Alcoa

Refresh delta (since 2026-06-18)

What changed

1. The stock lost roughly a third of its value. $80.86 on 2026-06-03 → 52-week high $84.38 in early June → $48.68 on 2026-07-02 (−39.8% in 30 days) → $51.43 on 2026-08-10. Against the prior dossier's ~$72 mark, that is −29%. Market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. is now $13.57B, trailing P/E 10.40, forward P/E 8.19.

2. Alcoa announced the largest acquisition in its history. On 2026-06-30 it signed an Umbrella Implementation Deed to buy South32's entire aluminium business ("AliGroup") for $3.1B cash + 17M Alcoa shares ($1.0B, ~6% of post-issuance shares) + up to $750M in contingent value rights — headline ~$4.85B. Assets: 86% of the Boddington bauxite mine and Worsley refinery (Australia), 100% of the Hillside smelter and idled Bayside (South Africa), 33% of MRN bauxite, 36% of the Alumar refinery and 40% of the Alumar smelter (Brazil). Funded by a committed $3.1B 364-day senior unsecured bridge. Expected close H1 2027. The stock fell 9.7% on the announcement.

3. The two segments completed their inversion — and alumina went below cash cost. Q2-26 Aluminum posted record Segment Adjusted EBITDA of $1,073M (32% of sales); Alumina posted −$96M, deepening from −$40M in Q1. Alumina's realized price of $334/t is now below its own adjusted operating cost of $368/t. Half the company is selling at a loss per tonne.

4. Guidance was cut on the alumina side. FY2026 alumina production 9.7–9.9Mt → 9.5–9.6Mt; shipments 11.8–12.0Mt → 11.5–11.6Mt — Pinjarra instability after Cyclone Narelle plus contamination. Aluminum production (2.4–2.6Mt) and shipments (2.6–2.8Mt) unchanged.

5. Section 232 did NOT roll back — it got more generous. No country exemptions were restored; a 2026-06-01 proclamation holds adjusted rates to 2027-12-31, and a 2026-07-20 proclamation created an onshoring incentive program halving the 50% primary-aluminum duty for companies with approved plans to build/refurbish/expand US smelting capacity (construction to start by 2029-01-20). The prior dossier's designated bear trigger never fired.

6. A gallium plant reached FID — a genuine critical-materials pivot. Alcoa, with the governments of Australia, Japan and the US (via JOGMEC + Sojitz), took final investment decision on a gallium plant at Wagerup; ~100 t/yr, up to ~10% of global gallium demand. Alcoa contributed $24M, equity-method.

7. Portfolio and balance-sheet housekeeping. San Ciprián smelter restart completed 2026-04-07; Alcoa bought out Trento EQT and goes to 100% of San Ciprián from 2026-08-01. Curtailed smelting capacity fell to 87 kmt from 341 kmt a year ago. Net debt fell to $873M (total debt $2,225M less cash $1,352M) from $1,089M at Q1. Equity rose to $7,370M from $6,118M at YE25.

What held

The structural thesis stands, and its central warning was right — but the mechanism was misdiagnosed. The prior dossier's core claim was "own the cycle, not the multiple; these earnings are borrowed." Earnings were indeed borrowed and the loan was indeed called inside eight weeks. But the prior dossier named Section 232 / the Midwest premium as "the current king variable" and the rollback as the bear trigger. That is not what happened:

• The tariff regime tightened, not loosened.
• What actually inflated and then deflated the price was a geopolitical supply shock: Middle East conflict disrupted roughly 9% of global aluminum production via the Strait of Hormuz, and prices hit a two-month low on 2026-06-15 when a US–Iran framework agreement raised the prospect of shipments resuming.
• That same conflict simultaneously raised Alcoa's own costs — the Q2 filing attributes Alumina segment weakness partly to "higher energy prices, primarily fuel oil and diesel, related to the Middle East conflict".

So the prior dossier was directionally correct and mechanically wrong. It also over-read Section 232 as a one-way tailwind. The Q2 MD&A lists "Tariffs on U.S. imports of aluminum from Canada" as an explicit drag on Aluminum Segment EBITDA in both the sequential and year-to-date bridges. Alcoa is long the Midwest premium and short the tariff on its Canadian tonnes. The net exposure is materially more balanced than "~60% of US volume rides an inflated premium" implied — and the 2026-07-20 onshoring program could cut the cost leg while leaving the revenue leg intact. This is a correction to the prior analysis, not a refinement of it.

What else held: the price-taker diagnosis (unchanged and now vividly demonstrated); the "never anchor on GAAP net income" discipline (Q2 proved it in reverse — see Lens 5); the balance-sheet repair (real, and it is precisely what bought management the capacity to do this deal); and the forensic all-clear (still zero SEC findings).

Phase A — Understand the business

Company Overview — REFRESHED (the company is being structurally recast)

Everything in the prior dossier's Lens 1 stands as a description of today's Alcoa: pure upstream, two segments, 25 locations, price-taker on LME + API. What changed is that management has committed to a different company by H1 2027.

