ArcelorMittal S.A. (NYSE: MT) — A Macro Trade in a Value Costume

Report date: 25 July 2026 Price referenced throughout: $66.88 (NYSE close, 24 July 2026) Sector: Materials · Steel (Integrated Steel & Mining) Coverage: First published analysis of this company Reporting regime: Foreign private issuer (Luxembourg S.A.); reports in USD under IFRS; files 20-F and 6-K. No 10-K, 10-Q, DEF 14A or Form 4 exists for this issuer.

This article takes no investment recommendation and no price target. The sole exception is the clearly-labeled Claude’s Take block immediately below, which is the author’s own subjective opinion. Sections 1–15 discuss valuation only as embedded expectations and scenarios.

⚡ Claude’s Take

The author’s own subjective, independent opinion. General information only — not investment advice, and not a recommendation to buy or sell any security. The analysis that follows (sections 1–15) carries no position and no price target.

Call: AVOID-here / not-a-short. Accumulate only toward ~$43–54 (roughly 0.60–0.75x book). Medium conviction. You are being asked to pay an above-base-case price for a 2%-return-on-capital business at the richest valuation in its own history.

The bull story is not stupid, and I want to state it fairly before dismantling it. Europe’s Carbon Border Adjustment Mechanism began charging imports a carbon cost in 2026, and a new tariff-rate quota regime takes effect on 1 July 2026. ArcelorMittal is the largest steel producer in Europe; if imports genuinely retreat, its utilisation, prices and returns all rise together, and that is real operating leverage on 28.4Mt of European shipments. The inflection has already started showing up: EBITDA per tonne went from $116 in Q1 2025 to $131 in Q1 2026, and the Company’s own published Q2 consensus (Visible Alpha, 13 brokers, released 24 July) looks for $2,037M of EBITDA against $1,679M in Q1. Liberia iron ore is setting production records. Net debt is a manageable ~1.4x EBITDA. And management has genuinely retired 38% of the fully diluted share count since September 2020. None of that is fiction.

Here is why I still won’t pay $66.88 for it. First, the reported earnings are substantially an accounting artifact. FY2025 operating income of $3,628M includes a $1,858M one-time acquisition gain — $1,736M of which was a bargain-purchase gain booked on acquiring Nippon Steel’s remaining 50% of the Calvert plant for one U.S. dollar, a forced divestiture in the Nippon/U.S. Steel deal. That single non-cash item is 51% of reported operating profit, it was the auditor’s Critical Audit Matter precisely because its size fell out of a discounted-cash-flow estimate, and it flowed through a 9.97% tax rate. Strip it and the other one-offs and FY2025 underlying operating income was roughly $2.56bn on $61.4bn of revenue — a 4.2% margin — and underlying EPS was about $2.30, not the reported $4.13. Second, the group does not earn its cost of capital and never durably has. FY2025 ROIC was 1.98%; it was 2.27% in 2024 and 2.54% in 2023. Europe is 47% of segment sales and 53% of tonnes and earned a 1.8% reported operating margin. The only genuinely high-return business in the entire group is Mining, at a 24.4% margin on 5% of sales. Meanwhile the Company just issued ten-year money at 5.375% — it is borrowing at more than twice its return on invested capital. Over the last eleven years it has taken roughly $18.4bn of impairments, about 36% of today’s market capitalisation, written off. Over its 20-year life the stock has compounded at 1.66% a year with a 96.3% peak-to-trough drawdown.

Third, and this is the part the market has wrong: this is not a value stock, and the tape proves it. Empirically MT carries essentially zero loading on the Value (−0.03), Momentum (−0.02) and Quality (0.00) style factors. What actually drives it is the dollar (−0.80), mining and materials beta (+0.75) and credit risk (+0.65 to +0.77) — it trades like a high-yield bond with an iron-ore overlay, and its closest statistical comparables are index ETFs, not operating peers. The +101% twelve-month move was dollar weakness, credit-spread compression and sector beta riding a steel-policy trade, not a re-rating of a franchise. And you are buying it after seven consecutive up quarters and +195% from the end-2024 close, at the 94th percentile of its own ten-year valuation history — 99.2nd on sales, 97.8th on book. The reflex that “MT is cheap because it trades below book” is exactly inverted: 0.93x book is the most expensive this company has been on book in a decade, and a business earning a 3.2% underlying ROE should trade far below book. Fair P/B is roughly ROE ÷ cost of equity — on the ten-year median ROE of 7.6% that is ~0.76x, not 0.93x. To justify today’s price the market must underwrite a durable 9–10% ROE, better than anything achieved outside 2017–18 and the 2021–22 windfall. My scenarios land at roughly $29 bear / $54 base / $82 bull: today’s price sits above my base case, which is the definition of no margin of safety.

Two things keep this off the short list and out of AVOID-entirely territory: the policy tailwind is real and could genuinely reset European economics for several years, and the balance sheet is investment-grade with $9.9bn of liquidity, so the bad case is a de-rating rather than an impairment. I’d also flag one governance point that gets no airtime: because the Mittal family has not sold, the ~30% reduction in shares outstanding has mechanically lifted the family stake from roughly 31% to 44.7% — shareholders funded ~$10.9bn of buybacks and one result is that a founding family moved ~13 points closer to outright control without buying a share. Framing: a macro-and-policy beta trade at a record own-history price, dressed as deep value. Conviction medium. Flips bullish if the European segment posts an operating margin sustainably above ~8% for two or three consecutive quarters post-TRQ — that is the number that turns the policy story into a genuine ROE step-up — with the buyback running at full tilt. Flips bearish on the TRQ being diluted, delayed or legally unwound, on ETS free-allowance costs continuing unreformed while CBAM relief lags, or on net debt/EBITDA breaching the 1.5x threshold that contractually halts the buyback. Tag: bought a mill for a dollar, booked half a year’s profit.

📈 Stock Price Action — Five-Year Event Map

ArcelorMittal has made a complete round trip and then some. The stock traded at $29.05 in July 2021, collapsed to a five-year low of $18.73 on 29 September 2022, and has since risen almost without interruption to a five-year high of $71.65 on 4 June 2026, closing at $66.88 on 24 July 2026 — 6.7% off that high, against a 52-week range of $30.04–$71.65. The defining feature of the recent tape is its persistence: seven consecutive up quarters, a +195% move from the end-2024 close of $22.66, and +123% off the 52-week low. The move is a policy-and-macro repricing, not an earnings-driven one — reported earnings per share actually fell through much of it.

Cycle narrative. (1) The 2021 windfall is the reference point that matters: at $19.3bn of EBITDA and a 25.5% ROIC the stock went essentially nowhere, because the market correctly refused to capitalise a once-in-a-generation spread. (2) Europe’s 2022 energy shock did what energy shocks do to an energy-intensive price-taker, and the shares halved. (3–4) The 2023–24 stretch is the honest picture of this business at normal: EBITDA in the $4.8–5.7bn range, ROIC of 2–2.5%, and a share price grinding sideways-to-down. (5) The turn was made in Washington, not Luxembourg — the 50% Section 232 tariff from 4 June 2025 re-rated the entire global steel complex, and note the irony that Section 232 is a net cost to ArcelorMittal of roughly $150M a quarter because it ships slabs into Calvert from Brazil and Mexico. (6–7) The genuine ArcelorMittal-specific catalyst was European: CBAM began pricing carbon into imports and the tariff-rate quota was agreed for 1 July 2026, which management describes as a structural reset; the CFO noted European index prices had moved up “almost EUR 100” with the benefit “not yet in our results.” (8) The most recent leg is confirmation-buying ahead of an as-yet-unreported Q2. Price moves are Fact; every attribution above is Interpretation.

1. Executive Summary

ArcelorMittal is the world’s largest integrated steel and mining company outside China — 55.6Mt of crude steel and 48.8Mt of iron ore in 2025, 34 steelmaking facilities across 14 countries, ~125,554 employees, $61.4bn of revenue. It is also, on the evidence of its own filings, a business that does not earn its cost of capital: FY2025 return on invested capital was 1.98%, the third consecutive year in the low single digits, against ten-year money issued in May 2026 at 5.375%.

The FY2025 accounts require careful handling. Reported operating income of $3,628M includes a $1,858M one-time acquisition gain, of which $1,736M was a bargain-purchase gain recognised on acquiring Nippon Steel’s remaining 50% of the Calvert facility for one dollar — a divestiture forced by the Nippon/U.S. Steel transaction. Adjusting for that gain and for the offsetting one-off charges (a $400M Votorantim purchase-price settlement, a $226M Bosnia impairment, $133M of restructuring, a $61M disposal loss), underlying operating income was approximately $2,557M — a 4.2% margin — and underlying EPS approximately $2.30 against the $4.13 reported.

The segment detail explains why. Europe is 47% of segment sales and 53% of tonnes, and earned a 1.8% reported operating margin. The only genuinely high-return business in the group is Mining, at a 24.4% margin on 5% of sales, reflecting orebody quality in Liberia and Canada. A third meaningful profit stream, the AMNS India joint venture, sits outside consolidation and contributed $635M through the associates line. On any honest reading, ArcelorMittal is an iron-ore miner plus a minority interest in an Indian steelmaker, bolted onto a very large, very low-return European steel utility.

There is a real positive change underway, and it is a policy change. Europe’s CBAM began charging imports for carbon in 2026 and a new tariff-rate quota regime takes effect 1 July 2026; management characterises the combination as structurally resetting European industry economics through higher utilisation. The early evidence is visible — EBITDA per tonne rose from $116 (Q1 2025) to $131 (Q1 2026), and the Company’s published Q2 2026 sell-side consensus looks for $2,037M of EBITDA and $1.06 of EPS. Set against it: ETS phase 4.2 has cut free carbon allowances from January 2026, and ArcelorMittal jointly with thyssenkrupp and voestalpine publicly warned in June 2026 that the current ETS trajectory “risks destroying Europe’s industrial base.” The same management team argues European policy has reset its earnings upward and that European policy threatens its existence. Both can be true; together they define where this company’s economics are actually decided.

Capital allocation is genuinely mixed. The share count is down 29.6% since 2020 (38% fully diluted since September 2020) — a large, real transfer of value. But the buyback was pro-cyclical (spent $5.2bn in 2021, $262M in 2025), was paused for roughly a year, and its restart in mid-2026 was part-funded by selling $667M of Vallourec shares. Against it sit ~$18.4bn of cumulative impairments over eleven years — about 36% of today’s market capitalisation — capex running at 1.47x depreciation, and a flagship European decarbonisation programme whose three EAF projects are credited by management with only ~$200M of incremental EBITDA on roughly 4Mt of capacity. One structural consequence deserves naming: because the Mittal family has not sold, buybacks have lifted its stake from roughly 31% to 44.7%.

Valuation is the crux, and it is not what the surface suggests. At $66.88 the stock trades at 0.93x book, 0.83x sales and ~9.4x trailing EV/EBITDA — figures that read cheap in isolation and are, on this company’s own ten-year history, the 97.8th percentile on book, the 99.2nd on sales and the 94.2nd on a composite basis. A price-taker earning a 3.2% underlying ROE should trade at a deep discount to book; fair P/B approximates ROE ÷ cost of equity, which on the ten-year median ROE of 7.6% gives ~0.76x. Today’s 0.93x embeds a durable 9–10% ROE. Scenario arithmetic lands near $29 bear / $54 base / $82 bull — with the current price above base.

No recommendation and no price target are offered in this article. The framework verdicts are: structurally bad industry; no durable competitive advantage in steel and a genuine but narrow resource advantage in mining; low-quality growth; economics that do not improve with scale; mixed-to-poor capital allocation; and a valuation that has already capitalised the policy improvement it is being bought for.

2. Business Overview

2.1 What the company is

ArcelorMittal S.A. is a Luxembourg-incorporated holding company created by the July 2006 merger of Mittal Steel and Arcelor — it reaches its 20th anniversary on 31 July 2026. It is the largest steel producer in Europe, among the largest in the Americas, and holds a growing Asian position through the AMNS India joint venture. Primary listings are Amsterdam, Paris, Luxembourg and Madrid, with New York registry shares trading as MT. It reports in U.S. dollars under IFRS, which is unusual for a European issuer and helpfully removes translation noise from the headline numbers.

Physical scale in FY2025: 55.6Mt of crude steel production, 54.0Mt of steel shipments, and 48.8Mt of iron ore production (including captive mines), across 34 integrated and mini-mill facilities in 14 countries, selling into approximately 126 countries. Employment is ~125,554. Revenue was $61,352M, down 1.7% year on year, on an average steel selling price of $898/tonne, down 2.3%.

2.2 What it sells

The product range is the full integrated-mill catalogue, which is to say almost entirely commodity and semi-commodity:

• Flat products — slabs (semi-finished), plate, hot-rolled and cold-rolled coil and sheet, hot-dipped and electro-galvanised coated steel, tinplate, and pre-painted/colour-coated coil. This is the automotive, appliance, packaging and energy end of the business and contains what value-added content exists, including advanced high-strength steels and electrical steels for electro-mobility.
• Long products — blooms and billets (semi-finished), bars, wire rod, structural sections, rails, sheet piles and wire products, sold into construction, infrastructure and machinery.
• Tubular products — seamless and welded pipe and tube for energy, construction and automotive.
• Mining products — iron ore as lump, fines, concentrate, pellets and sinter feed, plus coking and thermal coal historically. Iron ore operations span Brazil, Canada (AMMC), Liberia, Mexico, South Africa and Ukraine.
• Downstream solutions — the Sustainable Solutions segment aggregates distribution, construction systems (profiles, panels, façade siding, the Steligence platform) and processing.

