FCF). Values are present fair value estimates, not 12 month price targets. ## What Century Aluminum is Century makes primary aluminum, metal smelted from alumina with electricity. It runs the Sebree (Kentucky) and Mt. Holly (South Carolina) smelters in the US and the Grundartangi smelter in Iceland, and it consolidates the Jamalco alumina refinery in Jamaica, where a partner holds a minority share. It sold the Hawesville site in February 2026. In the second quarter it shipped 130,632 tonnes: 95,057 from the US and 35,575 from Iceland. Three features set the economics: - **Price taker.** Revenue is the LME price plus a regional premium. Costs, mainly electricity and alumina, do not move with either. - **One buyer.** Glencore, which also owns 30.0% of the common stock, accounted for about 44.0% of second quarter net sales and 54.0% of 2025 sales. - **Power.** Sebree buys electricity under a market-based contract, Mt. Holly has a power agreement with Santee Cooper through 2031, and Grundartangi uses Icelandic hydroelectric and geothermal power. Management guided a $10 million to $15 million energy headwind in the third quarter from warmer weather. A fourth item is a tax credit. The Section 45X production credit, 10% of certain eligible production costs, reduced cost of goods sold by $26.3 million in the second quarter and $51.3 million in the first half. It sits inside the cost line, so it is inside my cost per tonne below. It is also a law that can change. ## The latest quarter Second quarter net income attributable to Century was $249.3 million, or $2.39 per diluted share. Part of that is not operations: a $40.1 million gain on insurance proceeds and a $7.1 million derivative gain, offset by $76.6 million of Iceland equipment failure costs. Adjusted EBITDA attributable to Century was $326.9 million. The table shows the last four quarters. | $ millions unless noted | Q3 2025 | Q4 2025 | Q1 2026 | Q2 2026 | | --- | --- | --- | --- | --- | | Net sales | 632.2 | 633.7 | 649.2 | 752.1 | | Operating income | 58.3 | 33.0 | 86.1 | 211.6 | | Net income attributable to Century | 14.1 | 2.0 | 337.5 | 249.3 | | Adjusted EBITDA attributable to Century (non-GAAP) | 101.1 | 170.6 | 231.4 | 326.9 | | Shipments, tonnes | 162,442 | 140,257 | 122,865 | 130,632 | | Realized LME price, $ per tonne | | | 2,900 | 3,250 | | Realized US Midwest premium, $ per tonne | | | 2,200 | 2,480 | The first quarter's net income includes a $287.9 million gain on the Hawesville sale, which is not recurring. Adjusted EBITDA over the last four quarters adds up to **$830.0 million**. The first quarter realized prices are derived from the second quarter figures and the increases management quoted ($350, $280), so treat them as calculated, not reported. The realized prices come from a machine-generated transcript of the call, not from the earnings release, which gives neither. Guidance for the third quarter is adjusted EBITDA attributable to Century of $325 million to $345 million. Management's bridge from the second quarter: a lagged LME price of $3,330 and a Midwest premium of $1.09 per pound add $5 million to $10 million, volume and mix add $15 million to $25 million, energy and raw materials take away $15 million to $20 million, and operating expenses are flat. Realized hedge settlements, a $20 million to $25 million headwind, and taxes of $10 million to $15 million sit below adjusted EBITDA. Annualized, the midpoint is $1.34 billion, which is more than the last four quarters combined. Operations moved in the right direction. The final 90 pots at Mt. Holly restarted in late June, on time and on budget, which management says raises total US aluminum production by nearly 10%, though it also reported "some instability" after the restart that it expects to clear in the third quarter. Grundartangi Line 2 restarted at the end of July, about six months ahead of schedule, and runs at slightly reduced amperage until replacement transformers arrive in the fourth quarter. Insurance paid $46 million in the second quarter and $19 million in July for lost Iceland margin. ## Why a price deck, and what the cost per tonne is A smelter's profit is a spread: what the metal sells for minus what it costs to make. The framework asks for mid-cycle earnings, not peak earnings, so I do not capitalize $1.34 billion. I build EBITDA from three prices (the LME price, the Midwest premium and the European premium), a cost per tonne and a volume. I calibrated the cost on the two quarters where I have prices. For each, I took the weighted selling price (LME plus premium, weighted by the US and Iceland share of tonnes) and subtracted consolidated adjusted EBITDA per tonne shipped. The remainder is the all-in cost per tonne, net of alumina and other revenue. | Cost calibration, $ per tonne | Q1 2026 | Q2 2026 | | --- | --- | --- | | LME price | 2,900 | 3,250 | | US share of tonnes | 72% (estimate) | 72.8% | | Weighted LME plus premium | 4,571 | 5,178 | | Consolidated adjusted EBITDA per tonne shipped | 1,855 | 2,471 | | Implied all-in cost per tonne | 2,716 | 2,707 | The two quarters agree within $9. I use **$2,750** in the Base, about $40 above the 2026 average, because the third quarter outlook already