Alcoa

Refresh delta (since 2026-08-10)

What changed

1. The stock fell another ~15–19% and is now at the bottom of its 12-month band. $51.63 close (2026-08-10; the August dossier's $51.43 was an intraday mark) → $51.75 (09-08) → $44.43 (09-18) → $41.95 close 2026-10-02, $43.84 on the latest quote (2026-10-05, intraday — not a close); market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. $11.57B, trailing P/E 8.69, forward P/E 7.45, 52-week range $33.77–84.38, Street target $62.64. The driver was LME, not company news: aluminium touched ~$3,360/t on 2026-09-10 and slid to $3,213.5 (3M) on 09-29 and a $3,203 cash bid on 09-30 — against the $3,585/t 15-day-lag average that produced Q2.

2. The AliGroup black box opened — and the number lands at the top of the prior dossier's guess. The updated pro forma (as if closed 2025-01-01) shows LTM-to-June-2026 Adjusted EBITDA ex-special items of $3,154M combined vs $2,293M Alcoa standalone → ~$861M attributable to AliGroup; FY2025 $2,811M vs $1,965M → ~$846M; AliGroup LTM net income $307M (IFRS, standalone). The prior dossier's FY2027/28 arithmetic rested on an inferred $700–900M; it is now sourced. Pro forma net income attributable: $1,482M LTM (vs $1,277M standalone), $1,022M 6M-26 (vs $832M).

3. Permanent financing is done; the bridge is gone; the interest clock started early. $1.5B 6.625% notes due 2034 (issuer Alumina Pty Ltd) and $1.1B 6.875% notes due 2036 (Alcoa Nederland Holding) closed 2026-09-23; the $3.1B 364-day bridge was terminated the same day; cash portion funded $2.6B notes + cash on hand. A special mandatory redemption at 100% applies if the deal terminates or will not close by the end date. Annual coupon ≈ $175M, accruing from 23 September 2026 (first payment date not on the shelf) although the assets do not arrive until H1 2027. Pro forma total debt $5,420M; pro forma cash $668M at 6/30/26.

4. The seller's reasons are now on the record — the prior dossier's open question is answered. South32 is selling to become "an upstream base-metals focused" company (≥80% of pro forma EBITDA from copper, zinc, silver, lead), to shed ~$1.1B of rehabilitation obligations onto Alcoa, to cut ~$125M/yr of overhead and to reach ~$3.5B pro forma net cash. Its implied EVEnterprise valueWhat it would cost to buy the whole company: its market value plus its debt, minus the cash you would get with it. Often very different from market cap. is "up to $5.6B" incl. ~$750M net debt and leases assumed — "a through-the-cycle EBITDA multiple of ~6.8x" on ~$0.8B average FY21–FY25 EBITDA. Independent expert Grant Samuel values the Aluminium Business at $4.5–5.0B and the consideration at $4.8–5.0B — "towards the top end of the value range". Alcoa says "well below replacement cost"; the seller's expert says top of fair value. Both can be true; only the second is a valuation.

5. Alumina recovered above Alcoa's Q2 cash cost. WA FOB alumina $349/t on 2026-09-24; a 30 kt cargo traded at $369/t FOB WA on 2026-09-30 (Nov shipment) — vs API $307/t in Q2 and Alcoa's Q2 adjusted operating cost of $368/t. The "half the company sells below cash cost" condition that defined the August dossier is easing; Q3 realized price lags the index by about a month, so most of this shows up in Q4.

6. The calendar compressed: the South32 vote and Alcoa's Q3 print land on the same day. South32's AGM, which carries the disposal resolution (Resolution 7), is Thursday 15 October 2026, 12:00pm AWST (= 00:00 EDT 15 Oct). Alcoa reports Q3 after the NYSE close the same Thursday, 15 October 2026; call 5:00pm EDT (= Fri 16 Oct 08:00 AEDT — which is where the "Fri 16 Oct" date in the refresh brief comes from). The 5% ticking fee on $3.1B (~$13M a month) starts accruing from that approval; Alcoa's pro forma estimates it at $90M.

7. WA mining approvals may slip past year-end (disclosed on the Q2 call, missed by the prior dossier's partial transcript). Oplinger: "while my confidence in the outcome remains unchanged, the timing could extend beyond our original expectations" — the target had been ministerial approval by end-2026; a six-month contingency protects ore supply, beyond which Alcoa would "consider modifying mining operations and flow rate at the refineries to avoid an ore gap".