Recurring revenue in the software sense does not exist here. What exists instead is repeat volume under annual or semi-annual contract, concentrated in automotive. On the Q1 2026 call the CFO described US automotive contract resets as spread across the year — roughly “30% Q1, 30% from quarter 2 and then the rest 25% is Q3” — while European automotive contracts concentrate at the start of the year. That contracting structure creates lag, not pricing power: it delays the pass-through of both cost inflation and price recovery.

2.3 How the money is actually made — the segment reality

This is the single most important table in the memo, because the group headline conceals the structure completely.

Three observations follow directly and they frame everything downstream.

First, Europe dominates volume and contributes almost no profit. At $28.8bn of segment sales and 28.4Mt of shipments, Europe is 47% of sales and 53% of tonnes, and it produced a 1.8% reported operating margin — roughly $18 of operating profit per tonne shipped. Adjusted for the Bosnia impairment and restructuring it is ~3.1%. The entire bull case for this equity is a bet on this one number improving.

Second, North America’s headline 17.9% margin is an artifact. Of its $2,205M of operating income, $1,858M was the Calvert acquisition gain. Underlying North American operating income was approximately $347M on $12,335M of sales — 2.8% — and the 20-F is explicit that underlying performance actually declined, hit by Section 232 tariff costs and unplanned maintenance in Mexico.

Third, Mining is the quality asset and it is small. A 24.4% operating margin on $3,232M of sales, with iron ore production up 26.5% and shipments up 37.5% as Liberia phase 2 ramped. This is a genuine resource advantage — Liberian and Canadian orebodies cannot be replicated by a competitor’s capital spending — but it is 5% of group sales.

Add the equity-accounted AMNS India contribution ($635M through the associates line; 7.86Mt of shipments and $6,026M of sales on a 100% basis) and the picture resolves: of roughly $2.5bn of genuine FY2025 pre-financing earnings power, Mining and India together account for the large majority, while the consolidated steel business — 47Mt of European, North American and Brazilian tonnes — generates a low-single-digit margin.

2.4 Verdict

A vast, genuinely global, vertically semi-integrated commodity producer whose reported consolidated profitability is a poor guide to its economics. The business is honestly described as three things stapled together: a high-return iron-ore miner (5% of sales), an attractive equity stake in an Indian growth market (outside consolidation), and a very large low-return European steel utility (roughly half of everything) whose profitability is set by trade policy and carbon regulation. The scale is real; the earnings quality is not what the headline suggests.

3. Industry Dynamics

3.1 Structure: the textbook bad industry

Steel is the canonical structurally unattractive industry, and ArcelorMittal’s own returns are the proof rather than the exception. The characteristics are all present at once: an essentially fungible product where hot-rolled coil from one mill substitutes for another’s at a specification level; enormous fixed costs and high operating leverage, so that utilisation rather than price discipline drives marginal behaviour; very high capital intensity with long asset lives that keep uneconomic capacity alive far past the point of rationality; fragmented global supply confronting concentrated, sophisticated buyers in automotive; and chronic structural overcapacity. Published June 2026 analysis of Nucor and Steel Dynamics put OECD excess global capacity at roughly 640M tonnes — more than ten times ArcelorMittal’s entire output — with China exporting a record ~131M tonnes in 2025. That export flood is the marginal price-setter for most of the traded world, and no Western producer influences it.

Against that backdrop, note ArcelorMittal’s demand picture: FY2025 shipments of 54.0Mt were down 0.6% and average selling price was down 2.3%. This is not a growth industry, and the volume line has been flat-to-down for years.

3.2 The profit pool sits with policy, not producers

The decisive industry fact of 2025–26 is that Western steel profitability has been created by trade barriers, not by supply-side discipline or demand growth. Two distinct regimes matter to ArcelorMittal, and they cut in opposite directions.

In the United States, Section 232 is a net cost to ArcelorMittal. Tariffs went to 25% on 12 March 2025 and to 50% from 4 June 2025. For a domestic-only US mini-mill this is pure windfall — the desk’s prior work traced US import share falling from ~22% to ~15% and the hot-rolled-coil-to-scrap spread reaching a 2022 high. ArcelorMittal is positioned differently: it ships slabs into its Calvert finishing complex from Brazil and Mexico, so the tariff taxes its own internal supply chain. The CFO confirmed on the Q1 2026 call that the drag runs at roughly $150M per quarter and that “there is no change there.” Management is “analyzing” a newly published relief framework designed to stimulate US investment and pointedly would not confirm eligibility, saying “the answer is not really a clear yes.” The 20-F attributes part of North America’s underlying earnings decline directly to these tariff costs.

In Europe, policy is the bull case — and simultaneously the bear case. Two mechanisms:

• CBAM (Carbon Border Adjustment Mechanism) began imposing a carbon cost on imported steel in 2026. Management describes it as “proving to be very effective,” with European index prices up “almost EUR 100” since introduction, and — critically — states the benefit is “not yet in our results.”
• The tariff-rate quota (TRQ) tool was agreed and is expected effective 1 July 2026, replacing the prior safeguard regime and materially tightening import quotas. WSJ characterised it as the EU “doubling” steel tariffs. Management expects imports to fall, utilisation and prices to rise, and has prepared restarts of idled blast furnaces at Fos (France) and Dąbrowa Górnicza (Poland) to capture the volume, alongside the new Gijón EAF and expanded Sestao capacity.

The offsetting force is the EU Emissions Trading System. ETS phase 4.2 benchmarks cut free allowances from January 2026, and the timing is adverse: MT accrues the higher carbon cost immediately while the CBAM price benefit lags. The clearest evidence of how seriously the industry takes this came on 17 June 2026, when ArcelorMittal, thyssenkrupp Steel and voestalpine issued a joint public call for “urgent, pragmatic reform” of the ETS, warning that without adjustment the current trajectory “risks destroying Europe’s industrial base,” with Lakshmi Mittal making the case in the Financial Times.

That juxtaposition deserves to be stated plainly, because it is the honest characterisation of this industry: the same management team argues in the same quarter that European policy has structurally reset its earnings upward and that European policy threatens to destroy its industrial base. Both claims are defensible. Together they establish that ArcelorMittal’s European economics — roughly half the company — are determined in Brussels rather than on the shop floor, by mechanisms with no contractual permanence.

3.3 The capital cycle read

Applying the Marathon supply-side lens from the frameworks skill, the signal is unambiguous and negative. In a genuine capital-cycle upturn, high returns follow capacity exit. Here, returns are being lifted by trade barriers while capacity is being added on both sides of the Atlantic: roughly 9–12M tons of new US sheet capacity is landing into a market growing flat-to-2% (per the Nucor/Steel Dynamics analysis), and ArcelorMittal itself is adding 3.4Mt of EAF capacity by end-2026 (taking group EAF capacity to 30Mt), restarting two idled blast furnaces, and pursuing an 8Mt greenfield in Andhra Pradesh within a stated “40Mt vision” for India. Asset growth is accelerating precisely as reported returns improve — the textbook late-cycle configuration that the asset-growth anomaly warns against.

The barriers-to-entry test in the Greenwald framework is equally clear. High capital cost is a barrier to casual entry but not to state-sponsored or incumbent capacity, which is the only kind that matters in steel. There is no customer captivity at the commodity end, no network effect, and no proprietary technology that peers lack — thyssenkrupp, voestalpine, Tata, Nippon, POSCO, Ternium and the Chinese majors all make comparable steel. What genuinely restricts entry into ArcelorMittal’s European position is regulation and trade protection, which is to say the barrier is political and therefore revocable.

3.4 Verdict

Structurally bad industry, currently enjoying a policy-manufactured improvement. Fungible product, chronic global overcapacity of ~640M tonnes, record Chinese exports, powerful concentrated buyers, high fixed costs, and no supply-side discipline. Current profitability is being created by CBAM, TRQ and Section 232 — real, material, and entirely political. The capital cycle is being extended by trade barriers while the industry adds capacity, which is the configuration that historically precedes disappointment rather than follows it. An investor in ArcelorMittal is underwriting a regulatory regime, not an industry structure.

4. Competitive Position

4.1 Naming the moat — or its absence

The temptation with ArcelorMittal is to treat “world’s largest steelmaker outside China” as though size were itself a moat. It is not. In the Greenwald taxonomy there are three genuine competitive advantages — supply/cost advantage, demand-side customer captivity, and economies of scale combined with captivity — and the diagnostic question is always whether the advantage produces a persistent return premium. Let us test each honestly.

Economies of scale: present in form, absent in effect. Scale creates advantage when it is local — when one firm holds a dominant share of a defined market whose fixed costs it can spread more thinly than any entrant. ArcelorMittal’s scale is the opposite configuration: it is globally dispersed across 60 countries and 14 steelmaking nations. A 34-plant global footprint delivers real but modest benefits — raw-material purchasing leverage, group-wide R&D that a single-country mill cannot fund, the ability to shift slabs between regions, and genuine commercial reach into 126 countries. What it does not deliver is local market dominance. A competitor in Poland competes with the Polish plant, not with the group; each ArcelorMittal facility faces regional competitors plant-for-plant on regional economics. Fixed costs in steel are plant-level, not group-level, so group scale spreads relatively little of what actually matters.

Cost advantage: partial, and confined to mining. ArcelorMittal produced 48.8Mt of iron ore against 55.6Mt of crude steel — meaningful but incomplete self-sufficiency, and the group remains a net purchaser of coking coal and scrap. Where the cost advantage is genuine is in the orebodies themselves: the Mining segment’s 24.4% operating margin reflects the quality of Liberian and Canadian (AMMC) assets, and no competitor can replicate a good orebody with capital spending. This is a real, durable, non-replicable advantage — and it applies to 5% of group sales.

Customer captivity: thin, and confined to automotive qualification. Switching costs in commodity flat and long steel are close to zero. They exist meaningfully only in automotive-qualified grades, where re-qualifying a supplier for a safety-critical stamped part takes time and money. ArcelorMittal has genuine standing here — advanced high-strength steels, electrical steels, laser-welded tailored blanks (AMTBA), and a stated intention that the Dunkirk EAF conversion will “produce the same grades as we can today with the blast furnace” specifically to “protect” a “very quality high order book.” That is real. But it is a share of one segment’s volume, the contracts reset annually, and MT’s own CFO would characterise the outcome of US automotive negotiations only as “in line with our expectations.”

4.2 The financial test, which settles it

One standard is decisive here: if a claimed moat cannot be tied to a financial outcome that would deteriorate without it, it is not a moat. Apply it.

A firm with a durable advantage earns a premium return through the cycle, not only at its peak. ArcelorMittal earned 25.5% ROIC in the 2021 windfall and 1.98% in 2025. Its ten-year mean ROE of 12.6% collapses to a median of 7.6% once the 50.2% post-COVID outlier is set aside — and it just borrowed ten-year money at 5.375%. A business whose invested capital returns 1.98% while its marginal debt costs 5.375% is destroying value at the margin, and the ~$18.4bn of cumulative impairments over eleven years is the accumulated bill for exactly that.

Market-share stability, the other Greenwald test, also fails to show a moat: ArcelorMittal’s shipments have been flat-to-declining (54.0Mt, −0.6% in 2025; European crude steel production −6.6%) while Chinese exports hit records. A firm with genuine advantage in a growing niche gains share; a price-taker holds volume and takes whatever spread the cycle allows.

4.3 Versus named competitors

The comparison that matters most is against the US mini-mills covered in June 2026, because it isolates business model from cycle:

The pattern is stark. All three re-rated to the richest valuations in their own histories inside twelve months, on the same policy trade. ArcelorMittal earned roughly one-fifth of Steel Dynamics’ trough return on capital and delivered the largest share-price gain of the three. The structural reasons for the return gap are durable, not cyclical: the integrated blast-furnace route is more capital-intensive, more carbon-exposed (hence ETS-exposed), less flexible in downturns, and geographically anchored in Europe — the world’s highest-cost, most-regulated major steel market — whereas the EAF mini-mill route is flexible, scrap-based and domestically protected.

Against its true peer set — the factor evidence identifies Ternium (TX) as MT’s closest single-stock comparable, alongside European integrateds thyssenkrupp and voestalpine — ArcelorMittal is legitimately advantaged: larger, more diversified geographically, partially ore-integrated, and with a superior balance sheet. It is a strong competitor within a weak cohort.

4.4 Verdict

No durable competitive advantage in the steel operations; a genuine but narrow resource advantage in mining. Group scale is global rather than local and therefore produces purchasing and R&D benefits without market power. Customer captivity is confined to automotive qualification. The financial test is conclusive: 1.98% ROIC against a 5.375% marginal cost of debt, a 7.6% ten-year median ROE, flat-to-declining volume, and $18.4bn of impairments. What protects ArcelorMittal’s European position is trade regulation, not competitive advantage — and the distinction is the entire investment question, because regulation can be withdrawn by the body that granted it.