Pro forma for AliGroup, on management's own numbers:

• +~5.2 Mt/yr alumina refining capacity — a ~53% pro forma increase.
• +~900 kt/yr primary aluminum — a ~37% pro forma increase.
• New country of operation: South Africa (Hillside), a jurisdiction Alcoa has not operated in.
• Consolidation of two existing JV positions: Alumar refinery (to ~100%) and Alumar smelter, plus a stake in MRN bauxite.

Oplinger's stated rationale, verbatim from the SEC-filed excerpt: "we are acquiring that capacity to a valuation that is well below replacement cost." The demand frame he offered: non-China primary aluminum demand +~7 Mt over the next decade, alumina demand +~18 Mt.

Also new and easy to miss: Alcoa now carries a secondary listing on the Australian Securities Exchange (ticker AAI, via CHESS Depositary Interests), and the Q2 10-QForm 10-QThe quarterly version of the annual report. Lighter, and not audited. adds a risk factor for dilutionDilutionIssuing new shares, so each existing share owns a smaller slice of the same company. and price impact from it.

Read: this is no longer a simplification story. From the 2016 spin through the Alumina Ltd buy-in, the Kwinana closure and the Ma'aden exit, the direction of travel was fewer, cleaner assets. AliGroup reverses that vector — it is a scale-up into the same two commodities, financed with debt, at the moment one of them is unprofitable.

Supply Chain — CARRIED

Unchanged, see the previous dossier Lens 2. The named chain (CBG/Halco bauxite, caustic soda, calcined coke via Strathcona, Hydro-Québec / NYPA / Landsvirkjun / AGL power, WA Huntly + Willowdale permitting) is intact and none of it moved this quarter. Note for the next refresh: AliGroup rewrites this lens on close — Boddington, Worsley, MRN and Hillside each bring their own input and power dependencies, and the Hillside Eskom power contract is a first-order new item (see Lens 13).

Competitive Advantages — REFRESHED (one new, one claimed, one questioned)

The prior dossier's four moats (vertical integration + scale; long-life hydro; ELYSIS; captive Brazilian hydro + WA resource life) all stand. Three changes:

New — gallium is a real, if small, moat of a different kind. The Wagerup gallium plant is not a commodity expansion; it is a critical-minerals supply-security asset, funded by three allied governments precisely because China dominates gallium supply and has used export controls on it. At ~100 t/yr it targets ~10% of global demand. Financially it is immaterial near-term (Alcoa's contribution: $24M, equity-method). Strategically it is the first Alcoa asset whose value derives from who is allowed to sell it rather than where it sits on a cost curve — and it converts a waste stream in existing refinery liquor into a policy-protected product. For a company whose primary_topic is critical-materials, this is the most thesis-relevant thing in the filing.

Claimed — cost-curve improvement via AliGroup. Oplinger: the deal will "improve our position on the global alumina and aluminum cost curves", with ~$900M NPV of synergies and ~$50M of run-rate cost savings in year one, which he pointedly framed as "not high-level consultant projections". Take seriously but unverified: $50M of year-one run-rate against ~$4.85B of consideration is ~1% — the $900M NPV is doing nearly all the work and is not decomposed in any primary document I read.

Questioned — the hydro/carbon advantage gets diluted. Adding Hillside (South Africa, Eskom grid, coal-heavy) and more Brazilian/Australian alumina refining moves the pro forma portfolio's carbon intensity in the wrong direction relative to the prior dossier's "carbon-advantaged hydro smelting into CBAM" story — which that dossier named as the durable thesis the market was mispricing. This deal partially trades that away for scale.

Segments — REFRESHED (the inversion completed)

All figures.

Alumina segment

The single most important number in this refresh: $334 realized against $368 of cost. Price crossed below cash cost in Q2. API averaged $307/t, −1% sequentially and −38% YoY. Costs rose for two separable reasons — Pinjarra's post-cyclone instability (a transient, and management guides Q3 costs favourable on its recovery) and Middle East-driven fuel oil/diesel (a macro). Six-month segment EBITDA swung −$939M YoY.

Aluminum segment

A $2,271/t spread between realized price and operating cost — against $3,376/t realized for all of FY2025. LME on a 15-day lag averaged $3,585/t; Midwest premium +10% sequential, Rotterdam +47% sequential (+179% and +102% respectively year-to-date). Unit costs fell YoY (−$271/t) on restart-driven volume leverage.

Total Segment Adjusted EBITDA: $977M (Q2-26) vs $654M (Q1-26); $1,631M (6M-26) vs $1,034M (6M-25), +58%.

The structural read, updated. The prior dossier called the segments "counter-cyclical to each other on alumina price" and noted total EBITDA was nearly flat despite a violent mix shift. That offset has now broken down. Alumina no longer merely underperforms — it consumes EBITDA, so it can no longer cushion an aluminum downturn; it would compound one. The company is, for the first time in this cycle, a single-commodity earner with an alumina liability attached — and it is buying more alumina.