shows an energy and raw material headwind. I could not verify 2025 realized premiums, so I do not calibrate on 2025. Stock compensation, which the company adds back, is a real cost to owners. I deduct it at $35 million a year (my estimate; the 2025 add-back was $47.0 million and the first half of 2026 was $11.9 million). A noncontrolling interest in Jamalco takes about $16 million a year of EBITDA. Cash conversion is a separate check. Over the first half, operating cash flow was 42.9% of consolidated adjusted EBITDA. The difference is the $151.6 million Iceland add-back, $31.7 million of Mt. Holly items, $19.4 million of interest, hedge settlements, taxes and working capital. In 2025, operating cash flow was $183.6 million against adjusted EBITDA of $413.4 million, 44.4%. Until the one-off items stop, I do not treat adjusted EBITDA as free cash. ## Shares, dilution and the debt Century had 98,993,295 common shares outstanding on August 4, 2026. That is up from 93.3 million at the end of 2024, because the preferred stock has now fully converted into common. For valuation I add 1.2 million shares of dilution from share awards and 4.6 million shares from the 2028 convertible notes, both the incremental shares in the second quarter's diluted count, for **104.79 million** diluted shares. The convertible notes carry $86.3 million of principal. At 4.6 million shares, the implied conversion price is about $18.76 (my arithmetic, not a filed term). The notes are well in the money at $36.76, so I treat them as equity and remove their debt. | June 30, 2026, $ millions | Amount | Basis | | --- | --- | --- | | 6.875% senior secured notes due 2032 | 400.0 | Principal, 10-K and 10-Q | | Cash and equivalents | -343.4 | Reported | | Pension and benefit liability (noncurrent) | 92.1 | Reported, no tax shield assumed | | Derivative liabilities less assets | 65.1 | 71.6 less 6.5, reported | | Net debt used in every method | 213.8 | Convertible treated as equity | Restricted cash of $46.3 million is excluded. Finance leases are not in what I could read, so they are not in the bridge. Reported long-term debt is $480.0 million with no current maturities. The first maturity is the $86.3 million convertible in 2028, and the notes run to 2032. Liquidity at June 30 was $784.9 million, and management said cash exceeded total debt at the end of July. The balance sheet is not the problem here. Cash has been repaying debt: $326.1 million of revolver borrowings and $7.8 million of industrial revenue bonds in the first half. One asset sits outside the bridge. Century sold Hawesville for $200.0 million in cash plus a 6.8% non-dilutive minority interest in the entity developing the site, with a put option at $300.0 million exercisable from the first anniversary of data center operations starting. Management expects the site to be energized in the second half of 2027. The tenant is Anthropic, which signed a 20-year lease; Anthropic also makes the AI model used in this site's research tools, so I flag the link and give it no weight. The value of the put depends on who must pay it, which I could not find. I discount the $300.0 million at 12% for 2.25 years ($232.5 million) and credit **50%** of that, $116.2 million or $1.11 per share, in the Base, nothing in the Bear and all of it in the Bull. The larger capital question is Oklahoma. Century has a 40% stake in a joint smelter with Emirates Global Aluminium, 60%, for 750,000 tonnes a year at a reported cost of about $4 billion. Management targets a final investment decision by year-end 2026 and first metal by the end of 2029, and says the final power contract is still under negotiation and a $500 million Department of Energy grant is secured and pays out dollar for dollar against project spending. Century's dollar commitment is not disclosed in what I read. A pro rata share would be about $1.6 billion, or $1.1 billion after the grant (my estimate), which is 3.2 times June 30 cash. I value the existing mills only and assume the project earns its cost of capital, so it adds no value, and I count it as a funding risk, not a cash flow. ## Primary valuation: mid-cycle EBITDA after stock compensation at 6.0x The formula: common equity equals (mid-cycle EBITDA minus $35 million of stock compensation) times 6.0, minus net debt of $213.8 million, plus the credited share of the Hawesville put, divided by 104.79 million shares. The multiple sits in the framework's commodity cyclical tier of 5x to 8x, below its 6.5x midpoint, because Century is a single-commodity price taker with one dominant customer. I held the multiple at 6.0x in all three scenarios, so the price deck alone separates them and the downside is not counted twice. I had no verified peer table, so 6.0x is an assumption, tested from 5.0x to 7.0x below. The price decks are my judgments, not forecasts and not forward curve prices (I did not have the curve). The Midwest premium is the one I weigh hardest. For the Base I assumed a 40% chance that the 50% tariff regime stays and the premium holds near $1.00 per pound, 40% that it is partly relaxed and settles near $0.45, and 20% that it goes away near $0.20, which averages to about $0.62, rounded down to $0.60. The Base LME