What held

The structural read stands: Alcoa is a price-taker on LME and the alumina index; the earnings mix is still aluminum-carried; the supply-chain map is unchanged; the forensic all-clear holds (0 SEC findings, controls effective as of 6/30). The August dossier's two biggest analytical claims mostly survive contact with the new filings: (a) AliGroup's economics live and die on alumina — confirmed, and the CVR strikes now show exactly where the seller drew the line; (b) Hillside's power contract is the least-analysed liability — confirmed and now quantified by the seller's own expert (Lens 13). One claim is corrected: the "management sold equity at $58.79, a revealed preference that the stock was not cheap" argument was framed as if Alcoa chose the stock component in June. The S-4 background shows the $3.1B cash / $1.0B stock split was agreed in October 2025, and in May 2026 Alcoa proposed more stock ($2.6B cash / $1.5B stock) — South32 countered back to $3.1B/$1.0B. The revealed preference is real but it dates to May 2026, when the shares traded far higher — and the fixed share count has since worked for Alcoa (Lens 9).

Phase A — Understand the business

Company Overview — CARRIED (with deal-perimeter facts added)

Unchanged; see the previous dossier Lens 1. Additions from the new filings: the perimeter is 86% of Worsley Alumina (mine + refinery), 100% of Hillside (and the idle Bayside site), 33% of MRN, 36% of the Alumar refinery and 40% of the Alumar smelter; Mozal (63.7%) was in Alcoa's April 2025 offer but is excluded (Mozal went on care and maintenance on 2026-03-15). AliGroup's audited combined balance sheet at 6/30/26: total assets $5.2B, PP&E $3.5B, equity (parent net investment) $2.5B under IFRS.

Supply Chain — CARRIED

Unchanged, see the previous dossier Lens 2. One quantified addition for the post-close rewrite: Hillside draws >1,200 MW at a 99.9% baseload factor, making it "one of Eskom's largest single customers", and exports ~70–80% of output via Richards Bay and Durban.

Competitive Advantages — REFRESHED (the synergy claim got a breakdown)

The prior dossier said the $900M NPV synergy figure "is not decomposed in any primary document I read." It now is. The CFO split it into three groups:

• ~30% — procurement, logistics, commercial: ~$50M/yr, inside 12 months of close.
• Process technology (Alcoa's refinery and smelter creep practices applied to Worsley and Hillside): starts in years 2–3. Implied ~30% of NPV.
• ~40% — WA life-of-asset mine planning: three refineries and two adjacent mine leases planned as one, to "avoid or defer mine moves", each worth "hundreds of millions of dollars."

Read: the largest synergy block is the one most plausibly unique to Alcoa — no other buyer owns the adjacent WA leases — and it is also the slowest and least verifiable (decades of avoided capexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs., NPV'd back). It is the strongest argument that Alcoa can pay a top-of-range price and still create value. The seller's own memorandum concedes the point by keeping South32 exposed to it through the shares: "retain exposure to potential value creation by Alcoa, including from synergies in Western Australian alumina."

The gallium plant (FID July 2026, ~100 t/yr) remains the only non-commodity moat; Alcoa's $24M is its "only expected contribution". Unchanged in substance.

Segments — CARRIED

No new quarter has been reported; see the previous dossier Lens 4 for the Q2 table (Alumina −$96M; Aluminum record $1,073M). The 3Q26 sequential bridges management gave are in Lens 5/8 below.

Phase B — Measure performance

Earnings Result — REFRESHED (no new print; what the Q2 release and the deeper history now show)

Latest print is still Q2 2026 (reported 2026-07-16). The Q2 results release, now on the shelf, adds the company-level figures the prior dossier drew from the 10-QForm 10-QThe quarterly version of the annual report. Lighter, and not audited.: Adjusted EBITDA ex-special items $901M (vs $595M Q1), adjusted EPS $2.12, OCF $608M, FCF $422M, net debt $873M and adjusted net debt $1,446M at 6/30/26 vs $851M / $1,464M at 12/31/25 (the release's reconciliation compares to year-end, not Q1; Q1-26 net debt was $1,089M). The prior dossier's consensus comparison ($2.12 vs $2.32) stands; another compiler had $2.25 — a ~$0.13–0.20 miss either way. Management blamed the miss on realization: "LME prices declined sharply in the final two weeks of June" and the 15-day-lag sensitivities "do not account for the steep changes near quarter end".