5. Growth History and Forward Opportunities

5.1 The historical record: no growth, high volatility

Revenue over eleven years shows a commodity price series, not a growth business:

Revenue in FY2025 was 3.5% below FY2015 — a decade of nothing, in nominal terms, on a shrinking share count. EBITDA swung from negative $1.4bn to positive $19.3bn and back to $4.4bn. EPS has printed −$10.55 and +$13.53 within the same decade. Volume tells the same story: shipments were 54.0Mt in 2025, down 0.6%, with European crude steel production down 6.6% on the Dunkirk blast-furnace reline, maintenance and the Bosnia disposal.

There has been essentially no organic volume growth, and what unit growth exists has come from acquisition (Calvert consolidation, Tuper, Tekno, AMTBA) and from mining, where FY2025 iron-ore production rose 26.5% and shipments 37.5% as Liberia phase 2 ramped. Note that Mining is also where growth was highest-quality — it grew and it earned a 24.4% margin.

5.2 The forward opportunity set

Management’s forward case rests on four pillars, and they differ enormously in quality.

(i) The European volume and price recovery — the largest and most credible. If CBAM plus the 1 July 2026 TRQ genuinely reduces imports, ArcelorMittal captures the volume through better utilisation of existing plant, restarts of the idled Fos and Dąbrowa blast furnaces, and the new Gijón EAF and expanded Sestao. Because Europe currently earns a 1.8–3.1% margin on $28.8bn of sales, the operating leverage is enormous: every 100bp of European operating margin is ~$288M of operating income, and moving Europe from 3% to 8% would add roughly $1.4bn. The CFO expects H2 2026 shipments to exceed H1 — explicitly “unusual” against normal seasonality — and expects “a clear improvement in our EBITDA in all Steel segments next quarter.” This is genuine and it is already partly visible in the $131/t Q1 print. It is also entirely contingent on a policy instrument that has been in force for three weeks as of this article.

(ii) Mining expansion — the highest-quality growth. Liberia phase 2 is ramping to a target of at least 80Mt of shipments at full capacity in H2, with AMMC in Canada alongside. Management flagged a $0.2bn payment on signing a new Mineral Development Agreement in Liberia during Q1 2026. This growth carries the group’s best margins and best returns, and is the one place where ArcelorMittal is expanding a genuine advantage rather than a commodity position.

(iii) AMNS India — real but unconsolidated and self-funding-constrained. The joint venture shipped 7.86Mt on a 100% basis in 2025 (sales $6,026M, both down slightly on maintenance and weak H1 pricing). The roadmap is Hazira to 15Mt, then an 8Mt greenfield in Andhra Pradesh, within a stated “40Mt vision.” Management confirmed sequencing has changed — Andhra now precedes further Hazira phases, with Ephrem Ravi of Citi explicitly asking whether Hazira phase 2 was being delayed “in order to balance the balance sheet,” which the CFO effectively conceded (“you’re right… we have to phase it”). India is the single most attractive end-market in world steel and this is a credible long-term asset. But it sits outside consolidation, contributes through a single associates line, and its capital needs compete with the parent’s.

(iv) The EAF decarbonisation programme — growth in name only. ArcelorMittal will add 3.4Mt of EAF capacity by end-2026, lifting group EAF capacity to 30Mt (EAF is now ~26% of production, up from 19% in 2018). Projects: Gijón (long products, first heat Q1 2026, €213M), Sestao (to 1.6Mt, XCarb low-carbon flat), Dunkirk (newly approved, replacing a blast furnace, ~2Mt), plus the commissioned Calvert EAF (1.5Mt) in the US. Management’s total claimed incremental EBITDA from all strategic projects and completed M&A is $1.8bn from 2026 onward. But the three European EAFs specifically are credited with only ~$200M. Jefferies’ Tristan Gresser put the objection precisely on the Q1 call: “I was a bit surprised to see that you were only targeting $200 million of EBITDA for your 3 EAF projects… that’s close to 4 million tonnes of EAF steel. And it does not look like there are much of productivity gains or green steel premiums baked into that.” The CFO’s response leaned on incremental framing, French state support covering 50% of Dunkirk through white certificates, an avoided blast-furnace reline, and an explicit refusal to disclose assumptions because green-steel pricing is “commercially sensitive” — while conceding the green-steel premium market “will be limited.”

That exchange is the most informative two minutes of the call. A company spending materially to convert ~4Mt of capacity, achieving ~$200M of incremental EBITDA, requiring 50% government subsidy to clear its own return hurdle, and declining to publish the assumptions, is not funding growth. It is funding regulatory compliance and labelling it growth. The corroborating evidence arrived a week before the call: the 2025 Sustainability Report cut the 2030 CO₂ reduction target from −30% to −10% — management explaining that the new figure reflects only “announced projects,” which is an admission that the projects required to hit −30% do not clear the return bar.

5.3 Verdict

Low-quality growth, with two genuine bright spots. A decade of no revenue growth and no volume growth; historical unit growth bought rather than built. The credible forward opportunities are the European utilisation recovery (large, high-operating-leverage, and wholly policy-dependent) and mining expansion (smaller, but the group’s best returns and a real advantage being extended). AMNS India is a genuine long-term asset held at one remove and competing for capital. The flagship European decarbonisation programme is, on management’s own disclosed numbers and by its own refusal to disclose more, a compliance cost presented as growth — and the simultaneous downgrade of the 2030 carbon target confirms it.

6. Financial Quality

6.1 The quality-of-earnings problem

Any assessment of ArcelorMittal’s FY2025 must begin by dismantling the reported figures, because the headline is materially misleading.

Reported: operating income $3,628M, pre-tax income $3,602M, net income $3,152M, EPS $4.13, effective tax rate 9.97%.

The single largest component of that result was not earned in the ordinary course. On 18 June 2025 ArcelorMittal acquired Nippon Steel’s remaining 50% of AMNS Calvert — a divestiture required in connection with Nippon’s acquisition of U.S. Steel, approved by the US administration on 13 June 2025 — for cash consideration of one U.S. dollar. Under IFRS, where the fair value of net assets acquired exceeds consideration, the excess is recognised immediately in operating income. The result: an aggregate acquisition gain of $1,858M, comprising a $1,736M bargain-purchase gain and $122M on settlement of pre-existing relationships. That is 51% of reported operating income from a single non-cash entry.

Three features make this worse than a simple one-off:

1. It is an estimate, not a receipt. The size of the gain fell out of a discounted-cash-flow valuation of Calvert’s enterprise value. Deloitte designated it the Critical Audit Matter, describing the audit as “complex” and requiring “a high degree of auditor judgment” because the key assumptions — selling prices, cost of consumed slabs, discount rate — “are forward-looking and could be affected by future regulatory, economic and market conditions,” and because “changes in the assumptions directly impacted the amount of the gain.”
2. It was barely taxed. The 20-F states that “exceptional items, such as bargain purchase in North America segment, had minimal tax impact” — a principal reason the effective rate collapsed from 52.7% to 9.97%. So the item inflated pre-tax income and depressed the tax rate, compounding its effect on EPS.
3. It masked genuine deterioration. The 20-F says that excluding the gain, North American “underlying operating performance declined,” hurt by Section 232 costs and Mexican outages.

Normalising properly requires adding back the offsetting one-off charges too — this is not a one-sided exercise:

And the associates line needs the same treatment. Reported income from associates and JVs was $806M gross (less $123M of equity-method impairment, principally $81M on Tameh). But that $806M itself contained roughly $215M of one-off fair-value remeasurement gains from stepping up to control of previously-held stakes — AMTBA $145M, Tuper $35M, Calvert $13M, Atlas $11M, Perfilor $11M. Genuinely recurring associate earnings were therefore closer to $591M.

Rebuilding EPS from the bottom up: underlying operating income ~$2,557M, plus recurring associates ~$591M, less net financing costs of $709M, gives roughly $2,439M of underlying pre-tax income. At a normalised ~25% tax rate (against a 23.87% Luxembourg statutory rate) and after $91M of minority interests, underlying EPS was approximately $2.30 — some 44% below the reported $4.13. At $66.88, that is roughly 29x underlying 2025 earnings for a business with a 4.2% operating margin.

(A methodological note that matters for anyone re-deriving these figures: aggregator data is unreliable for this issuer. ROIC.ai reports FY2025 operating income of $1,461M and EBITDA of $4,406M; the 20-F reconciliation and segment note both show operating income of $3,628M, and the sum of the Company’s four reported quarterly EBITDA figures is $6,541M. The filing governs. ROIC.ai’s quarterly enterprise-value series is also unusable here — it shows 2Q25 EV of $10,466M against a market capitalisation of roughly $25bn at the time.)

6.2 Cash generation: structurally thin

Where the income statement is flattered, the cash flow statement is merciless.

Two years running, ArcelorMittal has converted a ~$61–62bn revenue base into roughly $450–470M of free cash flow — under 1% of the current market capitalisation. The 2023 figure of $3,032M looks better only because it included a $1,714M working-capital release; on an underlying basis the last three years cluster in the same thin band.

The final row is the one to sit with. In FY2025, EBITDA less capital expenditure covered interest expense 0.12 times. After spending what it must to keep the asset base intact, essentially nothing remained to service $577M of interest — it was funded from the balance sheet. Capex has run at 1.47x depreciation, meaning the asset base is growing in nominal terms while returns on it are at generational lows. That combination — heavy reinvestment, negligible free cash flow, sub-2% ROIC — is the financial signature of a capital-intensive business reinvesting into inadequate returns.

Q1 2026 continued the pattern seasonally: a $1.3bn free cash outflow on a $1.5bn working-capital investment, with capex of $1.3bn. Full-year 2026 capex guidance is $4.5–5.0bn, of which $1.7–2.0bn is “strategic.” Management’s preferred metric — “investable cash flow,” being operating cash flow less maintenance capex — was $2.0bn over the trailing twelve months, against which it spent $1.5bn on strategic capex, $0.7bn on shareholder returns and $0.2bn on M&A. That totals $2.4bn against $2.0bn: the Company is currently outspending even its own favourably-defined cash generation.

6.3 Returns and the value-destruction record

FY2025 ROIC of 1.98% is the third consecutive low-single-digit year (2.27% in 2024, 2.54% in 2023). The cycle high was 25.5% in 2021. Ten-year ROE averages 12.6% but medians 7.6% — the mean being distorted by a single 50.2% post-COVID year. Underlying FY2025 ROE, on the ~$2.30 of underlying EPS, was approximately 3.2%.

Set against a cost of capital that is directly observable: the Company priced $1.0bn of 5.375% notes due 2036 on 12 May 2026. Equity cost is higher still — beta is ~1.33 and lifetime volatility above 50%, so a 10%+ cost of equity is conservative. ArcelorMittal is investing at ~2% and funding at 5.375%+.

The cumulative consequence is visible in the impairment record. Charges recognised in the cash-flow statement, FY2015–FY2025: $6,184M (2015), $4M (2016), $987M (2017), $1,783M (2018), $2,745M (2019), $464M (2020), $1,095M (2021), $1,440M (2022), $2,043M (2023), $708M (2024), $903M (2025) — approximately $18.4bn, or about 36% of today’s market capitalisation, written off over eleven years. These are not accounting abstractions; they are the retrospective admission that capital deployed did not earn its keep. A company that impairs $18.4bn while earning 2% on capital is telling you something structural about its reinvestment opportunity set.

6.4 Balance sheet: genuinely sound, with one large soft asset

This is where the analysis turns positive, and it should be given full weight.

Leverage is modest for a cyclical, maturities are termed out, liquidity is ample at $9.9bn, and the Company retains investment-grade status. Net debt of $6.8bn against $54.5bn of equity is 12.6% — a genuine strength and the main reason the bad case here is a de-rating rather than a solvency event. Interest cover of 7.6x on EBITDA is comfortable.

Two qualifications. First, the seasonal jump matters more than usual right now. Net debt rose from $6.8bn to $9.3bn in a single quarter, putting net debt/TTM EBITDA at approximately 1.40x — against a capital-return policy that states explicitly: “If the ratio of net debt to EBITDA is greater than 1.5x, then ArcelorMittal will not carry out any share buyback.” The restarted buyback therefore sits inside a covenant with roughly 10bp of headroom, and depends on the H2 earnings recovery arriving on schedule.

Second, and more important for anyone anchoring on book value: $8.9bn of the balance sheet is a deferred tax asset. That is 16% of equity ex-minorities and roughly 17% of market capitalisation. The 20-F’s risk factors are explicit that recovering it requires “at least $38.9 billion” of future taxable income at certain operating subsidiaries — much of it, structurally, in the European operations that have been generating 2–3% margins. Stripping the DTA and goodwill/intangibles gives tangible book of roughly $40.3bn, or ~$53 per share, at which the stock trades at ~1.26x rather than 0.93x.

This is a subtle but important point about the bull case: the DTA and the investment thesis are the same bet, counted twice. The book value that makes the stock look cheap is 16% composed of a tax asset that only converts into cash if the European profit recovery that the thesis presupposes actually materialises. If Europe stays at 3% margins, both the earnings case and part of the book value fail together.

6.5 Verdict

Economics do not improve with scale, and reported earnings materially overstate them. FY2025 EPS of $4.13 was approximately $2.30 underlying, with 51% of reported operating income from a non-cash bargain-purchase gain on a plant bought for one dollar. Two consecutive years of ~$460M of free cash flow on ~$61bn of revenue; EBITDA-less-capex covering interest 0.12x; capex at 1.47x depreciation; ROIC of 1.98% against 5.375% marginal debt; $18.4bn of impairments over eleven years. The balance sheet is a genuine strength and is the reason this is not a distressed situation — but 16% of the book value that anchors the “below book” argument is a deferred tax asset contingent on the very recovery being underwritten.