price of $2,650 is 18% below the second quarter's $3,250, even though management expects a global deficit of about 1 million tonnes in 2026. | | Bear | Base | Bull | | --- | --- | --- | --- | | Operating thesis | Tariff partly relaxed, weaker aluminum | Premium compresses, aluminum below today | Tariff regime persists, aluminum stays firm | | LME price, $ per tonne | 2,400 | 2,650 | 3,000 | | US Midwest premium, $ per pound (per tonne) | 0.35 (772) | 0.60 (1,323) | 0.90 (1,984) | | European premium, $ per tonne | 300 | 350 | 450 | | Weighted selling price, $ per tonne | 2,975 | 3,567 | 4,344 | | All-in cost, $ per tonne | 2,750 | 2,750 | 2,750 | | Volume, thousand tonnes | 700 | 720 | 740 | | EBITDA attributable, $ millions | 141.5 | 572.3 | 1,163.9 | | Less stock compensation, $ millions | -35.0 | -35.0 | -35.0 | | Multiple | 6.0x | 6.0x | 6.0x | | Enterprise value, $ millions | 638.8 | 3,224.0 | 6,773.2 | | Less net debt, $ millions | -213.8 | -213.8 | -213.8 | | Hawesville put credit, $ millions | 0.0 | 116.2 | 232.5 | | Equity per share (104.79 million shares) | $4.06 | $29.83 | $64.81 | Volume is 420 thousand tonnes from the US (second quarter annualized 380 thousand, plus Mt. Holly) and 300 thousand from Iceland (second quarter 2025 shipped 81 thousand a quarter), less a small allowance for the restart; the US share is 58.3%. These are my estimates, and the regional split is why the Base margin per tonne ($817) is lower than the second quarter's $2,471. Each $0.10 per pound on the Midwest premium moves EBITDA by $93 million, or **$5.32 per share**. Each $100 per tonne on the LME price moves it by $72 million, or $4.12 per share, and each $100 per tonne on cost does the same in the other direction. Sensitivity of the Base primary value, by multiple and Midwest premium (LME $2,650): | Multiple | $0.30/lb | $0.45/lb | $0.60/lb | $0.75/lb | $0.90/lb | | --- | --- | --- | --- | --- | --- | | 5.0x | $11.5 | $18.1 | $24.7 | $31.3 | $38.0 | | 5.5x | $12.7 | $20.0 | $27.3 | $34.6 | $41.8 | | 6.0x | $13.9 | $21.9 | $29.8 | $37.8 | $45.7 | | 6.5x | $15.2 | $23.8 | $32.4 | $41.0 | $49.6 | | 7.0x | $16.4 | $25.7 | $35.0 | $44.2 | $53.5 | And by LME price and Midwest premium at 6.0x: | LME, $ per tonne | $0.30/lb | $0.45/lb | $0.60/lb | $0.75/lb | $0.90/lb | | --- | --- | --- | --- | --- | --- | | 2,300 | -$0.5 | $7.5 | $15.4 | $23.4 | $31.3 | | 2,500 | $7.8 | $15.7 | $23.7 | $31.6 | $39.5 | | 2,650 | $13.9 | $21.9 | $29.8 | $37.8 | $45.7 | | 2,800 | $20.1 | $28.1 | $36.0 | $44.0 | $51.9 | | 3,000 | $28.4 | $36.3 | $44.3 | $52.2 | $60.2 | The grid is the model. The current price of $36.76 sits on the line between "premium holds near $0.75" at an LME price of $2,650 and "LME near $2,800" at a $0.60 premium. A negative cell means the equity is worth nothing on that deck before any other consideration. ## Secondary valuation: scenario cash flow model The primary method capitalizes mid-cycle earnings and gives no credit for the strong years ahead of the mid-cycle. The cash flow model fixes that. It starts from today's prices and fades them to the same mid-cycle decks over three years. Fourth quarter 2026 uses the third quarter guidance midpoint of $335 million of EBITDA, less stock compensation, interest, tax and capital spending, for $252 million of free cash flow. After that: | Base case, $ millions | 2027 | 2028 | 2029 to 2031 | | --- | --- | --- | --- | | LME, $ per tonne | 2,950 | 2,750 | 2,650 | | Midwest premium, $ per pound | 0.95 | 0.75 | 0.60 | | European premium, $ per tonne | 450 | 400 | 350 | | Volume, thousand tonnes | 710 | 720 | 720 | | EBITDA attributable | 1,126 | 798 | 572 | | Free cash flow after stock compensation, interest, tax, capex and pension | 833 | 525 | 321 | Other inputs, all my estimates: stock compensation $35 million, cash interest $38 million, maintenance capital spending $90 million (2025 total capex was $98.8 million; the split between sustaining and growth is not disclosed), pension cash $15 million, and cash tax of 12%, 15% and then 18% of pretax profit, low at first because the US deferred tax assets still carry a full valuation allowance. Hedge settlements are left out of the cash flows because the $65.1 million derivative liability is already in net debt. The discount rate is **11%**, below the roughly 13% that a simple build with the observed beta of 2.08 would give, because the cash flows are partly contracted and the balance sheet is light. Terminal growth is 1%. The terminal value is 47% of the enterprise value. Oklahoma capital spending is excluded. The Base enterprise value is $3,993 million, less net debt of $213.8 million, plus the $116.2 million put credit, divided by 104.79 million shares: **$37.17**. In the Bear, free cash flow turns negative after 2028, so I set the terminal value to zero instead of capitalizing a loss forever, on the assumption that management would curtail. That gives $5.14. The Bull gives $81.38. | Base DCF per share | g = 0% | g = 1% | g = 2% | | --- | --- | --- | --- | | Discount rate 9% | $41.6 | $44.5 | $48.3 | | 10% | $38.2 | $40.5 | $43.3 | | 11% | $35.4 | $37.2 | $39.4 | | 12% | $33.0 | $34.5 | $36.2 | | 13% | $31.1 | $32.2 | $33.6 | A cross-check on cash: mid-cycle