---

Phase B — Measure performance

Earnings Result (latest print: Q2 2026) — REFRESHED

The print:

• Sales $3,966M (vs $3,018M Q2-25, +31% YoY; vs $3,193M Q1-26, +25% sequential)
• Net income attributable to Alcoa $407M (vs $164M Q2-25)
• GAAP diluted EPS $1.53 (vs $0.62 Q2-25)
• Total Segment Adjusted EBITDA $977M

vs consensus: MISSED on the line that matters. Adjusted EPS $2.12 vs $2.32 consensus — a $0.20 / 8.6% miss. Revenue $3.97B was within noise of estimates (Zacks consensus $3.91B → small beat; other compilers had $3.99B → small miss) — a live research-layer-vs-web conflict I am surfacing rather than resolving: the revenue actual is unambiguous at $3,966M from the filing; the estimate differs by compiler.

The GAAP/adjusted gap ran the OPPOSITE way to Q1 — and this is the quarter's most instructive detail. Q1's GAAP $1.60 was flattered by an +$88M Ma'aden mark-to-market gain. Q2's GAAP $1.53 was depressed by a −$123M Ma'aden mark-to-market loss (~$0.47/sh), as the Tadawul price fell from SAR 64.80 (3/31) to SAR 59.40 (6/30), carrying the stake from $1,485M to $1,360M. Total Other expenses, net was a $200M charge in Q2 vs $126M income in Q1 — a $326M sequential swing in a line with no operating content. The prior dossier's rule — work from Segment Adjusted EBITDA and operating cash flow, never headline EPS — is the reason this quarter reads as strong rather than flat.

Balance-sheet flags:

• Operating cash flow recovered hard: Q2 +$608M (6M $429M less Q1's −$179M). Q1's negative print was seasonal, as suspected.
• Receivables from customers jumped to $1,538M from $1,064M at YE25 (+$474M; the 6M cash-flow statement shows a −$440M working-capital drag). On +31% YoY sales this is largely mechanical, but it is the single largest working-capital movement and worth a second look next quarter.
• Inventories $2,340M (from $2,177M) — a milder build than Q1's.
• Net debt improved to $873M; total debt $2,225M (from $2,438M at YE25 — the remaining 2028 Notes were extinguished in May 2026). Interest expense fell to $36M from $56M YoY.
• Derivative liabilities fell to $1,374M ($494M current + $880M noncurrent) from $1,601M at YE25 — but see Lens 10, this is not the benign line it looks like.
• ARO improved to $1,025M from $1,120M; environmental remediation $209M.
• Equity rose to $7,370M from $6,118M, helped by +$454M of OCI (cash-flow hedges +$281M, FX +$148M).
• CapexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs. 6M $305M (Q2 $186M) against FY2026 guidance of ~$750M — tracking slightly light.

Credit ratings (pre-boundary but not previously captured): S&P upgraded to BB+ from BB (2026-03-03); Fitch affirmed BB+, outlook stable → positive (2026-03-11); Moody's Ba1 stable (2026-02-06). Oplinger states Moody's and S&P affirmed ratings and outlook on the pro forma transaction, and that cash consideration was sized to keep leverage ≤2.0× post-close.

Market reaction: negative. Shares fell ~4.2% after hours — the adjusted miss, the alumina guidance cut, and above all the deal. The tape is not disputing that Q2 was operationally the best aluminum quarter in Alcoa's history; it is disputing what management intends to do next with the balance sheet.

Unusual vs own history: two firsts in one print — a record Aluminum segment EBITDA and an Alumina segment selling below its own cash cost. Those coexisting is the whole story.

Earnings Calls (sentiment shift) — REFRESHED

transcripts/ was empty at the prior dossier and now holds one file: 2026-Q2.md, hand-written from the SEC Form 425 excerpt Alcoa itself filed on 2026-07-16.

The tonal arc, extended:

• 2024 — defensive; triage, curtailment, balance-sheet repair.
• FY2025 (Jan 2026) — operational confidence; production records; "met the high end of the adjusted net debt target."
• Q1 2026 — "resilience despite forecast miss"; leaning on a Q2 premium tailwind.
• Q2 2026 — a hard pivot from stewardship to ambition. The register changes completely. Oplinger opens the deal section with "the largest transaction for Alcoa Corporation" and "This acquisition is about creating long-term shareholder value."

The sentiment shift is the single most important qualitative delta, and it is a shift in kind, not degree. For three years the recurring vocabulary was defensive-operational: "operational stability," "disciplined capital allocation," "strong balance sheet / low debt," "reduce complexity," "optimize our asset portfolio." Q2's vocabulary is acquisitive: "strategic fit is compelling," "unlocks significant value," "accretive to our earnings per share and cash flow immediately after close," "well below replacement cost," "participate more fully in the long-term growth of the aluminum industry."

What they have started saying: replacement-cost framing, decade-scale demand forecasts (+7 Mt aluminum, +18 Mt alumina ex-China), and — notably — pre-emptive defence. A striking share of the prepared remarks answers objections nobody on the call had yet raised: "Let me provide some additional context on the transaction based on questions we have received from investors." He then walks through the cash/equity mix, the locked box, the ticking fee, the CVR, the South32 selling restriction (capped at 20% of ADV per day for three months) and the ≤2.0× leverage ceiling. Management knew the deal was badly received before the call began. The insistence that synergies are "not high-level consultant projections" is the tell — it is a rebuttal to a charge not yet made aloud.