7. Capital Allocation

7.1 The buyback: large, real, and pro-cyclical

The strongest item in ArcelorMittal’s capital-allocation record is the share count.

Shares outstanding fell 29.6% from 2020 to 2025; the Company claims 38% on a fully diluted basis since September 2020. Roughly $10.9bn was returned through buybacks over 2021–2025. On a per-share basis this is genuine value creation and management is right to highlight it — the CFO called the policy “really great” on the Q1 call, and shrinking the count of a cyclical when the cash is there is a defensible strategy.

The criticism is the timing. Spending tracked free cash flow almost mechanically — $5,170M in 2021 when EBITDA was $19.3bn and the stock was near highs, collapsing to $262M in 2025 when EBITDA was $4.4bn and the stock was in the $30s. That is the buy-high, abstain-low pattern. The counterfactual is unforgiving: capital deployed at 2025’s average price would have retired roughly twice as many shares per dollar as capital deployed in 2021. Contrast this with Steel Dynamics, which, as noted in June 2026, buys counter-cyclically — heavy at $80–130 in 2022–23 and throttled to a trickle at $251 in Q1 2026.

Worse, the buyback stopped. An analyst on the Q1 2026 call observed there had been “no buybacks for nearly a year”; the CFO responded that the Company was “close to restart,” would be “free cash flow positive this year,” and saw “no reason why we would not go above the minimum 50%.” The restart duly came: the second tranche (up to 10M shares) of the 2025–2030 programme commenced around 1 July 2026. But the funding is telling — it was part-financed by the 19 May 2026 secondary sale of ~23.9M Vallourec shares (~10% of Vallourec) at €24.00, raising ~$667M, with proceeds explicitly allocated to buybacks. ArcelorMittal bought into Vallourec in August 2024 and sold roughly a third of the stake some 21 months later to fund a return of capital. Selling an asset to fund a buyback is a portfolio reallocation, not earnings being returned. It may well be sensible — Vallourec was a financial stake, not a core operating asset — but it should not be scored as cash-generative capital return.

The dividend is modest and rising: $0.55/share for 2025 ($421M, two instalments), moved to quarterly with a first $0.15 interim paid in March 2026 against a proposed $0.60 annual. Management notes the dividend has doubled over five years. The policy is a progressive base dividend plus a minimum 50% of post-dividend free cash flow to buybacks, with the hard 1.5x net-debt/EBITDA brake discussed above.

7.2 The control question nobody discusses

Per the FY2025 20-F, the Mittal family’s holdings are: Lumen Investments S.à r.l. 275,840,595 shares; Nuavam Investments S.à r.l. 63,658,348; Mr Lakshmi N. Mittal 564,103 directly; Mrs Usha Mittal 25,500 — an aggregate 340,088,546 shares, or 44.7% of the 761,125,819 outstanding. Lakshmi Mittal is Executive Chairman; his son Aditya Mittal is CEO. The 20-F identifies the Chief Operating Decision Maker as “the Executive Office comprising the Executive Chairman, Mr. Lakshmi N. Mittal and the CEO, Mr. Aditya Mittal.” A 2006 Memorandum of Understanding governs the family’s relationship with the Company.

Now combine that with the buyback. Because the family has not sold into it, the ~30% reduction in shares outstanding has mechanically lifted its stake from roughly 31% (340M of 1,081M shares in 2020) to 44.7% today — an increase of roughly 13 percentage points, achieved without the family buying a share, funded by ~$10.9bn of company cash that belonged to all shareholders. (The 2020 percentage is inferred on the assumption that the family’s absolute holding was approximately unchanged; the current 44.7% is read directly from the filing.)

To be scrupulously fair: this is a consequence of buybacks, not evidence of intent, and every non-selling shareholder saw the same proportional increase — that is what a buyback does. Aligned family ownership at scale also has genuine advantages, including a long time horizon and an owner’s aversion to dilution. But the effect is material and one-directional. At 44.7% the family holds effective negative control over any change of control, and the 20-F itself flags that the “Significant Shareholder could also exercise significant influence over a change of control.” Investors should understand that continued buybacks walk this company toward family majority ownership, and price the governance implications accordingly. This is a capital-allocation and control observation, not an ESG framing.

7.3 M&A: disciplined recently, and cheap

The recent M&A record is, unusually, a point in management’s favour. FY2025 net cash paid for acquisitions was just $169M, and the vintage was opportunistic:

• Calvert — the remaining 50% for $1, with $638M injected by Nippon to repay Calvert debt and a $248M shareholder loan forgiven, plus a seven-year Nippon domestic slab supply agreement averaging 750kt/year. Acquiring a 5.3Mt hot-strip mill complex for a dollar because the seller was a forced divestor is genuinely good opportunism, whatever one thinks of the accounting gain.
• Tuper (Brazil pipes, control from 40%), Tekno (Brazil coil coating, $133M, being taken private), AMTBA (tailored blanks, to 90%), Atlas (265MW Brazilian solar JV, $47M).

These are sensible bolt-ons in adjacent value-added and captive-power positions. FY2024 was heavier at $1,352M, principally the Vallourec stake — a third of which has now been sold.

The disciplined counter-example also deserves credit: management terminated the Monlevade expansion in Brazil (writing off $43M of civil works in 2024) rather than complete a project that no longer cleared its hurdle.

7.4 Growth capex: where the doubt lies

The real capital-allocation question is not M&A but the $1.7–2.0bn a year of “strategic” capex. Management claims $1.8bn of incremental EBITDA from all strategic projects plus completed M&A from 2026 onward. That would be an excellent return on a cumulative programme — if delivered.

But the disclosed economics of the flagship European component are poor, as section 5.2 sets out: three EAF projects, roughly 4Mt of capacity, ~$200M of incremental EBITDA, with Dunkirk requiring 50% French state support via white certificates and an avoided blast-furnace reline to clear the hurdle, and management explicitly declining to publish return assumptions. When a sell-side analyst says he is “surprised” the number is so low and the CFO answers with framing rather than figures, and when the 2030 carbon target is simultaneously cut from −30% to −10% on the basis that only “announced projects” count, the reasonable inference is that this capital earns a low return and is being spent because regulation requires it.

Mining capex is the opposite case and should be scored positively: $926M in 2025 into assets earning 24.4% margins, delivering 26.5% production growth.

7.5 Incentives and insider behaviour

Executive compensation structure could not be assessed to the usual standard, and the reason is structural rather than an omission: as a foreign private issuer, ArcelorMittal files no DEF 14A proxy statement. Compensation is disclosed in the 20-F’s “Management and employees—Compensation” section and in Luxembourg governance filings, at a level of granularity well below a US proxy. The 20-F does disclose, creditably, that independent non-executive directors receive no share options, RSUs or PSUs, consistent with Luxembourg’s 10 Principles of Corporate Governance.

Insider transaction direction is unknown, and this is a genuine gap. ArcelorMittal files no Forms 3/4/5 — a five-year EDGAR sweep confirms zero such filings — because FPIs are exempt from Section 16. Insider dealing is instead disclosed under EU Market Abuse Regulation Article 19(3) as “Designated person notifications” lodged with the Luxembourg Stock Exchange’s OAM database. Three such notifications were issued in 2026: 13 May, 27 May and 3 June. The press releases are pointers only and do not state direction or size, and the underlying OAM records were not retrieved for this article. This article therefore makes no inference about insider conviction in either direction — a point worth flagging because the absence of visible insider selling in a stock that has doubled should not be mistaken for absence of selling.

7.6 Verdict

Mixed-to-poor, with the good and bad both substantial. In credit: a genuinely large 38% fully diluted share-count reduction, cheap and disciplined recent M&A (Calvert for a dollar, Monlevade cancelled), high-return mining capex, a rising dividend, and a conservatively managed investment-grade balance sheet. In debit: the buyback was pro-cyclical and its restart is part-funded by an asset sale; ~$18.4bn of impairments over eleven years is the accumulated record of capital deployed below cost; capex runs at 1.47x depreciation into a 2% ROIC; the flagship European decarbonisation programme has disclosed returns close to trivial and undisclosed assumptions; the 2030 carbon target was cut rather than the capital found; and the cumulative effect of buybacks has moved a founding family from ~31% to 44.7% ownership on shareholders’ cash. Management has allocated capital opportunistically well in acquisitions and poorly in the aggregate deployment of retained earnings into its own asset base.

8. Changes and Headwinds — Last Two Years

8.1 The policy reset (2025–26) — the dominant change

US Section 232. Tariffs at 25% from 12 March 2025, raised to 50% from 4 June 2025, with the derivative-product loophole closed in April 2026. For ArcelorMittal this is a net headwind of ~$150M per quarter, unchanged as of the Q1 2026 call, because it ships slabs into Calvert from Brazil and Mexico. It also removed US quotas in March 2025, which the 20-F identifies as depressing Brazilian slab prices. Management supports the policy in principle (“for over 20 years we have been arguing that the global steel industry has been suffering from overcapacity”) while seeking North American single-market treatment so that steel melted and poured in Canada and Mexico moves tariff-free. USMCA renegotiation is a live 2026 variable; the CFO declined to speculate.

EU CBAM and TRQ. CBAM began charging imports for carbon in 2026; the new tariff-rate quota regime takes effect 1 July 2026. Management calls this a structural reset. European index prices are up “almost EUR 100” since CBAM, with the benefit “not yet in our results.” Preparations to capture volume: idled blast furnaces at Fos and Dąbrowa Górnicza readied for restart (a furnace in Poland restarted around late April 2026), Gijón EAF first heat in Q1 2026, Sestao expanded to 1.6Mt. Management flagged elevated pre-TRQ imports in Q2 as buyers rushed material in before quotas tightened, and expects inventories to “normalize relatively quickly.”

EU ETS phase 4.2 cut free allowances from January 2026 — the cost is being accrued now while the CBAM benefit lags. The 17 June 2026 joint call by ArcelorMittal, thyssenkrupp Steel and voestalpine for urgent ETS reform, with Lakshmi Mittal writing in the FT that the trajectory “risks destroying Europe’s industrial base,” is the clearest signal of how binding this constraint is.

8.2 Portfolio and operational changes

• Calvert consolidated (18 June 2025) — 50% acquired for $1; 5.3Mt hot strip mill, 3.6Mt pickling/cold rolling, 2.1Mt finishing; new 1.5Mt on-site EAF produced first heat 14 June 2025 and was running at 20–25% of capacity in Q1 2026, with management targeting completion of the ramp by end-2026. The Cleveland-Cliffs 1.5Mt/year slab agreement was terminated and expired 9 December 2025, replaced by a seven-year Nippon agreement averaging 750kt/year. A second Calvert EAF is under “serious consideration.”
• Bosnia exited (30 October 2025) — ArcelorMittal Zenica and Prijedor sold for nil consideration, triggering a $205M impairment and a $61M disposal loss. A steel plant and iron ore mine given away is a data point about asset quality in the European long tail.
• Brazil bolt-ons — Tuper (May 2025), Tekno (November 2025, being delisted), Atlas solar (December 2025). Brazil also absorbed a $400M final settlement of the Votorantim purchase price.
• Vallourec partially exited (19 May 2026) — ~$667M raised, proceeds to buybacks.
• Mexico operational disruption, resolved — the illegal blockade of Lázaro Cárdenas (May–July 2024) cost ~800kt of production; a long-products blast furnace was down for preventive maintenance in H2 2025 and restarted late January 2026; there was an unplanned Lázaro Cárdenas DRI outage in Q3 2025. Management reported North American profitability “almost doubling” and full long-products capacity from Q2 2026.
• Gijón blast furnace B taken offline after a late-September 2025 incident; Dąbrowa blast furnace 3 idled from July 2025 on market conditions.

8.3 Ukraine — a persistent, under-appreciated drag

ArcelorMittal Kryvyi Rih remains consolidated and remains impaired in all but name. FY2025: 1.5Mt of steel shipments, $1.7bn of sales (of which $0.5bn domestic), 1.7Mt of crude steel and 7.6Mt of iron ore. Operating rates are 73% for open-pit mining and just 35% for steel. PP&E carrying value is unchanged at $0.7bn. Management states Ukraine was approximately EBITDA-neutral across FY2025 but free-cash-flow negative because of capex, and turned EBITDA-negative in Q1 2026 on very high energy costs (improving into Q2). The 20-F’s impairment testing applies an elevated country risk premium only through end-2026, reverting to a pre-war premium thereafter for the terminal value — an assumption that, if the conflict persists, would require revisiting and could produce a further impairment. Separately, Ukraine was not exempted from CBAM, which the CFO endorsed (“there shouldn’t be an exemption”) while noting the focus on domestic sales and pig-iron transfers within the group.