free cash flow in the Base is $320 million ($3.05 per share), built from $572 million of EBITDA less $35 million of stock compensation, $38 million of interest, $75 million of tax, $90 million of capex and $15 million of pension cash. At 10x to 12x, that gives $31.65 to $37.75 per share, with $34.70 at 11x. It is not independent of the price deck, so it carries no weight, but it shows the Base is not a product of the EBITDA multiple alone. At $36.76 the stock trades at 12.0x mid-cycle free cash flow. ## History and growth diagnostics History does not help much, and I give it no weight. An aggregator shows year-end EV/EBITDA of 10.9x to 16.3x from 2021 to 2025, but those are on depressed earnings, with trailing multiples inflated at the bottom of the cycle, and the same source shows 7.65x today. On the company's adjusted EBITDA, the stock trades at 4.76x the last four quarters ($830.0 million) and 2.95x annualized third quarter guidance. | $ millions | Operating cash flow | Capex | Free cash flow | Per share | | --- | --- | --- | --- | --- | | 2024 | -24.6 | 82.3 | -106.9 | -$1.15 | | 2025 | 183.6 | 98.8 | 84.8 | $0.90 | | First half 2026 | 236.0 | 134.4 | 101.6 | $1.03 | | Base mid-cycle, annual | | | 320 | $3.05 | | Annualized Q3 guidance, not used | | | 949 | $9.06 | Free cash flow growth quality (the framework's SM-20 check) is a rebound from a negative base, so a growth rate is not meaningful. Classification: cyclical rebound from a low base, with a low-reliability forecast because the whole path depends on a price deck. Per-share figures use weighted basic shares for 2025 and the first half of 2026, year-end shares for 2024, and 104.79 million diluted for the Base. At the Base free cash flow, the yield on today's market value of $3,852 million is 8.3%. At the annualized third quarter run rate it would be 24.6%, which tells me the market does not believe the run rate lasts. Stress test: if Base EBITDA were 20% lower, free cash flow would be $226 million ($2.15 per share, a 5.9% yield); at 40% lower, $132 million ($1.26, 3.4%). In the primary method those cuts give $23.28 and $16.73 per share. The growth-adjusted price to earnings check (PEG, SM-19) is not meaningful. Trailing GAAP earnings per share of $6.02 include the $287.9 million Hawesville gain and $73.1 million of insurance gains, and growth off a cyclical rebound is not a growth rate. Cyclical distortion: high. It does not set the Base, the MOS or the verdict. ## Valuation triangulation The two methods differ by 21.9% in the Base, which is inside the 20% to 30% warning band. The gap has a clear cause: the cash flow model gives credit for 2027 and 2028 earnings that sit well above mid-cycle, and the primary method does not. I weight the primary method 60% because it is the framework's engine and rests on fewer assumed years. I weight the cash flow model 40% because 47% of its value is terminal and its explicit path is my construction. A 50/50 split would give $33.50 and a 70/30 split $32.04, so the weighting moves the Base by less than $1. | Method | Bear | Base | Bull | Weight | | --- | --- | --- | --- | --- | | Primary: mid-cycle EV/EBITDA after SBC | $4.06 | $29.83 | $64.81 | 60% | | Secondary: scenario cash flow model | $5.14 | $37.17 | $81.38 | 40% | | Divergence | 23.5% | 21.9% | 22.7% | | | Cross-check: P/FCF at 10x to 12x | | $31.65 to $37.75 | | 0% | | Historical multiples | not usable, 0% weight | | | 0% | | Adopted | $4.49 | $32.77 | $71.44 | | The divergence formula is the gap between the two values divided by their midpoint. All three scenarios share the same net debt and put credit, and the cash flow model's later years are the primary method's mid-cycle deck, so their agreement is partly built in. It is a warning, not a pass. Confidence stays at grade D regardless. ## What is the market already pricing in? The diluted market value is 104.79 million shares times $36.76, or $3,852 million. Adding net debt of $213.8 million and subtracting the Base put credit of $116.2 million leaves an operating enterprise value of **$3,950 million**. | Measure | Implied by $36.76 | Base | | --- | --- | --- | | Operating enterprise value, $ millions | 3,950 | 3,224 | | EV / mid-cycle EBITDA after SBC | 7.35x | 6.0x | | EV / mid-cycle EBITDA before SBC | 6.90x | 5.63x | | Mid-cycle EBITDA after SBC needed at 6.0x, $ millions | 658 | 537 | | Midwest premium needed at LME $2,650, $ per pound | 0.73 | 0.60 | | LME price needed at a $0.60 premium, $ per tonne | 2,818 | 2,650 | | EV / last four quarters adjusted EBITDA ($830.0 million) | 4.76x | | | EV / annualized Q3 guidance midpoint ($1,340 million) | 2.95x | | The market does not capitalize the current run rate: at 2.95x annualized guidance it is assuming that most of today's spread goes away. But it is not assuming enough going away to leave a margin of safety. It already expects mid-cycle EBITDA about 22% higher than mine, or a premium about 13 cents a pound higher, or an aluminum price about $170 higher. The same price could also reflect a lower discount rate or credit for Oklahoma, so this is not a unique forecast. What it rules out is the idea that the market is ignoring the cycle. ## Bear, base and bull scenarios The