What they have stopped saying: the prior dossier flagged management as "conspicuously quiet on what they'll do with the recovering free cash flow." That question is now answered — and the answer is not a buyback. It is a $4.85B acquisition. The dormant $500M repurchase program remains dormant; the dividend held at $0.10/quarter (ex-date 2026-08-11).

Net sentiment: confident, well-prepared, and defensive in a way the prior calls were not. Credible operationally; the promotional risk has risen from ~zero to non-trivial, because management is now selling a transaction rather than reporting a business.

Comps — REFRESHED (the de-rating is the story)

Provenance-critical lens. Multiples are with date or with arithmetic shown; anything unsourced is stated as such.

Honest note: I did not re-source the peer set this run — the prior dossier's peer figures are now ~8 weeks stale and I decline to recycle them as if current. The Alcoa column is fresh; the peer columns should be treated as not sourced rather than carried forward.

EV/EBITDA derivation: market cap $13.57B + net debt $0.873B − Ma'aden stake $1.360B ≈ EV ~$13.08B; ÷ FY2026E Total Segment Adjusted EBITDA ~$3.17B ≈ ~4.1×. On a deliberately conservative $2.5B: ~5.2×.

The read. At the prior dossier the stock screened at ~8.7× trailing EV/EBITDA and ~10× forward P/E, and the verdict was "the multiple is fine; the E is the problem — it's borrowed from a tariff." Eight weeks later the multiple has roughly halved to ~4.1× while FY2026 EBITDA is tracking above the prior base case. The market has repriced the E and the multiple simultaneously — which is what happens when a cyclical announces a large debt-funded acquisition at what investors suspect is a cycle peak.

South32 is now the most informative comp available and nobody has modelled it that way: Alcoa is paying ~$4.85B for assets South32 was willing to sell. The most useful valuation question is no longer "what multiple is AA worth" but "why was the seller a seller." That is not answered in any primary document I read, and I am flagging it as unresolved rather than guessing.

Stock-Price Catalysts — REFRESHED (three ±5% moves since the boundary)

• 2026-06-03 → early June: $80.86 → 52-week high $84.38. Peak of the war-premium rally. LME driven by Middle East disruption to ~9% of global aluminum production.
• 2026-06-10: sharply negative Q2 operating update — flagged a ~$60M unfavourable hit to Alumina segment EBITDA and cut alumina production guidance by 200–300 kt after Cyclone Narelle and Pinjarra contamination.
• 2026-06-15: aluminum hit a two-month low on the US–Iran framework agreement raising prospects of resumed Strait of Hormuz shipments. The war premium began unwinding.
• 2026-06-30 / 2026-07-02: AliGroup announced; stock −9.7% on financing and dilution concerns.
• Trough 2026-07-02: $48.68 — −39.8% in 30 days.
• 2026-07-16: Q2 print; −4.2% after hours on the adjusted-EPS miss + guidance cut.
• 2026-07-20: Section 232 onshoring incentive proclamation — halved primary-aluminum duty for approved US investment plans. Not visibly priced.
• 2026-08-07/09: LME recovered to ~$3,256–3,276/t, a six-week high, on supply constraints and non-China output −6.7% YoY in July; AA at $51.43 on 2026-08-10, +2.51% on the day.

What the market actually reacts to for AA — revised. The prior dossier ranked the drivers: (1) Section 232 / Midwest premium, (2) LME and API, (3) energy shocks, (4) portfolio events. The last eight weeks invert that ranking. Section 232 moved twice and the stock barely noticed; LME (driven by geopolitics) and the acquisition did all the damage. The corrected ranking: (1) LME, whatever is currently moving it — in 2026 that was a war, not a tariff; (2) balance-sheet/M&A events; (3) operational guidance cuts, which now bite harder because alumina has no margin cushion; (4) Section 232, which is largely priced and increasingly two-sided.

This remains a macro/policy stock wearing a company ticker — but the prior dossier picked the wrong macro.

---

Phase C — Judge people & books

Management — REFRESHED (the capital-allocation question got answered)

CEO William F. ("Bill") Oplinger — no change; ~2.9 years as CEO, ~13 years as a senior officer, finance-and-operations lifer. No management changes found in the filings or the web sweep.

What changed is the capital-allocation record, and it changed decisively. The prior dossier's critique was passivity: "management is hoarding optionality instead of returning capital," with the $500M buyback dormant and a $1.4B Ma'aden stake idle. That critique is now void — and replaced by its opposite.