8.4 Leadership and other developments

• Geert Van Poelvoorde retires as CEO of ArcelorMittal Europe at the end of July 2026, moving to Chairman of the Board of ArcelorMittal Europe Steel (announced 6 July 2026). A leadership transition at the segment carrying the entire recovery thesis, at the exact moment the TRQ takes effect, is worth monitoring.
• 2030 CO₂ target cut from −30% to −10% (2025 Sustainability Report, 23 April 2026).
• $1.0bn of 5.375% notes due 2036 priced 12 May 2026.
• Dividend moved to quarterly; AGM/EGM on 5 May 2026 approved all resolutions with 82.28% of voting rights represented.
• AWS strategic collaboration (22 June 2026) on industrial automation and AI. Directionally sensible; immaterial to the financials and not a thesis input.
• Safety genuinely improving — LTIFR of 0.45x in Q1 2026 against 0.63x in Q1 2025, described as the lowest quarterly figure in group history. Financially material in a heavy-industrial context through downtime and liability avoidance.
• AMNS India resequenced — Andhra Pradesh greenfield (8Mt) now precedes further Hazira phases, with balance-sheet capacity the acknowledged constraint.

8.5 Verdict

These changes are net thesis-strengthening on revenue and net thesis-neutral on returns — and they concentrate rather than diversify the risk. The European policy reset is the most favourable development in a decade for ArcelorMittal’s largest segment, and Calvert for a dollar was genuinely well-played opportunism. But the improvement is regulatory, not competitive; Section 232 remains a $150M-a-quarter cost; ETS is a live and unresolved threat that management itself calls existential; Ukraine is an EBITDA-negative, capex-consuming asset carried at $0.7bn on an assumption that the war ends by 2027; Bosnia was given away for nothing; and the 2030 carbon target was cut rather than the returns found. The most important structural change is that the equity story has become more, not less, dependent on a single policy variable in a single region.

9. Risk Analysis

The risk profile in one sentence: the dominant risks are not operational or financial but regulatory and valuation — a business that does not earn its cost of capital, currently priced at the top of its own history on the strength of two European policy instruments (one three weeks old, one actively opposed by the company itself), with a sound balance sheet that makes permanent capital loss unlikely and a multiple that makes disappointing returns quite likely.

10. Valuation Discussion

No price target and no recommendation appear in this section. What follows is embedded-expectations and scenario analysis.

10.1 Where the stock trades

10.2 The percentile point, and why it inverts the usual argument

The most common bullish framing of ArcelorMittal is that it trades below book value and therefore embeds pessimism. On this company’s own ten-year history, that framing is inverted. At 0.93x book, ArcelorMittal is at the 97.8th percentile of its own decade — it has been cheaper on book 98% of the time. On sales, 0.83x is the 99.2nd percentile, effectively an all-time high. The composite sits at the 94.2nd percentile.

Why has MT historically traded at 0.3–0.6x book? Because that is what a business earning 2–7% on equity is worth. The discount to book was never an anomaly to be arbitraged; it was the market correctly capitalising sub-cost-of-capital returns. Paying 0.93x is not buying a dollar of assets for 93 cents — it is paying 0.93x for assets that generated a 3.2% underlying return on equity last year.

Note also that the P/E percentile (85.8th) is the least extreme of the three, and per standard practice that is exactly the metric to discount here: FY2025 GAAP EPS of $4.13 is distorted upward by the bargain-purchase gain and the 9.97% tax rate. Read the P/B and P/S percentiles, which are the reliable signals, and both are at or near record highs.

10.3 Embedded expectations: what must be true at 0.93x book

For an asset-heavy price-taker, the cleanest expectations framework is the relationship between price-to-book and sustainable return on equity: fair P/B ≈ ROE ÷ cost of equity.

The embedded expectation is therefore a sustainable ROE of roughly 9–10%. ArcelorMittal has achieved that only in 2017–18 and in the 2021–22 windfall. It requires the European segment — 47% of sales at a 1.8% reported margin — to move durably to something like 8%+ operating margins and stay there, for the long run, on the strength of CBAM and the TRQ.

That is not an absurd proposition. It is precisely what management argues, and the early data is consistent: EBITDA/tonne of $131 in Q1 2026 against $116 a year earlier, with the CFO stating this is “around 50% higher than our historical average margins” and that the CBAM benefit is not yet in the numbers. If Europe reaches 8% operating margins on $28.8bn, that is ~$1.4bn of incremental operating income, and combined with the claimed $1.8bn of strategic-project EBITDA the 9–10% ROE becomes reachable.

The point is not that the bull case is impossible. It is that at $66.88 you are paying for it in full, in advance, three weeks after the enabling policy took effect, with no discount for the possibility that it disappoints. The market is underwriting the successful outcome as the base case.

What the market appears to be pricing correctly: the 2026 cyclical and policy-driven recovery (Q2 consensus of $2,037M EBITDA is a genuine step-up); the value of the mining franchise and the Liberia ramp; the balance sheet’s safety; and the per-share benefit of a 38% share-count reduction.

What it appears to be pricing incorrectly: the durability of the European reset, treating a three-week-old trade instrument as a permanent structural change; the quality of FY2025 earnings, anchoring on $4.13 rather than ~$2.30; the cost of the decarbonisation programme, treating ~$200M-EBITDA EAF conversions as growth capex; and the composition of book value, treating an $8.9bn DTA contingent on the same recovery as though it were hard asset value.

10.4 Scenario analysis

Assumptions stated explicitly. All scenarios use through-cycle EV/EBITDA multiples below today’s 9.4x trailing, on the view that a 2%-ROIC price-taker should not sustain a premium multiple. Equity value = EV less ~$9.3bn net debt less ~$2.0bn minorities, on 761M shares.

Two features of this distribution matter. First, the current price of $66.88 sits above the base case (~$54) and roughly 82% of the way from base to bull. The market is not pricing the middle outcome; it is pricing most of the good one. Second, the spread is wide and roughly symmetric in percentage terms around a level below today — approximately −57% to +23% from the current price. For a business with a 96.3% lifetime maximum drawdown, that asymmetry is unattractive.

A useful sanity check from the cash-flow side: at a ~0.9% free cash flow yield on FY2025’s $471M, the equity is priced for free cash flow to multiply several times over. Management’s own “underlying free cash flow annualising at over $2 billion” — a figure that excludes seasonal working capital and all strategic growth capex — is a 3.9% yield on today’s market capitalisation, and it excludes precisely the $1.7–2.0bn of annual spending that the growth case depends on. On genuinely all-in reported free cash flow, the yield is under 1%.

10.5 Comparables

The cross-sectional argument that “MT is cheap versus the US mini-mills” is factually true and analytically empty. MT trades at a lower multiple because it earns roughly a fifth of Steel Dynamics’ trough return on capital, runs the more capital-intensive and carbon-exposed blast-furnace route, and is anchored in the highest-cost, most-regulated major steel market on earth. A lower multiple on structurally worse economics is not a discount; it is a correct relative price. The relevant comparison — supported by the factor-similarity evidence identifying Ternium as MT’s closest single-stock analogue, alongside thyssenkrupp and voestalpine — is to other integrateds, against which MT screens as the strongest operator in a weak cohort rather than as a bargain.

10.6 Verdict

Fairly-to-fully valued, with the improvement already capitalised and no margin of safety. At 0.93x book — the 97.8th percentile of its own decade — and 0.83x sales (99.2nd), ArcelorMittal embeds a sustainable ROE of roughly 9–10% against a ten-year median of 7.6% and an FY2025 underlying figure of ~3.2%. Scenario arithmetic centres on ~$54 with a range of roughly $29–82, placing the current price above base and most of the way to bull. The reflexive “below book therefore cheap” argument is inverted on this company’s own history. The bull case is coherent and could be right; the price no longer pays you to take it.

11. Variant Perception

11.1 What consensus believes

The consensus view, visible in the sell-side commentary and the Company’s own published Q2 consensus (13 brokers looking for $2,037M of EBITDA and $1.06 of EPS), runs roughly as follows: ArcelorMittal is a cheap, under-owned global cyclical at 0.93x book and under 1x sales, at an inflection point, where European trade protection (CBAM plus the 1 July TRQ) structurally resets the profitability of its largest segment. Earnings are recovering visibly ($116 to $131 per tonne), iron ore is ramping at record rates in Liberia, the balance sheet is investment-grade, buybacks have restarted, and 38% of the share count is already gone. Zacks has been publishing MT as simultaneously a “Strong Momentum Stock” and a “Top Value Stock.” Seeking Alpha frames it as “Steel Import Controls Are A Massive Tailwind.” The dissenting quantitative voice is GuruFocus, whose GF Value has flagged MT as “overvalued” repeatedly through 2026 at a fair value of ~$28.

11.2 The strongest bull case, stated at full strength

Europe is 47% of segment sales and 53% of tonnes, earning a 1.8% reported operating margin. That is not a normal margin — it is a margin created by a decade of subsidised imports taking share from domestic mills. CBAM now prices carbon into those imports and the TRQ now restricts their volume. If European utilisation normalises, ArcelorMittal — the largest producer in Europe, with idled blast furnaces at Fos and Dąbrowa ready to restart and new EAF capacity at Gijón and Sestao — captures that volume with enormous operating leverage on a largely fixed cost base. Every 100bp of European operating margin is ~$288M. Moving Europe from 3% to 8% adds ~$1.4bn, and management’s $1.8bn of strategic-project EBITDA sits on top. Add a mining business earning 24.4% margins and growing production 26.5%, an unconsolidated Indian JV with a 40Mt ambition in the world’s best steel market, and 38% fewer shares to divide it all by, and a normalised EBITDA of $10bn+ is achievable — on which today’s price is under 7x, against a stock that has averaged closer to 6x through cycles with far worse fundamentals. The balance sheet is investment-grade with $9.9bn of liquidity, so you are taking cyclical risk, not solvency risk. On this reading MT at 0.93x book is early, not late.

11.3 The strongest bear case, stated at full strength

ArcelorMittal has been the world’s largest steelmaker outside China for twenty years and has compounded shareholder capital at 1.66% a year with a 96.3% maximum drawdown over that period. It earned 1.98% on invested capital in 2025 while issuing ten-year debt at 5.375%, and has impaired roughly $18.4bn — 36% of its market capitalisation — over eleven years. Its reported FY2025 EPS of $4.13 was approximately $2.30 underlying, because 51% of operating income was a non-cash gain on a plant acquired for one dollar, taxed at nearly nothing. It converts $61bn of revenue into under $500M of free cash flow, spends 1.47x depreciation on capex, and covers interest with EBITDA-less-capex 0.12 times. Its flagship European decarbonisation programme earns ~$200M of EBITDA on ~4Mt, requires 50% state subsidy, and its assumptions are undisclosed — while the 2030 carbon target was cut from −30% to −10%. Its entire bull case rests on two European policy instruments, one of which is three weeks old and the other of which (ETS) the company itself publicly warns “risks destroying Europe’s industrial base.” And it trades at the 97.8th percentile of its own decade on book and the 99.2nd on sales after seven consecutive up quarters and a +195% move. Buy the improvement here and you are paying a record own-history price for a policy option on a business that has never durably earned its cost of capital.

11.4 Where I differ from consensus — the variant perception

My variant view is that MT is not a value stock and the tape proves it — and that this is the single most useful non-consensus fact available about the security.

Consensus discusses ArcelorMittal as deep value. The empirical factor evidence says otherwise, decisively. In the Base+Sector model (R² 0.459), MT’s loadings on the three style factors that would define such a thesis are approximately zero: Value −0.025, Momentum −0.015, Quality −0.002. In the fuller all-factor model (R² 0.626) Value is +0.072 and Quality −0.028. What actually explains MT’s returns is macro and credit: USDollar −0.80 (a levered short-dollar position), Sector Materials +0.84 and Industry Mining +0.78 (commodity beta), CreditRisk +0.65 to +0.77 (it trades like high yield), Liquidity −0.58 and InterestRate −0.39. Its nearest statistical neighbours are not operating peers but index funds — iShares Global Materials (0.844), Cambria Global Value (0.838), Global Metals & Mining (0.832) — with Ternium the closest single stock at 0.821.

Three consequences follow, and they are the heart of this article:

1. The +101% year was not a value re-rating or a momentum-factor bid. It was dollar weakness, credit-spread compression and mining/materials beta, layered on the steel-policy trade. Investors who believe they own a cheap asset re-rating toward intrinsic value are, factually, holding a short-dollar, long-credit-risk, long-commodity macro position. When those macro drivers reverse — a stronger dollar, wider credit spreads, higher real rates — the position unwinds regardless of what CBAM does to European steel spreads.
2. The “cheap on book” argument is inverted. Consensus reads 0.93x book as a margin of safety. It is the 97.8th percentile of MT’s own ten-year range. A 3.2%-underlying-ROE business trading at 0.93x book is expensive, not cheap, and the historical 0.3–0.6x range was the market being right rather than wrong.
3. The earnings the market is extrapolating are half accounting. Consensus anchors on FY2025 EPS of $4.13 and a 17.6x multiple. On underlying earnings of ~$2.30, the trailing multiple is ~29x. The improvement from here is real, but the starting point is materially lower than the reported base suggests.

Where I agree with consensus: the policy shift is genuine and material; the mining franchise is a real quality asset; the balance sheet is safe; and this is emphatically not a short — the momentum is violent, the policy tailwind is live, Q2 will likely be strong, and the realistic bad case is a de-rating rather than an impairment.