scenario table above holds the details. Adopted values after triangulation: | | Bear | Base | Bull | | --- | --- | --- | --- | | Primary value | $4.06 | $29.83 | $64.81 | | Cash flow model value | $5.14 | $37.17 | $81.38 | | Adopted value per share (60/40) | $4.49 | $32.77 | $71.44 | | Versus $36.76 | -87.8% | -10.9% | +94.3% | These are scenarios, not probabilities, and not one-year targets. I publish no probability-weighted return because I cannot support weights for a policy decision. The Bull case is roughly what today's market prices would deliver if they lasted: the third quarter run rate of $1.34 billion is above the Bull's $1.16 billion. The Bear still leaves a company with a positive equity value, because the balance sheet is light, so it measures a loss of earnings, not a solvency event. One assumption at a time on the Base primary value of $29.83: | Test | Primary | Cash flow model | | --- | --- | --- | | Base | $29.83 | $37.17 | | Cost per tonne +$100 / -$100 | $25.71 / $33.96 | $31.7 / $42.7 | | Stock compensation $50 million / $25 million | $28.98 / $30.41 | $36.41 / $37.94 (at $45 and $25 million) | | Hawesville put credit 0% / 100% | $28.73 / $30.94 | $36.06 at 0% | | Pension and derivative liability not deducted | $31.33 | | | Convertible settled in cash at par, not in shares | $30.34 | | | Diluted shares 110 million | $28.42 | | | Maintenance capex $110 million / $130 million | | $35.33 / $33.48 | | Midwest premium $0.30 per pound | $13.94 | | The bear case to take seriously is not the $4.49. It is the Midwest premium falling to $0.30 per pound with the aluminum price unchanged, which leaves $13.94 on the primary method, 62% below the price, in a world where the company still earns about $295 million of EBITDA. Tariff changes do not need a recession to do that. ## How the buy price is set A Confidence score of 11/25 is grade D, which starts the MOS at 50%. I add 5 points for cyclicality, because the earnings are a spread between two prices that Century does not control, and 5 points for concentration, because one counterparty buys about half of the output and owns 30% of the company. I subtract 5 points for the balance sheet, with $784.9 million of liquidity and cash above debt in July. The method gap of 21.9% is under the 30% threshold, so it adds nothing. The result is **55%**, under the 60% ceiling, so a price is issued. Maximum buy price: $32.77 times (1 - 0.55) = **$14.75**. That is about 60% below today's price. At $14.75 the price already covers a Midwest premium near $0.30 per pound on the primary method ($13.94), so a buyer there is paid for the policy reversal in advance. The 52-week low is $25.40, and $14.75 is a price the shares have not traded at in the last year. Hold is the correct answer at $36.76, and it does not mean the business is poor. It means the price pays for it in full. ## Dividend and capital return Century pays no dividend and I assume none. Management ranked its priorities as liquidity, sustaining capital, then organic growth such as Oklahoma, and said capital returns are under consideration but announced nothing. No buyback is credited. The 2025 refund of the Section 45X credit, $94.3 million received in July, and the insurance receipts of $19 million in July are cash that will show up in the third quarter balance sheet. I do not credit them in net debt, which is as of June 30. Cash built in the third quarter (guided adjusted EBITDA of $325 million to $345 million, less capital spending, hedge settlements and interest) is also not credited until the November report; I count it as upside not yet in the Base. ## Reasons to own CENX - **Earnings at today's prices are enormous.** Third quarter guidance annualizes to $1.34 billion of adjusted EBITDA, 35% of the diluted market value, and the Bull case is not far from it. - **A scarce position.** Century has US smelting capacity in a market where imports face a tariff, and it is one of two partners in the first new US smelter in decades, with up to 300 thousand tonnes a year of imports at a 25% tariff rate from 2027 under the July 20 executive order, according to management. - **The operating restarts worked.** Mt. Holly's last 90 pots restarted on time and on budget, and Grundartangi Line 2 was back about six months early. - **A light balance sheet.** Net debt is $213.8 million after pension and hedge liabilities, with no maturity before 2028. - **Optionality outside the Base.** The Hawesville put at $300.0 million is credited at 50%, and the 10-Q says a release of the US tax valuation allowance is reasonably possible within 12 months, which would cut book tax expense but not cash. The Bull case of $71.44 is above the 52-week high of $70.43. It does not need heroic assumptions; it needs today's prices to last. ## What could go wrong? - **Thesis breaker: the Midwest premium falls.** A $0.30 premium at an LME price of $2,650 leaves $13.94 on the primary method. The premium is a function of a tariff that policy can change. - **Aluminum price.** Each $100 per tonne is $72 million of EBITDA. At $2,300 and a $0.60 premium the primary method gives $15.4. - **Oklahoma funding.