1. Track record — still good, and it is what funded this. FY2023 −$651M → FY2024 $60M → FY2025 $1,157M → 6M-26 $832M. Debt down to $2,225M; net debt $873M; the 2027 and 2028 Notes retired; S&P upgrade to BB+; equity rebuilt to $7,370M. The deleveraging was real and it bought the optionality now being spent. Judge the deal knowing the balance sheet earned the right to attempt it.
2. The deal's structure is genuinely well-negotiated, and this deserves credit separately from whether the deal is wise. Four mechanisms shift risk to the seller:

• Locked box back to 2026-04-01 — AliGroup's cash generation accrues to Alcoa pre-close, estimated >$200M at 2026-06-30, offsetting consideration.
• CVR up to $750M over four years, paid only if alumina/aluminum prices exceed strikes — i.e. Alcoa pays more only in the world where it can afford to. Oplinger: "if markets are exceptionally strong, we will retain most of the value for our shareholders while a portion of that value will be shared with South32."
• ~$1B in stock — genuine risk-sharing; South32 holds equity exposure to the outcome.
• Orderly-sale restriction: South32 capped at 20% of ADV/day for three months post-completion, limiting the overhang. Against that, Alcoa pays a 5% annualized ticking fee on $3.1B from South32 shareholder approval (Oct/Nov) to close — management estimates $80–100M. That is a real, quantified cost of a long close.

1. Capital return remains last in the queue. Dividend held at $0.10/quarter ($26M paid in June; ex-date 2026-08-11). The $500M buyback is still dormant — now through a full 40% drawdown in the shares. Management has had the stock at $48–51 and chose to spend $3.1B of cash and issue equity at a $58.79 VWAP instead of buying its own. Issuing shares near a 12-month low to buy commodity assets is the single most challengeable allocation decision in this refresh, and it is the mirror image of the prior dossier's complaint.
2. Red flags: none of the governance kind — no related-party dealing, no restatement, controls effective, no strategy whiplash (the deal is consistent with the stated Investor Day thesis). The honest critique has simply inverted: from too passive to a large, levered, pro-cyclical bet with a ~12-month close window during which prices can move against every assumption.
3. Professional manager, confirmed — with a caveat. A finance-pedigree CEO negotiated an unusually good structure. Structure is not the same as timing, and timing is what a commodity acquisition is judged on.

Forensic Red Flags — REFRESHED

Re-run against the Q2 filing. Prior findings that stand unchanged: clean index-priced revenue recognition; conservative, transparent accounting; the trading gross-up in "Other segment items" (Alumina $343M, Aluminum $574M in Q2) remaining lower-quality than production margin; asbestos and CERCLA as routine heavy-industrial matters.

Improved since the prior dossier:

• Earnings-vs-cash quality reversed favourably. FY2025's problem was ~$1B of GAAP inflation from the Saudi gain and Ma'aden marks. In Q2-26 GAAP was understated: a $123M Ma'aden MtM loss sits in the $200M Other-expenses charge. Underlying cash generation was strong — Q2 OCF +$608M.
• Restructuring finally stopped being a recurring "one-time" cost. $184M (2023) → $341M (2024) → $918M (2025) → $14M for all of 6M-2026, with Q2 a net benefit of −$4M. The prior dossier's "when one-time charges recur every year, they're an operating cost" critique has genuinely resolved. The Kwinana cash outlays (~$525M through 2031) continue as previously disclosed.
• ARO fell to $1,025M from $1,120M — the first decline after the Kwinana-driven build.

New or newly-important flags:

• The power-derivative liability is a disclosed forward hit to Sales, not a mark against equity. The prior dossier read the ~$1.5B net derivative position as "a large mark against equity via OCI." The Q2 filing is more specific and more concerning: "Assuming market rates remain constant with the rates at June 30, 2026, a realized loss of $376 related to power contracts is expected to be recognized in Sales over the next 12 months". These are embedded derivatives in power contracts hedging 893 kmt of aluminum. This is the mechanism behind the recurring MD&A line "unfavorable impacts from certain energy contracts linked to metal pricing." It means Alcoa's realized upside is structurally capped: as LME rises, these contracts claw back a portion through revenue. The headline $4,752/t realized price is already net of some of this, and $376M more is queued. Any model built off Q2's spread without this haircut is wrong. Level 3 power-contract liability: $1,200M at 6/30/26.
• AliGroup accounting items to track: transaction costs of $11M (Q2) / $13M (6M) expensed into SG&A — expect these to grow materially through a H1-2027 close; the ticking fee ($80–100M); the CVR, which will require fair-value measurement and can introduce earnings volatility; and the $3.1B bridge, whose replacement with permanent financing is a stated intention, not a completed transaction. The bridge is the live financing risk: it is a 364-day facility.
• Locked-box exposure runs both ways. Alcoa takes AliGroup's economics from 2026-04-01 — including any deterioration. With alumina below cash cost, an alumina-heavy acquired portfolio may be accruing losses, not the ">$200M" of value, if prices weaken before close. The disclosure gives the upside estimate; it does not model the downside.
• Receivables +$474M since YE25 (−$440M cash-flow drag). Mechanical on +31% sales, but the largest working-capital line — monitor.

Regulatory findings (required, re-verified):

• SEC Litigation Releases / AAERs: None. Re-fetched 2026-08-10, period 2021-08-10 → 2026-08-10: total_sec_findings: 0.
• Controls: CEO/CFO concluded disclosure controls effective as of 2026-06-30; no changes in internal control over financial reporting during Q2.
• New regulatory exposure to watch: antitrust/merger clearance. AliGroup is conditioned on "receipt of required regulatory approvals" across multiple jurisdictions (Australia, Brazil, South Africa, and likely others). Alcoa pays an $82M break fee in specified circumstances; South32 pays $41M or $82M. This is a genuinely new regulatory risk class for Alcoa, which has not faced a merger review of this scale since the Alumina Ltd buy-in.
• Conclusion: no accounting-fraud or securities findings; accounting remains conservative and the quality of earnings improved this quarter. The live risks are environmental permitting (Western Australia, unchanged) and now merger clearance.