11.5 The 3–5 assumptions that actually matter

1. That the EU TRQ and CBAM durably reduce import penetration and are not diluted, circumvented or litigated away. Everything else is second-order.
2. That the European segment’s operating margin moves durably from ~1.8–3.1% toward 8%+ — the specific mechanism by which the policy change becomes a 9–10% group ROE.
3. That ETS free-allowance costs are reformed or offset, rather than eroding the CBAM benefit — the company itself is publicly lobbying on this, which tells you it is unresolved.
4. That the $1.7–2.0bn of annual strategic capex earns something like the claimed $1.8bn of EBITDA rather than the ~$200M-per-4Mt economics disclosed for the European EAFs.
5. That the macro factors actually driving the stock (dollar, credit spreads, mining beta) do not reverse while the fundamental story plays out.

11.6 Falsification tests

Falsifies the bull case: two or three consecutive quarters post-1 July 2026 in which European shipments and realised spreads fail to improve materially, or European operating margin stalls below ~5%, despite the TRQ being in force — demonstrating the policy does not convert to returns. Also falsifying: net debt/EBITDA breaching 1.5x and contractually halting the buyback; or a further Ukraine or European impairment indicating the asset base is worth less than carried.

Falsifies the bear case: European operating margin sustained above ~8% for two or three consecutive quarters with group ROIC moving decisively above ~8%, accompanied by all-in reported free cash flow (not “underlying,” not excluding strategic capex) exceeding ~$3bn annualised, and the buyback running above the 50% minimum. That combination would demonstrate a genuine structural step-up in returns rather than a cyclical bounce, and would justify a materially higher multiple on book.

12. Fact vs. Interpretation

13. Open Questions

1. What direction were the three 2026 insider transactions? Designated-person notifications were issued 13 May, 27 May and 3 June 2026, but the underlying records sit in the Luxembourg Stock Exchange OAM database and were not retrieved. As an FPI, MT files no Form 4s, so the usual insider read is unavailable. Absence of visible selling in a stock that has doubled is not evidence of absence.
2. What is the actual capital cost and expected IRR of the Dunkirk, Sestao and Gijón EAF conversions? Gijón is disclosed at €213M; Dunkirk and Sestao are not separately disclosed, and management explicitly declined to give return assumptions. Without the capital denominator, the ~$200M EBITDA figure cannot be converted into a return.
3. Will ArcelorMittal qualify for US tariff relief on Mexican and Canadian slabs? The CFO said the framework was newly published, that the answer “is not really a clear yes,” and deferred to the Q2 call. This is worth ~$150M a quarter and materially changes the North American economics and the second Calvert EAF decision.
4. How much of the Q2/H2 2026 European improvement is genuine demand versus pre-TRQ inventory distortion? Management conceded imports were “still high” in Q2 as buyers front-ran the quotas, and that inventories are “higher than normal levels.” The first clean read on TRQ economics may not arrive until Q4 2026 or Q1 2027.
5. What is the sustainable EBITDA of the Calvert complex now that it is consolidated? Revenue since acquisition was $2,599M with net income of just $7M. The bargain-purchase gain was derived from a DCF valuing Calvert well above $1 — yet the asset earned almost nothing in its first six months under full ownership. Which figure represents its earning power?
6. What are AMNS India’s balance sheet and funding needs, and how much parent capital will Andhra Pradesh consume? Management effectively conceded the sequencing was set to “balance the balance sheet.” India is the best growth asset and the least visible.
7. What happens to the $8.9bn DTA if the European recovery disappoints? The 20-F requires $38.9bn of future taxable income at certain subsidiaries. No sensitivity is disclosed.
8. Will the ETS be reformed? Management is publicly lobbying for it while telling investors European economics have structurally reset. The resolution of that tension is the single largest swing factor in the European margin case.
9. Is the Ukraine impairment assumption defensible? Value-in-use testing applies an elevated country risk premium only through 2026, reverting to pre-war levels for terminal value. What is the impairment exposure if the conflict continues past 2027?
10. Who succeeds Geert Van Poelvoorde as CEO of ArcelorMittal Europe, and does the transition at the end of July 2026 create execution risk exactly as the TRQ takes effect?
11. What is the detailed executive compensation structure and are incentives tied to ROIC? FPI status means no DEF 14A. Whether management is paid on tonnage, EBITDA or returns on capital could not be established to the usual standard, and it matters given a 1.98% ROIC alongside expansionary capex.

14. What Must Be True

14.1 For the bull case to work

1. The EU TRQ and CBAM must durably reduce import penetration and survive dilution, circumvention and legal challenge, holding for years rather than quarters.
2. The European segment must move from ~1.8–3.1% to 8%+ operating margins and hold there, converting policy into roughly $1.4bn of incremental operating income.
3. ETS free-allowance costs must be reformed or absorbed, so the carbon cost does not consume the CBAM benefit.
4. Group ROIC must move from 1.98% to sustainably above ~8–10%, delivering the 9–10% ROE that today’s 0.93x book embeds.
5. The $1.7–2.0bn of annual strategic capex must earn something close to the claimed $1.8bn of EBITDA, notwithstanding the ~$200M-on-4Mt European EAF economics.
6. All-in reported free cash flow must multiply from ~$470M toward $3bn+, sustaining the buyback above the 50% minimum without asset sales.

Falsification test for the bull case: By roughly Q1 2027 — after two to three full quarters with the TRQ in force — the Europe segment should show materially higher shipments and an operating margin above ~5% and heading toward 8%, with group ROIC visibly inflecting above ~5% and all-in FCF annualising above ~$2bn. If Europe is still posting sub-5% margins with the TRQ live and inventories normalised, the policy has not converted into returns and the bull case is broken regardless of the tape. A buyback suspension on the 1.5x net-debt/EBITDA brake would independently falsify it.

14.2 For the bear case to work

1. The policy benefit must prove partial, temporary or offset — TRQ leakage, ETS costs, or renewed import pressure once quotas are absorbed.
2. European margins must stall in the mid-single digits or below, leaving group ROIC in the low single digits.
3. Chinese overcapacity must continue setting the world price, keeping ~640M tonnes of excess capacity and record exports in play.
4. The multiple must de-rate from record own-history levels — 0.93x book and 0.83x sales reverting toward the 0.5–0.7x book that a sub-cost-of-capital business historically commands.
5. The macro factors actually driving the stock must reverse — a stronger dollar, wider credit spreads or higher real rates unwinding the −0.80 dollar and +0.65–0.77 credit-risk exposures.

Falsification test for the bear case: Two to three consecutive quarters of Europe operating margin sustained above ~8%, group ROIC decisively above ~8%, and all-in reported free cash flow (not “underlying,” not excluding strategic capex) above ~$3bn annualised, with the buyback running above the policy minimum and funded from operations rather than disposals. That combination would evidence a genuine structural step-up in returns rather than a cyclical-plus-policy bounce, would legitimise a P/B above 1.0x, and would make the “does not earn its cost of capital” objection obsolete.

15. Source Appendix

See Appendix B — Source Appendix below for the full itemised source list with URLs and access dates. Primary sources relied upon:

• ArcelorMittal 20-F for FY2025, filed 6 March 2026 (SEC CIK 0001243429) — the principal source for all FY2025 financials, segment data, note 2.2.4 (acquisitions, including the Calvert bargain purchase), note 2.3 (Bosnia divestment), note 2.6 (associates), note 3.1 (segment reporting), the auditor’s Critical Audit Matter, capital return policy, major shareholders, and risk factors including the deferred tax asset.
• ArcelorMittal Q1 2026 results press release, 30 April 2026.
• ArcelorMittal Q1 2026 earnings call transcript, 1 May 2026 (via ROIC.ai MCP).
• Form 6-K Exhibit 99.1 — Q2 2026 sell-side analyst consensus, filed 24 July 2026.
• Company press releases — Vallourec sell-down (19 May 2026), bond pricing (13 May 2026), joint ETS statement with thyssenkrupp and voestalpine (17 June 2026), Europe CEO retirement (6 July 2026), 2025 Sustainability Report (23 April 2026), AGM/EGM results (5 May 2026), designated-person notifications (13/27 May, 3 June 2026).
• AZI price history CSV and AZI valuation_index own-history percentiles, accessed 25 July 2026.
• FactorsToday factor loadings, leaderboard, related stocks and specific volatility, accessed 25 July 2026.
• ROIC.ai MCP — multi-year statements and ratios, used as cross-check; the filing governs where they differ, and section 6.1 documents three material discrepancies.
• Prior published analysis — NUE (20 June 2026), STLD (26 June 2026), CLF (11 July 2026) for peer comparison and industry figures (OECD excess capacity, Chinese exports, US import share).
• Third-party press — WSJ (30 April 2026, 19 May 2026), Zacks, GuruFocus, Seeking Alpha, for tape and sentiment context only; no analytical claim rests on them.

All sources cited in this article are public.

Sections 1–15 contain no investment recommendation and no price target. The Claude's Take block at the head of this article is the author’s own subjective opinion. This article is general information only and is not investment advice.

Appendix A — Diligence Questionnaire

ArcelorMittal S.A. (NYSE: MT) · Report date 25 July 2026 · Price referenced $66.88 (24 July 2026)

Supplemental to the research memo; not counted toward the memo length standard. Answers are grounded in the research notes and labelled Fact / Interpretation / Assumption where it matters.

General

What thoughtful questions have other investors asked about this company?

The Q1 2026 call (1 May 2026) is a good record of where sophisticated holders are actually focused, and three questions stood out as genuinely probing:

• Tristan Gresser (Jefferies) on EAF economics — the sharpest question of the call: “I was a bit surprised to see that you were only targeting $200 million of EBITDA for your 3 EAF projects… that’s close to 4 million tonnes of EAF steel. And it does not look like there are much of productivity gains or green steel premiums baked into that.” The CFO answered with framing (incremental basis, 50% French white-certificate support, an avoided blast-furnace reline) and explicitly declined to give assumptions, calling green-steel pricing “commercially sensitive” while conceding the premium market “will be limited.” [Interpretation] This is the most revealing exchange in the transcript and the strongest evidence that the European decarbonisation programme is a compliance cost rather than growth capital.
• Tom Zhang (Barclays) on the missing buyback — “I know your capital allocation policy hasn’t changed. We haven’t seen any buybacks for nearly a year now.” The CFO replied that the Company remained “very optimistic” about being FCF-positive, saw “no reason why we would not go above the minimum 50%,” and was “close to restart.” [Fact] The second tranche duly commenced around 1 July 2026.
• Ephrem Ravi (Citi) on Indian sequencing — asked whether Hazira phase 2 was being delayed in favour of the Andhra greenfield “in order to balance the balance sheet,” which the CFO effectively conceded: “I think you’re right. I mean, of course, we have to phase it.”

Other recurring investor preoccupations: the Q1→Q2 profit bridge and European carbon costs (Alan/BofA); European inventory levels and pre-TRQ import front-running (Bastian Synagowitz, Deutsche Bank; Cole Hathorn, Jefferies); whether Section 232 relief could apply retroactively to the first Calvert EAF (Gresser, Andrew Jones/UBS); US automotive contract resets and aluminium-to-steel substitution (Timna Tanners, Wells Fargo); the realistic timeline for the abandoned −30% carbon target and German stimulus (Maxime Kogge, Oddo); Ukraine’s EBITDA drag and CBAM exposure (Jones; Zhang); and Q2 net-debt modelling (Dominic O’Kane, JPMorgan).

[Interpretation] What is striking is what is not asked: nobody on the call questioned the $1,858M Calvert bargain-purchase gain, the resulting quality of FY2025 EPS, the 1.98% ROIC, or the valuation percentile. The sell-side conversation is entirely about the next two quarters’ bridge.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low?

[Interpretation] Neither — they are recovering off a trough, and 2026 is the first up-leg. The evidence: FY2025 EBITDA of $6,541M (company definition) against $19,335M in 2021 and $12,801M in 2022, with FY2023–25 clustered in the $4.8–6.5bn band. Quarterly EBITDA per tonne bottomed at $111 in Q3 2025 and has risen through $123 (Q4 25) to $131 (Q1 26). Q2 2026 consensus of $2,037M would be the best quarter since 2Q25. Reported FY2025 EPS of $4.13 is misleading as a cycle marker because roughly $1.83 of it was the Calvert gain and the associated tax effect.

Driven by the external environment or internal actions?

[Interpretation] Overwhelmingly external, and specifically regulatory. The three forces moving earnings are US Section 232 (a ~$150M/quarter cost to MT), EU CBAM (a benefit, live 2026) and the EU TRQ (a benefit, live 1 July 2026). Internal actions are real but second-order: the Liberia iron-ore ramp (genuine and management-driven), Calvert consolidation and its EAF ramp, Mexican operational recovery, and cost/asset optimisation. Management’s own framing on the call — “more favourable policy conditions translate into a stronger operating environment” — concedes the primacy of policy.

How stable are revenues?

[Fact] Highly unstable. Revenue over eleven years: $63.6bn (2015), $79.8bn (2022), $61.4bn (2025) — FY2025 was 3.5% below FY2015 in nominal terms. EPS has printed −$10.55 (2015) and +$13.53 (2021) within the same decade. Volume is the stable element: shipments have sat in a narrow 54–55Mt band, with FY2025 at 54.0Mt (−0.6%). [Interpretation] Volume stability with violent revenue and earnings swings is the signature of a pure price-taker: the tonnes move, the spread does everything.

Outlook for products/services?

[Fact] Management expects production and shipments to improve across all regions in 2026, European shipments to rise as domestic mills recapture import share, and — unusually — H2 to be stronger than H1. Q2 2026 was guided to improve in all steel segments. Iron ore is targeted at 80Mt+ of shipments at full capacity in H2.