** A pro rata share of about $1.6 billion before the grant is 3.2 times the cash on hand. The final power contract is not signed. A funding plan that dilutes shareholders or loads the balance sheet would change the character of the company. - **Customer concentration.** Glencore bought 54.0% of 2025 sales and owns 30.0% of the shares. - **Operations.** Iceland had an equipment failure that cost $151.6 million in the first half, Mt. Holly shows "some instability" after the restart, and Jamalco still has lower quality bauxite for a couple more quarters. - **Energy and tax credits.** The Section 45X credit is $51.3 million of cost reduction in the first half and depends on law. Power for Sebree is priced to the market. - **Hedges.** The derivative liability is $65.1 million net, and realized settlements cost $30 million in the second quarter. - **Data limits.** Realized prices come from a machine transcript, the finance leases and hedge book are not in what I read, and the price decks are mine. I assign no probability to any of these, and the price does not prove they are all discounted. ## Management execution The record is good on projects and mixed on everything else. The Mt. Holly restart finished on time and budget and Grundartangi Line 2 came back ahead of schedule. First quarter guidance of $215 million to $235 million for adjusted EBITDA was met at $231.4 million. Against that, Iceland lost two quarters of production to an equipment failure, the company added back $151.6 million of costs in a half-year, stock compensation of $47.0 million in 2025 is large relative to earnings, the share count rose 6% as preferred converted, and the Oklahoma deal ties up capital before first metal in 2029. The Hawesville sale, $200.0 million of cash plus a put on the retained stake, is a clean piece of capital allocation. **Assessment: 10/20, mixed. The next test is the third quarter report, expected November 5, 2026, which should show a cash build, Mt. Holly stability and a funding plan for Oklahoma.** ## Stock Analyza scorecard | Economic classification (industrial criteria) | Score | | --- | --- | | Normalized return on invested capital | 0/2 | | Revenue growth, 5 years | 0/2 | | Free cash flow per share growth, 5 years | 0/2 | | Balance sheet | 1/1 | | Unit economics | 0/1 | | Recurring or repeatable economics | 0/2 | | Reinvestment runway | 1/2 | | Total: Standard | 2/12 | A zero for missing proof is an evidence penalty, not a claim that returns are zero. I have no verified five-year return or growth series, and the one clean year-on-year comparison (net sales $2,220.3 million in 2024, $2,527.9 million in 2025) is a cyclical rebound. Business quality is **49/100**: | Business quality component | Score | Reason | | --- | --- | --- | | Moat and pricing power | 7/20 | Price taker; scarce US capacity and cheap Icelandic power help | | Economic returns versus cost of capital | 8/20 | Very strong now, weak across the cycle; 2025 net income was $15.8 million | | Balance sheet | 11/15 | Cash above debt in July, $784.9 million of liquidity, pension and hedges to carry | | Cash and earnings quality | 6/15 | Large add-backs; operating cash is 43% of adjusted EBITDA | | Growth and reinvestment runway | 9/15 | Mt. Holly, Iceland and a 40% share of a new smelter, all with execution risk | | Management and capital allocation | 6/10 | Good project execution, costly outage, large stock pay | | Dilution, governance and minority alignment | 2/5 | 30% holder is also 44% to 54% of sales; share count up 6% | | Total | 49/100 | | The valuation score is **26/100**: | Valuation component | Score | Reason | | --- | --- | --- | | Discount to Base | 4/40 | Price is 12.2% above Base | | Protection against the Bear case | 3/20 | Bear is 87.8% below the price; the balance sheet limits it to an earnings loss | | Agreement between methods | 8/15 | 21.9% gap, in the warning band | | Historical relative valuation | 5/10 | No usable synchronized history, neutral score | | Reverse valuation asymmetry | 3/10 | Price needs mid-cycle EBITDA about 22% above Base | | Data and model quality | 3/5 | Price decks are estimates, some filing sections unread | | Total | 26/100 | | By the framework's price-gap rule (within 15% of Base) the stock is fairly valued, which maps to Hold. The score band for 26 is lower because it also weighs protection against the Bear and how much of the Base the market already assumes. Both readings say the same thing: no discount. ## Confidence score | Confidence | Score | | --- | --- | | Data quality / source provenance | 3/5 | | Predictability | 2/5 | | Valuation robustness | 2/5 | | Accounting transparency | 3/5 | | Scenario dispersion | 1/5 | | Total | 11/25, grade D | Filings are primary and current, but realized prices come from a transcript, finance leases and the hedge book are not in what I read, and the cost and volume run rates are estimates. Predictability is low because the profit depends on two prices and one policy, with a partial offset from a strong balance sheet. Robustness is low because $0.10 per pound moves the value by $5.32, about 16% of Base. Accounting is transparent but heavily adjusted. Dispersion is the widest possible: $4.49 to $71.44. The score is what the components add up to; it does not suggest