The ten-quarter view, now in our figures:

*Sales less COGS ex-DD&A, ÷ sales. Read: Q1 is a structural cash-burn quarter in three of three years (working capital), and Q2-26's gross margin is the second-highest of the ten quarters — below 2025-Q1's 27.6%, a tariff-premium quarter. The record EBITDA was volume plus price, not a margin regime change.

Q3 2026 — what management already told the market: Alumina segment +~$10M sequentially (Pinjarra +$30M recovery, energy +$5M, less Alumar/Juruti maintenance; fuel oil assumed $90/bbl); Aluminum segment flat ex-price (productivity offsets ~$15M higher carbon and seasonally lower Brazil energy sales; Section 232 Canada tariff cost −$10M, alumina input +$10M); operational tax $80–90M; FY D&A raised to ~$660M, other corporate to ~$180M. Caustic soda spiked in Q2 and hits Alcoa on a ~6-month lag — a Q4 headwind the CFO flagged.

Q3 consensus: n/a — not reliably sourced. Two web compilers disagree ($1.50 adjusted EPS on $3.92B revenue, 13 analysts, vs $2.27) and neither is dated cleanly. My own number is in Lens 11.

Earnings Calls (sentiment trend) — REFRESHED (six full calls now on the shelf)

The prior dossier had one partial excerpt. The shelf now holds full transcripts for 2025-Q1 through 2026-Q2. A simple count of recurring vocabulary across the six calls:

What the arc says.

1. Tariffs went from the whole conversation to a footnote (60 mentions → 7). In Q2-26 the only tariff exchange was a volume clarification: "No change in the rate, simply volume." Management and analysts both treat Section 232 as settled — which is why the September Section 338 escalation against Canada (Lens 8) is a live risk nobody on the call raised.
2. The deal was invisible until it wasn't. The 2026-Q1 call (16 April) has zero acquisition mentions, while the S-4 shows Alcoa had been negotiating with South32 since October 2025 and would send a revised offer on 3 May. Asked on that Q1 call whether the conflict changed the M&A-versus-buybacks weighing, Oplinger answered "The conflict hasn't changed our capital allocation framework" and walked through the framework. Accurate, and uninformative.
3. The Q2-26 Q&A is calmer than the prepared remarks. The prior dossier read the call as defensive from the 425 excerpt alone. The full Q&A is dominated by operations — Pinjarra (an oxalate outbreak compounded by the cyclone gas cut), casthouse capacity (~95% full in Europe and North America), restarts (Alumar ~95%), San Ciprián (smelter EBITDA covered refinery losses, but the site still consumes cash), and the Massena East data-centre sale, now exposed to New York's data-centre moratorium. Only one analyst question (Hillside power) touched the deal. The sell-side did not challenge the transaction on the call; the tape did.
4. New in Q2-26 and worth carrying: Oplinger's macro framing is that the LME retreat is "sentiment" — 3–3.5 Mt of Gulf capacity still offline, China creeping to 45–46 Mt (above the 45 Mt cap) "given the high metal price." That is a falsifiable claim, and September's further LME slide is evidence against it so far.

Net sentiment: operationally confident, financially conservative in tone, strategically ambitious — and notably more forthcoming on synergies at a September conference than on the July call.

Comps — REFRESHED (two new reference points from the deal itself)

The read: Alcoa is buying at ~6.5–6.8× EBITDA while its own equity trades at ~5×. Paying a higher multiple than your own is normally dilutiveDilutionIssuing new shares, so each existing share owns a smaller slice of the same company.; here the pro forma says it is accretive (Lens 11) only because ~$3.1B of the price is debt at ~6.7%, not equity at ~5×. That is a financing arbitrage, not a valuation bargain, and it reverses if LME falls far enough for the interest to bite. The Comps sheet in model.xlsx remains unpopulated with sourced peer multiples.

Stock-Price Catalysts — REFRESHED (with the print-day watch list)

Moves since the boundary:

• Mid-August → 2026-09-08: flat at ~$51 while LME rose to a six-week high; the deal filings (S-4 on 09-01, effective 09-08) did not move it.
• 2026-09-10/11: LME peaked ~$3,360 then fell 2.06% in a day; AA started a leg down.
• 2026-09-15: insider tax-withholding priced at $46.52 (Form 4Form 4The filing an insider must make within two business days of buying or selling their own company’s shares., Hastings) — ~−10% in a week.
• 2026-09-18: $44.43; 2026-09-30: $42.00 (the $42.06 figure is 10-01); 2026-10-02: $41.95. LME ended September ~$3,200–3,215; Chinese pre-holiday position-squaring, a weak US consumer-confidence print and near-even Fed hike odds were cited.
• Policy, not yet visible in the price: Section 338 duties on Canadian goods (50%, effective 2026-08-22) were made to stack on Section 232 from 2026-09-15, at up to 75% combined on derivative aluminum products — profiles, bars, rods, tubes, pipes. Whether any of Alcoa's Canadian value-added output (rod, billet) falls inside that perimeter is not determinable from what I read — flagged for the call.