---

Phase D — Project & stress-test

Forward Projection (FY2026 / FY2027 / FY2028 EPS) — REFRESHED

Model integrity — stated before any number is cited. our model ran and wrote model.xlsx, but our model returns no computed values: "The workbook has 51 formulas and no cached results." It also reports only 2 quarters of history (revenue growth defaulted to a seeded 8%), an opening balance sheet "partly sourced" (8/10 lines — missing Other current assets and Other current liabilities), and a blank share price. I am therefore citing no figure from the workbook. Every projection below is `` with arithmetic shown, built from the filings. The workbook is a scaffold for the next refresh, not a source for this one.

Actuals in hand: 6M-26 sales $7,159M; Total Segment Adj EBITDA $1,631M; net income attributable $832M; GAAP diluted EPS $3.13; adjusted EPS ~$3.52 (Q1 ~$1.40 + Q2 $2.12). Shares 264M basic; 263.91M outstanding.

Macro marks: LME ~$3,256–3,276/t (2026-08-07/09) vs Q2's 15-day-lag average of $3,585/t — i.e. spot sits ~9% below the price that produced Q2's record. API ~$307/t. Non-China output −6.7% YoY in July.

Base case FY2026. H2 aluminum realized price falls with LME toward ~$4,350/t (from $4,752/t) on ~700 kt of produced shipments per quarter; a ~$400/t decline ≈ −$280M/quarter vs Q2 → Aluminum EBITDA ~$790M in each of Q3 and Q4. Alumina improves off −$96M on Pinjarra recovery and lower diesel, but stays negative at a $307 API against a $352 6M cost → ~−$30M/quarter. H2 Total Segment Adj EBITDA ≈ 2 × $760M ≈ $1,520M; FY2026 ≈ $1,631M + $1,520M ≈ $3,150M.

Bridge to EPS: $3,150M − Transformation ~$100M − Corporate ~$200M + intersegment ~$18M ≈ $2,868M; − D&A ~$680M ≈ $2,188M; − interest ~$145M ≈ $2,043M; − restructuring/other ~$30M ≈ $2,013M; − Other expenses net ~$74M (6M actual, H2 assumed nil) ≈ $1,939M; − tax at the 6M effective rate of 15.8% ($155M/$981M) ≈ $1,633M; ÷ 264M ≈ GAAP EPS ~$6.20; adjusted EPS ~$6.15–6.55.

Cross-check: $51.43 ÷ $6.28 ≈ 8.2×, against the independently published forward P/E of 8.19. The two agree closely, which is mild corroboration of the base case.

Note the inversion vs the prior dossier: its base case was FY2026 Segment Adj EBITDA $2.4–2.7B and EPS ~$6.30–6.90. Actual EBITDA is tracking ~$3.15B — above the prior range — while EPS lands at the low end, because the mix shifted (aluminum far better, alumina far worse) and D&A/other rose. The company beat the prior EBITDA case and the stock fell 29%. That gap is the refresh.

Bull FY2026: LME holds ≥$3,400, premiums hold, alumina claws to breakeven → EBITDA ~$3.5B, EPS ~$7.20.

Bear FY2026: LME reverts to ~$2,800 on a durable Hormuz reopening, premiums soften, alumina stays −$100M/quarter → EBITDA ~$2.4B, EPS ~$4.00.

FY2027 — the hard year to model, and I will not pretend otherwise. AliGroup is expected to close H1 2027, so 2027 is a stub year: partial contribution, full financing cost, and ~17M new shares (~281M diluted, +6.4% dilution).

• Standalone 2027 at a normalizing LME ($3,000) and softer premiums: **$5.10**.
• Plus a half-year of AliGroup: if the acquired assets generate $700–900M of annual EBITDA mid-cycle, a half-year ≈ $400M, less ~$100M incremental interest (on $3.1B at ~6.5%, half-year) and ~$100M D&A, taxed → ~+$150M ≈ +$0.53.
• Less dilution (~6%).
• FY2027 base ≈ $5.30. Error bars here are wide — the acquired assets' standalone earnings are not disclosed in any Alcoa primary document I read, so the $700–900M is my inference from capacity, not a sourced figure. Treat FY2027 as directional only.

FY2028 — first full year of the combined company: ~$5.50–6.00 at mid-cycle prices, and materially higher if alumina recovers, because the acquired portfolio is alumina-weighted (+53% alumina vs +37% aluminum). This is the crux: FY2028 is a leveraged bet on alumina mean-reverting. If API returns to ~$400/t, the acquired 5.2 Mt swings from marginal to strongly profitable and EPS could clear $8. If API stays ~$307/t, Alcoa will have bought 5.2 Mt of below-cash-cost capacity with $3.1B of debt.