solvency risk. ## What I would watch from here ### Green - US Midwest premium at or above $0.90 per pound, with the tariff regime unchanged - Lagged LME price at or above $2,900 per tonne - Quarterly adjusted EBITDA at or above $325 million - Implied all-in cost per tonne at or below $2,750 - Cash at or above total debt, with no new debt for Oklahoma beyond the grant - Oklahoma final investment decision with a signed power contract and a funding plan that does not issue shares at a discount - Mt. Holly stable in the third quarter, Grundartangi at full amperage after the fourth quarter transformers - Diluted shares at or below 105 million ### Yellow - Midwest premium between $0.50 and $0.90 per pound - Lagged LME price between $2,400 and $2,900 - Quarterly adjusted EBITDA between $200 million and $325 million - Implied cost per tonne between $2,750 and $2,900 - Net debt above zero but below $250 million, excluding Oklahoma - Oklahoma decision delayed beyond the first quarter of 2027 or power contract unsigned - Mt. Holly instability continuing past the third quarter - Diluted shares between 105 and 110 million ### Red - Midwest premium below $0.50 per pound, or tariff relief that removes the premium - Lagged LME price below $2,400 per tonne - Quarterly adjusted EBITDA below $200 million - Implied cost per tonne above $2,900 - Net debt above $500 million, or an equity raise at a discount to fund Oklahoma - Loss or reduction of the Section 45X credit, or a dispute over the Hawesville put - A new unplanned outage at Iceland or the US smelters - Diluted shares above 110 million On the evidence I have, no line is red today. The Midwest premium ($1.09 per pound outlook), the LME price ($3,330 lagged), quarterly EBITDA ($326.9 million) and implied cost ($2,707) are green, and cash was $343.4 million against $400.0 million of notes with the convertible treated as equity. Yellow today: Mt. Holly instability, and Oklahoma, where the decision is targeted for year-end but the power contract is not final. The next hard date is the third quarter report, expected November 5. ## The three most important thesis breakers 1. The Midwest premium falls toward $0.30 per pound, through tariff relief or new supply. At an LME price of $2,650 the primary value is $13.94, and each $0.10 is worth $5.32 per share. 2. Oklahoma is funded in a way that costs existing holders: new shares at a discount, or debt that leaves the existing smelters carrying the project. A pro rata cost of about $1.6 billion is 3.2 times June 30 cash, and 10 million new shares issued at $20 and spent on a project that only earns its cost of capital would cut the primary value from $29.83 to $28.98. 3. On the upside, the tariff regime lasts, the premium stays above $0.90 and the LME price holds near $3,000. That is the Bull case, $71.44, and it needs both prices to hold for more than a year, not one quarter. ## Adversarial review I built this analysis from the filings and checked it in two passes. Inputs were read from Century's Form 10-Q for the quarter ended June 30, 2026 (income statement, cash, debt, share count, stock compensation, tax credits, Hawesville terms), its August 6 earnings release (shipments, adjusted EBITDA and reconciliation, guidance), its fourth quarter 2025 release and 2025 Form 10-K (annual figures, notes, power contracts), a transcript of the second quarter call (realized prices, third quarter bridge, Oklahoma and tariff comments), and the closing price. The closing price of $36.76 on October 9 differs from a broker quote of $36.72 by 0.1%, so no price warning applies. Every table was recomputed from raw inputs a second time with a separate script, and the figures here reproduce. The first pass produced a different answer, and four things changed it: | Item | First pass | Corrected | | --- | --- | --- | | Primary Base (stock compensation not deducted from EBITDA) | $31.84 | $29.83 | | Cash flow model discount rate (beta 2.08) | 10%, $40.45 | 11%, $37.17 | | Bear: cost +$100, 5.0x multiple, lower volume stacked | negative equity | price deck only, $4.06 | | Triangulated Base | $35.28 | $32.77 | | Upside to Base | -4.0% | -10.9% | | Maximum buy price | $15.88 | $14.75 | Two corrections lowered the Base and one raised the Bear. Because the corrections ran one way, I ran a bias test with the most generous reasonable combination: a 7.0x multiple, the full put credit and a 9% discount rate. That gives a primary value of $36.1 and a cash flow value of $45.6, a triangulated $39.9, which is 8.5% above the price. Even then the buy price would be $18.0, so the Hold verdict does not depend on my conservative choices. The strongest case against my Base is that the tariff is structural and the 1 million tonne deficit supports an LME price near today's, in which case the Base is too low by something close to the Bull. The strongest case for a lower value is that the premium is a policy rent, and a policy rent should not earn a 6.0x multiple at all. The values marked as estimates are: the whole price deck (LME, Midwest and European premiums, with my probability weights for the premium); the 720 thousand tonne volume and the 58.3% US share; the $2,750 cost per tonne; the $35 million