How big will this market be — growing, shrinking, domestic or international?

[Fact] Global crude steel demand is mature and roughly flat in the developed world; the growth is in India (AMNS India’s 40Mt vision) and, historically, China. ArcelorMittal sells into ~126 countries, and all its sales are technically export sales since customers sit outside Luxembourg. [Interpretation] The addressable market is not growing meaningfully in ArcelorMittal’s core geographies — Europe is 47% of sales in a shrinking-to-flat market. What is changing is not market size but who is allowed to serve it, which is the trade-policy point. Growth capital is correctly being directed to India and to iron ore rather than to European steel volume.

Business Quality & Competitive Moat

Is the industry getting more or less competitive?

[Interpretation] Less competitive within protected regions, and no less competitive globally. Trade barriers (Section 232, CBAM, TRQ) are carving the world into regional markets where domestic incumbents face fewer import competitors — genuinely favourable for ArcelorMittal in Europe. But global overcapacity of ~640M tonnes and record Chinese exports of ~131M tonnes (2025) are undiminished, and capacity is still being added: ~9–12M tons of new US sheet capacity, MT’s own 3.4Mt of new EAF capacity by end-2026, and an 8Mt Indian greenfield. [Interpretation] Competitive intensity is being suppressed by regulation, not resolved by consolidation — a fragile form of relief.

How profitable is the business (ROIC, ROE)?

[Fact] Poorly. ROIC 1.98% (2025), 2.27% (2024), 2.54% (2023), 13.5% (2022), 25.5% (2021), −1.76% (2020). ROE 6.45% (2025) reported, ~3.2% on underlying earnings. Ten-year ROE mean 12.6%, median 7.6%. [Fact] The Company issued $1.0bn of ten-year notes at 5.375% in May 2026. [Interpretation] Investing at ~2% while funding at 5.375% destroys value at the margin, and ~$18.4bn of impairments over eleven years is the accumulated evidence.

How profitable is the industry — how many competitors, what barriers to entry?

[Interpretation] The industry is structurally unprofitable through the cycle, which is why it is chronically consolidating and chronically subsidised. Competitors are numerous and mostly national champions: in Europe thyssenkrupp, voestalpine, Tata Steel Europe, SSAB, Salzgitter; in the Americas Nucor, Steel Dynamics, Cleveland-Cliffs, Ternium, Gerdau, CSN; in Asia Nippon Steel, POSCO, JFE, Tata, JSW, and the Chinese majors. Barriers to entry are high in capital terms and low in practice, because the entrants that matter are state-backed or incumbent expansions rather than start-ups. [Interpretation] In the Greenwald framework the genuine barrier around ArcelorMittal’s European position is regulatory — trade protection and carbon border pricing — not structural, which means it is revocable by the authority that granted it.

Can the business be easily understood?

[Interpretation] The operations, yes — melting iron ore and scrap into steel is comprehensible. The accounts, emphatically no, and this is a genuine analytical hazard. FY2025 required unwinding a $1,858M bargain-purchase gain, $215M of fair-value remeasurement gains inside the associates line, a $400M purchase-price settlement, a $226M impairment, $194M of restructuring and disposal losses, a 9.97% effective tax rate, and a distinction between IFRS operating income ($3,628M) and the Company’s own EBITDA definition (which adds back D&A, impairment, special items and associates income). Third-party aggregators mis-state all of it — ROIC.ai reports FY2025 operating income of $1,461M and EBITDA of $4,406M against filed figures of $3,628M and $6,541M. Six reportable segments plus a materially profitable equity-accounted Indian JV outside consolidation completes the complexity.

Can it be undermined by foreign low-cost labour?

[Interpretation] Labour is not the vector — steel is capital- and energy-intensive, not labour-intensive. The genuine vector is foreign low-cost energy, foreign state subsidy and foreign carbon-cost arbitrage, which is precisely what CBAM exists to neutralise. ArcelorMittal’s European plants compete against imports whose cost advantage derives from cheaper power, weaker environmental compliance and state support. It is being undermined by foreign low-cost carbon, and the policy response is the entire investment thesis.

Do brands matter?

[Interpretation] Barely. The XCarb low-carbon range and Steligence construction platform are genuine attempts at branded differentiation, and 10 new XCarb RRP products were commercialised in 2025 with Environmental Product Declarations. But management itself conceded on the call that the green-steel premium market “will be limited,” and the three EAF projects are credited with only ~$200M of EBITDA precisely because no meaningful premium is assumed. Steel is bought on specification, price and delivery.

What is the nature of competition?

[Interpretation] Price and regional availability, mediated by trade policy. Within automotive there is a qualification-based, relationship-driven layer where technical capability matters. Everywhere else it is spread competition against the marginal importer.

Customers’ switching costs?

[Interpretation] Near zero in commodity flat and long products. Real but modest in automotive-qualified grades, where re-qualifying a supplier for a safety-critical part takes time and money — which is why ArcelorMittal is explicit that the Dunkirk EAF must “produce the same grades as we can today with the blast furnace” to protect a “very quality high order book.” [Fact] US automotive contracts reset on a rolling basis through the year (~30% Q1, ~30% Q2, ~25% Q3); European contracts concentrate at the start of the year. Annual resets cap the value of whatever captivity exists.

Financial Condition & Balance Sheet

Assets not fully recognised on the balance sheet?

[Interpretation] Probably yes, in two places, and both are hard to size. First, iron ore reserves are carried at historical cost less depletion, not at value — the Liberian and Canadian orebodies generating a 24.4% Mining margin are almost certainly worth more than their carrying value. The 20-F includes formal Reserves and Resources statements by experts. Second, land and long-held plant sites in Europe are carried at depreciated cost after decades of inflation; gross PP&E of $72.4bn against net $41.0bn indicates a heavily depreciated base. Set against that, the many impaired assets are carried at recoverable amount, so there is no hidden upside there.

Off-balance-sheet liabilities?

[Fact] Several disclosed items warrant attention:

• Decommissioning costs — the 20-F explicitly states that retiring assets over the next ten years in the transition to low-carbon steelmaking “may lead to certain decommissioning costs,” that these were considered in value-in-use calculations, but that no decommissioning provision has been recognised “as the obligating event has not occurred yet.” [Interpretation] This is a real future cash cost of the EAF transition that does not appear as a liability today.
• Pension and other post-retirement obligations — flagged in the risk factors as underfunded at some operating subsidiaries; sizeable but conventionally disclosed.
• Renewable power purchase agreements — accounted for variously as executory contracts under the own-use exemption or as IFRS 9 financial instruments; commitments disclosed in note 9.4.
• A put option granted to the AMTBA minority, exercisable 2030–2033, recognised as a $31M financial liability.
• Contingent and deferred consideration on the Tekno and Atlas acquisitions.
• Legal and arbitration proceedings — provisioned per note 9.1, contingencies described in note 9.3.

How conservative is the accounting?

[Interpretation] Mixed, and the answer differs by line. Conservative: the impairment record is aggressive rather than reluctant — ~$18.4bn taken over eleven years, including writing Bosnia down and disposing of it for nil consideration, and impairing Acciaierie d’Italia by $1,405M and Tameh by $81M. A company that writes down early is not hiding losses. Aggressive: the FY2025 income statement leans heavily on non-cash gains — the $1,736M bargain purchase (an estimate derived from a DCF, designated the auditor’s Critical Audit Matter precisely because “changes in the assumptions directly impacted the amount of the gain”) plus ~$215M of fair-value remeasurement gains on step-ups to control, all flowing through a 9.97% tax rate. Requiring judgement: the $8.9bn deferred tax asset, recoverable only against “at least $38.9 billion” of future taxable income concentrated in low-margin European subsidiaries; and the Ukraine value-in-use test, which applies an elevated country risk premium only through 2026 and reverts to a pre-war premium for terminal value.

How CapEx-hungry is the business?

[Fact] Very. FY2025 capex $4,337M on $61.4bn of revenue (7.1%) and against $2,945M of depreciation — a ratio of 1.47x. FY2026 guidance is $4.5–5.0bn, of which $1.7–2.0bn is “strategic.” [Fact] FY2025 CFO of $4,808M less capex left $471M of free cash flow; FY2024 left $447M. [Fact] EBITDA less capex covered interest expense 0.12x in 2025. [Interpretation] This is the defining financial characteristic of the business: it must reinvest nearly all of its operating cash flow simply to sustain and modestly grow an asset base that earns 2%. The EAF transition raises rather than lowers this burden, because it requires replacing functioning blast furnaces.

Capital Allocation & Management

How much FCF does the business generate, how does management use it, what is the philosophy?

[Fact] All-in reported free cash flow was $471M (2025) and $447M (2024), against $6,897M (2021) and $6,735M (2022). Management prefers “investable cash flow” (operating cash flow less maintenance/normative capex), reported at $2.0bn over the twelve months to Q1 2026, and separately cites “underlying free cash flow annualising at over $2 billion” excluding seasonal working capital and strategic growth capex.

[Fact] Trailing-twelve-month uses as disclosed at Q1 2026: $1.5bn strategic capex, $0.7bn shareholder returns, $0.2bn M&A — $2.4bn against $2.0bn of investable cash flow. [Interpretation] The Company is currently outspending even its own favourably-defined cash generation, funding the gap from the balance sheet and from the $667M Vallourec disposal.

[Fact] The stated philosophy is a progressive base dividend plus a minimum 50% of post-dividend free cash flow to buybacks, with an absolute brake: no buybacks if net debt/EBITDA exceeds 1.5x. Current ratio ~1.40x.

Significant acquisitions recently?

[Fact] Many, mostly small, and the 2025 vintage was unusually cheap. FY2025 net cash for acquisitions was only $169M: Calvert (remaining 50% for $1, with Nippon injecting $638M to repay Calvert debt and forgiving a $248M shareholder loan, plus a seven-year 750kt/year slab supply agreement); Tuper (Brazil pipes, control from 40%); Tekno (Brazil coil coating, $133M, being delisted); AMTBA (tailored blanks, to 90%); Atlas (265MW Brazilian solar JV, $47M). FY2024 was $1,352M, principally the Vallourec stake — ~10% of Vallourec was sold in May 2026 for $667M. [Interpretation] Recent M&A is a genuine credit to management: opportunistic, small, in adjacent value-added or captive-power positions, and the Monlevade expansion was cancelled rather than completed when returns no longer cleared.

Buying back shares?

[Fact] Yes, substantially. Shares outstanding fell from 1,080.7M (2020) to 761.1M (2025), −29.6%; the Company claims −38% fully diluted since September 2020. Spend: $5,170M (2021), $2,937M (2022), $1,208M (2023), $1,300M (2024), $262M (2025) — ~$10.9bn cumulative. The programme paused for roughly a year and the second tranche (up to 10M shares) of the 2025–2030 programme restarted around 1 July 2026, part-funded by the Vallourec sale. [Interpretation] Large and genuinely value-accretive per share, but pro-cyclical — heaviest at high prices and absent at the bottom, the opposite of Steel Dynamics’ counter-cyclical pattern. Capital deployed at 2025 prices would have retired roughly twice the shares per dollar as capital deployed in 2021.

Issuing large amounts of new shares to insiders?

[Fact] No. Share-based dilution is not a material feature here; the count is falling sharply. The 20-F discloses an Equity Incentive Plan, a Performance Share Unit Plan and a Special Grant for executives, and states that independent non-executive directors receive no share options, RSUs or PSUs, consistent with Luxembourg’s 10 Principles of Corporate Governance. Directors and senior management other than Lakshmi Mittal each beneficially own less than 1%.

Compensation policy of directors/management?

[Fact / gap] This could not be assessed to the usual standard, and the reason is structural: as a foreign private issuer ArcelorMittal files no DEF 14A proxy statement. Compensation is disclosed in the 20-F’s “Management and employees—Compensation” section at a granularity well below a US proxy. [Open question] Whether long-term incentives are tied to return on invested capital — as at Steel Dynamics, where the plan scores after-tax ROIC, operating margin and cash conversion against named peers — or to tonnage/EBITDA could not be established. [Interpretation] This matters a great deal for a company running 1.98% ROIC while committing $1.7–2.0bn a year of expansionary capital, and the absence of visibility is itself a finding.

Motivations of management?

[Fact] ArcelorMittal is family-controlled and family-run. The Mittal family holds 340,088,546 shares — 44.7% of the 761,125,819 outstanding (Lumen 275,840,595; Nuavam 63,658,348; Lakshmi Mittal 564,103 direct; Usha Mittal 25,500). Lakshmi Mittal is Executive Chairman; his son Aditya Mittal is CEO. The 20-F identifies the Chief Operating Decision Maker as “the Executive Office comprising the Executive Chairman, Mr. Lakshmi N. Mittal and the CEO, Mr. Aditya Mittal.” A 2006 Memorandum of Understanding governs the family’s relationship with the Company, and the risk factors acknowledge the Significant Shareholder “could also exercise significant influence over a change of control.”