EPS path: 2026 ~$6.20 · 2027 ~$5.30 · 2028 ~$5.75. I could not source current sell-side consensus for FY2027/FY2028 this run — n/a, not sourced (the prior dossier's $6.40/$6.30 figures predate both the deal and the price collapse and should not be carried forward). The one current market-wide datapoint: 16 analysts, 12-month target $62.98, ~22% above spot.

Per unattended rules, no Brier forecast logged (our model create skipped).

Bull vs Bear — REFRESHED

Bull case. You are being handed a business printing record aluminum profitability — $1,073M segment EBITDA at 32% margins, unit costs falling on restart volume, curtailment down to 87 kmt from 341 kmt — at ~4.1× EV/EBITDA and ~8.2× forward earnings, roughly half the multiple of eight weeks ago, with net debt of just $873M, an S&P upgrade in hand, and the alumina segment at what is plausibly a cyclical floor (below cash cost is not a durable state; capacity curtails). On top sit four options the price does not obviously reflect: AliGroup at a claimed discount to replacement cost with ~$900M NPV synergies and a seller-friendly-to-Alcoa risk structure (locked box, CVR, equity component); the Section 232 onshoring program (2026-07-20), which could halve Alcoa's Canadian-import tariff cost while leaving the Midwest premium intact — a lever nobody has priced; gallium at Wagerup, a government-backed critical-minerals asset; and ELYSIS (first production 2027). LME has already recovered to a six-week high with non-China output down 6.7% YoY.

Bear case (3 ways it permanently impairs or de-rates).

1. The deal is a pro-cyclical, debt-funded bet on the commodity that is currently unprofitable. Alumina sells at $334/t against $368/t of cost, and Alcoa is adding 5.2 Mt (+53%) of it with $3.1B of bridge debt on a ~12-month close. Every mechanism management touts as risk-sharing (locked box, CVR) is calibrated to strong prices; the locked box transfers AliGroup's economics to Alcoa from 2026-04-01, which in a weak alumina tape means absorbing losses, not the ">$200M" cited. Add $80–100M of ticking fees for the privilege of waiting.
2. Realized upside is structurally capped by the power contracts. A $376M realized loss on power contracts is already scheduled to hit Sales over the next 12 months at 6/30/26 rates, against 893 kmt of hedged aluminum. Alcoa does not fully monetize an LME rally — the higher metal goes, the more these embedded derivatives claw back. Bulls extrapolating the $4,752/t realized spread are extrapolating a number with a known, disclosed, $376M drag behind it.
3. Still a price-taker, now with more operating and financial leverage, and a diluted carbon story. Two macro commodities set by China and the LME, USD revenue against AUD/BRL/EUR/CAD/NOK/ZAR costs, and a pro forma portfolio that adds coal-grid South African smelting — trading away part of the hydro/CBAM advantage the prior dossier identified as the durable thesis.

Pre-mortem (18 months out, thesis broke — what happened?). The Hormuz reopening held; LME settled back to $2,700–2,800; the Midwest premium normalized as the onshoring program pulled new US supply forward; alumina stayed in surplus through 2027 as Indonesian and Indian refineries ramped. AliGroup closed in H1-2027 into that tape — Alcoa issued 17M shares near the lows and drew $3.1B, arriving at ~$4B of net debt against a combined portfolio in which the alumina half loses money. Leverage breached the 2.0× ceiling, the ratings outlook reversed, the dividend came under scrutiny, synergies slipped, and the bridge was refinanced at a worse coupon. No fraud, no blow-up — management simply bought the cycle instead of selling it, and did so with a 12-month settlement lag during which the world moved.

Are multiples too high? No — and this is a genuine reversal of the prior dossier's answer. At ~4.1× EV/EBITDA and ~8.2× forward P/E on earnings that are above the prior base case, Alcoa is not expensive on any conventional reading. The question has changed from "is the multiple too high?" to "is the E about to be spent?" The market is not saying AA's earnings are fake; it is saying management is about to convert them into an illiquid, levered, alumina-weighted bet. That is a governance-and-timing discount, not a valuation-and-cycle discount, and it is a different thing to underwrite.

Contrarian view (what the market is refusing to see). Two things, pulling opposite ways.

In the bulls' favour: the 2026-07-20 Section 232 onshoring program is a genuinely mispriced, asymmetric lever. The market still models Section 232 as a one-way tariff windfall at risk of repeal — the framing the prior dossier also used. The filings show it is two-sided: the tariff is an explicit cost to Alcoa on Canadian imports (named in both Q2 EBITDA bridges) as well as the source of the Midwest premium. A program that halves the duty for companies investing in US smelting — and Alcoa has 54 kmt curtailed at Warrick plus Massena West — could cut the cost leg while leaving the revenue leg intact. Nobody is modelling that.

Against them: the market is also refusing to ask the one question that matters most, and I cannot answer it from primary sources — why was South32 a willing seller of an entire aluminium division at a price Alcoa calls "well below replacement cost"? Replacement cost is the wrong yardstick for assets whose product sells below cash cost; nobody replaces capacity into a loss. Two sophisticated operators looked at the same assets and reached opposite conclusions. That disagreement, not the synergy number, is the real due-diligence question, and it is unresolved in every document I read.