stock pay; $90 million of maintenance capex; $85 million to $90 million of depreciation; cash taxes; the 6.0x multiple; the 11% discount rate and 1% terminal growth; the $15 million of pension cash; the 50% put credit and 12% put discount rate; the $18.76 conversion price; the $1.6 billion Oklahoma share; and the Q1 US share of tonnes. Not available to me: the hedge book and LME sensitivity (Item 7A), finance leases, the conversion terms of the 2028 notes, sustaining capex, power contract prices, the put counterparty, a verified peer table, a futures curve, and 2025 realized premiums. Part of the 10-Q was beyond what my reader could load, so the debt note, finance leases and hedge sensitivity were not read directly. Realized prices and the July 20 tariff order come from a machine-generated call transcript and are not independently verified. The weakest points are the premium weights and the cost per tonne at the regional mix. The cost calibration comes from two quarters at a 73% US mix, and the Base has a 58% US mix, so an Iceland cost advantage or disadvantage moves the Base by a few dollars. The publication gate status is PASS WITH WARNING: conditional research, with the price deck an assumption set and not a fact. ## Final verdict: hold at $36.76 Century Aluminum is earning more than it ever has, and the market has noticed: at $36.76 the stock sits 12.2% above my Base and pays for a Midwest premium of about $0.73 per pound against my $0.60. Annualized third quarter guidance is $1.34 billion of EBITDA, and the price is 2.95 times that, so the market is not capitalizing the spot spread. But a margin of safety is about being paid for being wrong, and here the price leaves none. The Bear is 87.8% lower, and the buy zone starts at $14.75. **Verdict: HOLD at $36.76. Triangulated Base value $32.77, downside to Base 10.9%, Confidence D (11/25). Required margin of safety 55%, maximum buy price $14.75.** At 76% of the metal price, the premium is the profit, and it is the one input in this model that nobody at Century sets. ## More analyses on Patreon This analysis is free. Premium members get every in-depth analysis, the full archive and a vote on which stocks get analyzed next: https://www.patreon.com/stockanalyza ## Sources Primary sources include Century Aluminum's [Form 10-Q for the quarter ended June 30, 2026](https://www.sec.gov/Archives/edgar/data/949157/000162828026054308/cenx-20260630.htm) (balance sheet, share count, debt, pension, tax credits, Hawesville terms), the [second quarter 2026 earnings release](https://s25.q4cdn.com/716080898/files/doc_financials/2026/q2/2026-06-30-Q2-Exhibit-99-1-Earnings-Release-Final.pdf) (shipments, adjusted EBITDA reconciliation, guidance, liquidity), the [fourth quarter 2025 earnings release](https://www.sec.gov/Archives/edgar/data/949157/000162828026009831/exhibit99120251231q4earnin.htm) (annual results, cash flow, debt activity) and the [2025 Form 10-K](https://www.sec.gov/Archives/edgar/data/949157/000162828026013788/cenx-20251231.htm) (notes outstanding, power contracts, customer concentration). Realized prices, the third quarter bridge, tariff and Oklahoma comments come from a [transcript of the second quarter 2026 call](https://www.fool.com/earnings/call-transcripts/2026/08/13/century-aluminum-cenx-q2-2026-earnings-call-transcript/). The Oklahoma smelter terms are reported by [Yahoo Finance](https://finance.yahoo.com/news/century-aluminum-cenx-ega-partner-140434431.html) and [alcircle](https://www.alcircle.com/news/ega-and-century-aluminium-set-to-build-first-new-us-smelter-in-nearly-50-years-117055). The Hawesville lease is reported by [Spectrum News](https://spectrumnews1.com/ky/lexington/news/2026/07/06/ai-company-anthropic-commits-to-long-term-hawesville-campus). Closing price, quarterly history and aggregator multiples come from StockAnalysis for [CENX](https://stockanalysis.com/stocks/cenx/) and its [ratio history](https://stockanalysis.com/stocks/cenx/financials/ratios/). The reference market price is the October 9, 2026 regular session close of $36.76. Cash, debt and balance sheet items are as of June 30, 2026, and shares as of August 4, 2026. Third quarter results were not yet available. The price deck, volumes, cost per tonne, multiples, discount rate, stock pay, capital spending, tax rates, put credit and Oklahoma share are analytical estimates, not company guidance. **Disclaimer.** This analysis is a structured fundamental research output, not personalized investment advice. It is provided solely for research and educational purposes and is not financial, legal or tax advice. Century Aluminum earns most of its profit from two commodity prices and a policy-driven premium, depends on one dominant customer, and is considering a multi-billion-dollar smelter project. Its shares can lose most of their value, and investors can lose part or all of their invested capital. Framework: SF-13 Industrial and manufacturing, primary aluminum smelter, commodity cyclical tier. Economic class: Standard. Engine: mid-cycle EV/EBITDA after SBC with a scenario DCF. Confidence: D/11. Data status: PASS\_WITH\_WARNING. Adversarial gate: PASS\_WITH\_WARNING, conditional research only. Version: Master v3.3.