[Interpretation] The alignment is genuine and has real advantages — a 44.7% owner has a long horizon, an aversion to dilution, and every incentive to compound per-share value, which plausibly explains the seriousness of the buyback programme. But two consequences deserve naming. First, because the family has not sold into the buyback, the ~30% reduction in shares has mechanically lifted its stake from roughly 31% to 44.7% — shareholders funded ~$10.9bn and one result is a founding family moving ~13 points closer to outright control without buying a share. (The 44.7% is Fact; the ~31% 2020 base assumes an unchanged absolute holding.) Second, at 44.7% the family holds effective veto over any change of control, so the usual disciplining threat of takeover is absent. [Interpretation] Continued buybacks walk this company toward family majority ownership, and minority holders should price that.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer?

[Fact] None of those, and the distinction matters for tax. MT trades on the NYSE as New York registry shares, not depositary receipts — although data vendors commonly and incorrectly label it “ArcelorMittal SA ADR” (both ROIC.ai and FactorsToday do so). It is a Luxembourg S.A., so distributions are subject to Luxembourg withholding tax (currently 15%, potentially reclaimable under treaty), which is the practical consideration for US holders. It is not an MLP and issues no K-1; US holders receive ordinary dividend treatment subject to withholding. It reports in U.S. dollars under IFRS, so there is no currency translation in the reported figures, though the underlying business carries material euro, Brazilian real, Canadian dollar, Indian rupee, South African rand, Mexican peso, Polish zloty, Argentine peso and Ukrainian hryvnia exposure. Note also that Argentina has been treated as highly inflationary since July 2018, with the Acindar accounts inflation-adjusted (a $67M charge to net financing costs in 2025).

Dividend policy?

[Fact] A progressive base annual dividend plus a minimum 50% of post-dividend free cash flow allocated to buybacks, with no buybacks permitted above 1.5x net debt/EBITDA. FY2025 dividend was $0.55/share ($421M) paid in two instalments (June and December 2025). The Company moved to quarterly payments with a first $0.15 interim in March 2026, against a proposed $0.60 annual — a yield of roughly 0.9% at $66.88. Management notes the dividend has doubled over five years. [Interpretation] The dividend is small and is not the reason to own this; the buyback is the primary return mechanism, and it is contingent on free cash flow and the leverage test.

How profitable is the business?

[Fact] Marginally, and less than reported. FY2025 reported operating margin 5.9% and net margin 5.1%; underlying operating margin ~4.2% and underlying EPS ~$2.30 against $4.13 reported. ROIC 1.98%. Segment margins diverge enormously: Mining 24.4%, Brazil ~9.0% underlying, North America ~2.8% underlying, Europe 1.8% reported, Sustainable Solutions 1.4%. [Interpretation] Group profitability is essentially Mining plus the equity-accounted Indian JV; the consolidated steel business is close to break-even on an operating basis in Europe.

Is net income diverging from cash from operations?

[Fact] Yes, and in the direction that warrants scrutiny. FY2025 net income (incl. minorities) was $3,243M against cash from operations of $4,808M — a ratio of 1.53x, which superficially looks healthy. But the reconciliation shows why: $2,375M of “other non-cash adjustments” were subtracted in arriving at operating cash flow (the Calvert gain and remeasurement gains being non-cash), alongside $2,945M of D&A and $903M of impairments added back. [Interpretation] The divergence is therefore benign in direction but instructive in cause — reported net income contained roughly $2.0bn of non-cash gains that the cash flow statement duly removes. This is precisely the corroboration for the section 6.1 conclusion: the cash flow statement confirms the earnings were not earned in cash. After capex, only $471M of free cash flow remained against $3,243M of reported net income — a conversion of 15%.

Risks & Downside

What factors would cause the stock to decline?

[Interpretation] In rough order of probability-weighted impact:

1. EU TRQ or CBAM diluted, delayed, circumvented or litigated — the entire European recovery case, live only since 1 July 2026.
2. ETS free-allowance costs persisting unreformed while CBAM benefits lag, eroding the European margin gain — a risk the Company itself is publicly lobbying against.
3. Multiple de-rating from the 97.8th percentile on book and 99.2nd on sales after seven consecutive up quarters and +195% from end-2024.
4. Macro reversal — the factor evidence shows −0.80 dollar and +0.65–0.77 credit-risk loadings, so a stronger dollar, wider credit spreads or higher real rates hurt the stock irrespective of steel fundamentals.
5. Renewed Chinese export pressure against ~640M tonnes of global excess capacity.
6. Cyclical disappointment — a Q2/Q3 miss versus the $2,037M consensus, or European volumes failing to materialise once pre-TRQ inventories are absorbed.
7. Buyback suspension on the 1.5x net-debt/EBITDA brake (currently ~1.40x).
8. Further impairment — Ukraine (a value-in-use test assuming a pre-war country risk premium after 2026) or the $8.9bn deferred tax asset.

Risk of a catastrophic loss?

[Interpretation] Low from the balance sheet, but the historical record demands humility. Against catastrophe: investment-grade rating, $9.9bn of liquidity, net debt/equity of 12.6%, net debt/EBITDA ~1.40x, maturities termed out, interest cover of 7.6x on EBITDA, and 34 plants diversified across 14 countries with no customer above 5% of sales. [Fact] But this security has a lifetime maximum drawdown of −96.3% and has compounded at 1.66% a year over roughly twenty years. [Interpretation] The 2008–09 and 2015–16 episodes show what a levered integrated steelmaker does in a genuine global downturn: a 96% drawdown was reached with far more leverage than today, so the realistic modern bad case is a severe de-rating and dilution-free drawdown rather than insolvency — but “severe” here could plausibly mean −50% or worse, as the 2022 episode (−37% in two quarters) demonstrated at much lower valuations.

Chance of a total loss?

[Interpretation] Very low on a 3–5 year horizon. The Company owns hard assets, produces 55.6Mt of steel and 48.8Mt of iron ore annually, carries modest leverage, and has a 44.7% family owner with every incentive to avoid dilution or distress. Total loss would require a multi-year global depression combined with an aggressive levered expansion — the opposite of current policy. The realistic downside is the ~$29/share bear scenario (roughly −57%), not zero.

Recent News & Events

Has the business environment changed recently?

[Fact] Yes, decisively, and it is the reason this article exists. Three changes:

1. EU CBAM began imposing a carbon cost on steel imports in 2026; European index prices are up “almost EUR 100” per management, with the benefit “not yet in our results.”
2. The EU tariff-rate quota took effect 1 July 2026, materially tightening import quotas — characterised by the WSJ as the EU “doubling” steel tariffs. ArcelorMittal has readied idled blast furnaces at Fos (France) and Dąbrowa Górnicza (Poland) for restart, commissioned the Gijón EAF and expanded Sestao to capture the volume.
3. US Section 232 at 50% since 4 June 2025, with the derivative loophole closed in April 2026 — a ~$150M/quarter cost to ArcelorMittal because it ships slabs into Calvert from Brazil and Mexico.

[Fact] Countervailing: ETS phase 4.2 cut free carbon allowances from January 2026, and on 17 June 2026 ArcelorMittal, thyssenkrupp Steel and voestalpine jointly warned that the ETS trajectory “risks destroying Europe’s industrial base,” with Lakshmi Mittal making the case in the Financial Times. [Interpretation] Management is simultaneously telling investors European policy has structurally reset earnings upward and telling Brussels European policy is existential. Both are defensible; together they locate this company’s economics in the regulatory arena.

Significant acquisitions?

[Fact] Covered above — Calvert (50% for $1, June 2025), Tuper, Tekno, AMTBA, Atlas in 2025; Vallourec partially exited May 2026 for $667M; Bosnia (Zenica and Prijedor) disposed for nil consideration in October 2025, triggering a $205M impairment and $61M disposal loss.

Change in accounting policies?

[Fact] No change in accounting policy was identified. But FY2025 saw unusually heavy application of business-combination accounting, which materially shaped the reported result: the $1,736M Calvert bargain-purchase gain plus $122M on settling pre-existing relationships (in operating income), and ~$215M of fair-value remeasurement gains on stepping up to control of AMTBA, Tuper, Calvert, Atlas and Perfilor (in the associates line). Several acquisition-date fair values remain provisional at 31 December 2025 (Calvert, Tekno, Atlas), meaning measurement-period adjustments are still possible. Deloitte designated the Calvert bargain purchase the Critical Audit Matter. A Luxembourg statutory rate reduction from 24.94% to 23.87% effective 1 January 2025 produced an $82M deferred tax charge, and $8M of Pillar Two top-up tax was recognised.

Recent changes — new markets, facilities, management?

[Fact]

• Facilities: Calvert consolidated with its new 1.5Mt EAF (first heat 14 June 2025, at 20–25% of capacity in Q1 2026, ramp targeted complete by end-2026); a second Calvert EAF under serious consideration; Gijón EAF first heat Q1 2026 (€213M); Sestao expanded toward 1.6Mt; Dunkirk EAF newly approved (replacing a blast furnace, ~2Mt, 50% supported by French white certificates); group EAF capacity heading to 30Mt (3.4Mt added by end-2026), with EAF now ~26% of production versus 19% in 2018. Idled/restarting: Fos and Dąbrowa blast furnaces readied (a Polish furnace restarted around late April 2026); Gijón blast furnace B offline after a late-September 2025 incident; Dąbrowa blast furnace 3 idled from July 2025.
• New markets: AMNS India resequenced — the 8Mt Andhra Pradesh greenfield now precedes further Hazira phases, within a 40Mt vision; ArcelorMittal Tubular Products Jeddah expansion with Sinosteel for OCTG/line-pipe heat treatment from December 2025; a Brazilian continuous annealing and galvanising line; a new electrical steels plant planned at Calvert.
• Management: Geert Van Poelvoorde retires as CEO of ArcelorMittal Europe at the end of July 2026, becoming Chairman of the Board of ArcelorMittal Europe Steel (announced 6 July 2026) — a transition at the segment carrying the entire recovery thesis, exactly as the TRQ takes effect.
• Other: $1.0bn of 5.375% notes due 2036 priced 12 May 2026; dividend moved to quarterly; AGM/EGM 5 May 2026 approved all resolutions with 82.28% of voting rights represented; a strategic AWS collaboration on industrial automation and AI (22 June 2026, immaterial to the financials); the 2030 CO₂ target cut from −30% to −10%; LTIFR improved to 0.45x from 0.63x, described as the best quarter in group history; and the 20th anniversary of the Mittal Steel/Arcelor merger falls on 31 July 2026.

[Open question] Three designated-person (insider) transaction notifications were issued in 2026 — 13 May, 27 May and 3 June. As a foreign private issuer ArcelorMittal files no Forms 3/4/5, and these notifications are lodged with the Luxembourg Stock Exchange OAM database rather than EDGAR. Their direction and size were not retrieved, so no inference about insider conviction is drawn in either direction.

Appendix B — Source Appendix

ArcelorMittal S.A. (NYSE: MT) · Report date 25 July 2026 · All sources accessed 25 July 2026 unless otherwise stated

Primary sources are listed first. Every non-obvious factual claim in this article and Appendix A traces to an entry below.

A. Company primary filings (SEC — CIK 0001243429)

ArcelorMittal is a foreign private issuer and files 20-F annual reports and 6-K current reports. It files no 10-K, no 10-Q, no DEF 14A and no Forms 3/4/5. A five-year EDGAR sweep (scripts/edgar.sh since MT 2021-07-01) returned 228 6-K, 5 20-F, 8 SD, 8 SC (13D/G), 7 424B5, 3 FWP, 1 F-3ASR and 1 25-NSE — and zero Section 16 filings.

The FY2025 20-F was mirrored locally to output/MT/sources/MT_20F_2025.htm and read in place.

B. Company earnings releases, transcripts and press releases

C. Quantitative data sources

D. Prior published sector analysis used for peer cross-read

Prior published analysis held locally in output/. Used for peer comparison and for industry-level figures (OECD excess capacity, Chinese export volumes, US import share, new US sheet capacity).

E. Third-party press and commentary

Used for tape, sentiment and event-dating context only. No analytical conclusion in this article rests on any item in this section.

Note on news retrieval: the ROIC.ai get_company_news feed for ticker “MT” returns unrelated items for MT Højgaard Holding A/S (a Danish construction group) and for a crypto token using the $MT symbol. These were identified and excluded.

F. Analytical frameworks

G. Data availability limitations

Stated plainly, because they bound several conclusions:

1. Proprietary-database access unavailable. No proprietary or subscription-only research is cited anywhere in this article. No sell-side steel industry primer was available; industry structure was built from primary sources and from prior published analysis.
2. No Form 4 corpus and no proxy statement. As a foreign private issuer, ArcelorMittal files no Forms 3/4/5 and no DEF 14A. The insider-transaction read that normally performed is unavailable; the three 2026 designated-person notifications point to the Luxembourg OAM database and their direction was not retrieved. No inference about insider conviction is drawn. Executive compensation structure — in particular whether long-term incentives are tied to return on invested capital — could not be established to the usual standard.
3. Q2 2026 not yet reported at the report date. The most recent 6-K (24 July 2026) publishes Q2 consensus, not results. All Q2 figures cited are sell-side consensus, clearly labelled.
4. Aggregator data unreliable for this issuer. Documented in section 6.1 and item 27 above. All operating income, EBITDA, segment and balance-sheet figures in this article are taken from the 20-F or the quarterly releases; enterprise value was built by hand.
5. EAF project capital costs incomplete. Only Gijón (€213M) is separately disclosed. Dunkirk and Sestao capital costs are not, and management explicitly declined to provide return assumptions, so the ~$200M incremental EBITDA figure cannot be converted into a return on capital.