Devil's Advocate (short-seller) — REFRESHED

Dismantling the bull case.

• What structurally breaks the money machine — restated, because the prior answer was wrong. The prior dossier named Section 232 repeal. Eight weeks later the tariff had tightened and the stock had fallen 39% anyway. The real fragility was never the tariff — it was that ~$1,400/t of the realized aluminum price is a spread that exists only while somebody else's supply is disrupted. Q2's $4,752/t was built on a war. A US–Iran framework agreement, signed by people who have never heard of Alcoa, took a third of the equity out in three weeks. The short case is not "the subsidy ends" — it is "the disruption ends." That is faster, less negotiable, and impossible to hedge.
• The tell everyone missed: management sold equity at $58.79. Alcoa agreed to issue ~17M shares valued on a 10-day VWAP to 2026-06-26 of $58.79 — after the stock had already fallen from $84.38 and while the buyback sat dormant. A management team that believed its shares were cheap would not have used them as currency at that level. The share consideration is a revealed preference about valuation, and it points down.
• Why the moat is weaker than bulls think. "Below replacement cost" is doing enormous rhetorical work. Replacement cost is a valuation floor only for assets someone would rationally replace — and at $334/t realized against $368/t of cost, nobody would build an alumina refinery today. Buying below replacement cost is not a bargain when the asset earns below cash cost.
• The new single-asset landmine bulls have not read: Hillside's power contract. On the call, an analyst noted that South African power rates for other customers average six times what the Hillside smelter pays, with the contract up for renewal in ~five years. Oplinger did not dispute the 6× figure — he answered with process ("South32 has already begun discussions with Eskom") and declined to speculate: "five years out on a transaction that we haven't closed yet is difficult to speculate". A smelter's economics are its power contract. Alcoa is acquiring a 100%-owned smelter carrying an off-market power deal on a visible cliff, in a jurisdiction it has never operated in, with a counterparty (Eskom) of well-known financial and operational fragility. This is the least-analyzed liability in the transaction.
• The realized price is not what it appears. $376M of realized power-contract losses are scheduled into Sales over the next 12 months at 6/30/26 rates. Bulls annualizing Q2's spread are annualizing a number with a disclosed drag queued behind it.
• Most dangerous competitor bulls underestimate: unchanged — Chinese state-linked smelters at the 45.5 Mt cap setting the marginal price, plus Indonesian and Indian refinery additions that are already the reason alumina is below cash cost. Alcoa's response to that surplus is to buy 5.2 Mt more of it.
• Worst capital-allocation choice — and it is now concrete, not a passivity critique. With the stock at $48–51, a dormant $500M buyback, net debt of only $873M, and a $1.36B liquid-ish Ma'aden stake, management chose to commit $3.1B of cash and ~$1B of equity to commodity assets. Buying someone else's alumina instead of your own equity at a 12-month low is the decision this refresh should be judged on in two years.
• Assumptions that must hold for today's ~$51: LME ≥$3,000; alumina troughs and recovers rather than grinding sideways at $307; AliGroup closes on terms and gets multi-jurisdiction clearance; the bridge is refinanced at a tolerable coupon; leverage stays ≤2.0×; Pinjarra stability holds; WA mining exemption extends past Aug 2027; Hillside power is renegotiated survivably. That is eight conditions, up from the prior dossier's five — the deal added three.
• What if results disappoint 20–30%? A 25% haircut to FY2026E EBITDA takes ~$3.15B to ~$2.36B and EPS to ~$4.00; at an unchanged 8.2× that is a ~$33 stock, −36%. Note this is a smaller proportional downside than the prior dossier's −40% at a higher price — the de-rating has already absorbed part of it.
• Single scenario that permanently impairs: the Hormuz normalization holds and alumina stays in surplus through 2027 and AliGroup closes into it. Alcoa emerges with ~$4B of net debt, 281M shares, and a portfolio in which the enlarged half loses money — re-rating to trough multiples on trough earnings with a balance sheet that no longer has the flexibility that made 2025 survivable. Plausibility: moderate. Lower than the prior dossier's "moderate-to-high," because the price has already fallen 29% and the tariff risk it feared proved illusory — but the consequences are now worse, because leverage is being added rather than removed.

Management Questions — CARRIED

Unchanged, see the previous dossier Lens 14 for the 15 questions. Q1 (Section 232 sensitivity) and Q2 (why no buyback) have been partially answered by events — see Lenses 9 and 12. Six questions the transaction makes urgent for the next call, recorded here for the next refresh: (a) what are AliGroup's standalone EBITDA and cash cost per tonne at spot API; (b) what happens to the locked box if alumina stays below cash cost to close; (c) what is the assumed permanent-financing coupon replacing the bridge, and what leverage results if LME is $2,800 at close; (d) what is the Hillside Eskom renewal assumption and the EBITDA sensitivity to a 2×, 4× or 6× power-price reset; (e) has Alcoa filed an onshoring plan under the 2026-07-20 Section 232 program, and what is the annual tariff-cost saving if approved; (f) at what share price does the buyback beat M&A on your own hurdle rate.

---