Pattern update: the August ranking holds — LME first, M&A/balance-sheet second, operations third, tariffs fourth — and September made the first one more absolute: no company event moved AA by 5%, yet it fell −18.9% in ~3.5 weeks (09-08 $51.74 → 10-02 $41.95) on the metal.

THE PRINT — Thursday 15 October 2026, after the NYSE close; call 5:00pm EDT. What to watch, in order of information value:

1. Alumina segment: is it still negative? Guided +$10M from −$96M. Q3 realized price should rise modestly (index lagged ~1 month; Sept FOB WA $349–369). A Q3 alumina result near breakeven, or Q4 commentary implying positive, is the single largest read-through to AliGroup's value (alumina-weighted, +53% capacity).
2. Aluminum realized price vs ~9% lower LME. Q2 15-day-lag LME $3,585; Q3 lagged average probably ~$3,250–3,300. Watch the realized price, the Midwest/Rotterdam premiums, and whether the power-contract embedded derivatives give back part of the $376M next-12-month realized loss disclosed at 6/30 (they should shrink as LME falls — a partial natural hedge).
3. Interest and cash. First quarter with the new notes (from 09-23): ~$4M of coupon in Q3, ~$44M a quarter from Q4, against interest income on the parked cash. Watch period-end cash (should be ~$1.4B + ~$2.6B notes proceeds, less fees) and whether management states the pre-close negative carry.
4. South32 vote result (decided earlier the same day in Perth) and the ticking-fee start date; any regulatory-approval update (South Africa, US, Australia, EU, Brazil — per the CFO).
5. WA mining approvals: does the "could extend beyond our original expectations" language harden into a date slip past end-2026, and is the six-month contingency being drawn?
6. GAAP noise: Ma'aden mark-to-market (stake $1,360M at 6/30) will again swing GAAP EPS; deal costs (SG&A) and any bridge-commitment fees written off on termination. Anchor on segment EBITDA and OCF, not GAAP EPS.
7. Q4 guide: caustic (6-month lag, now correcting), carbon (still high), Section 338 exposure, Massena East sale status, buyback language (none expected).

Phase C — Judge people & books

Management — REFRESHED

No management changes. The 2026 annual meeting (May 6) re-elected the 11-nominee board and approved the amended Stock and Incentive Compensation Plan.

Insiders sold small amounts near $51; nobody bought at $42–46. EVP & General Counsel Andrew Hastings sold 6,000 shares at $50.90 (2026-08-17); EVP & CHRO Tammi Jones sold 15,628 at $50.60 (08-27); EVP & Chief Commercial Officer Renato Bacchi sold 700 at $51.02 (08-31); Hastings' 09-15 entries are vesting and tax withholding at $46.52. Immaterial in size, but the August–September tape has zero open-market insider purchases through a ~20% drawdown.

Capital allocation — the record is more nuanced than the August dossier said. The S-4 background:

• Oct 2024: Oplinger approaches South32's Graham Kerr at LME Week; NDA effective 2024-10-01.
• Apr 2025: Alcoa offers $4.8B cash + liabilities, including Mozal — rejected 25 April; Alcoa walks away on 5 May.
• Jul 2025: South32 re-opens the talks.
• Oct 2025: Alcoa's revised offer — $3.1B cash + $1.0B stock + up to $750M CVR, ex-Mozal — is accepted on cash and stock by 28 October; CVR terms settled 9 December.
• May 2026: Alcoa proposes $2.6B cash / $1.5B stock; South32 counters back to $3.1B / $1.0B.

Two things follow. First, the headline price was fixed in October 2025, before the 2026 war premium. This was not a deal struck at the top of a price spike, which the August dossier implied by calling it "pro-cyclical"; it was struck mid-cycle and signed near the top. Second, the fixed share count has already transferred value to Alcoa: 17,008,960 shares were sized at a $58.79 VWAP (~$1.0B); at $43.84 they are worth ~$746M — about $250M less consideration than the headline, which is the "risk-sharing" Oplinger described working as designed. The ~$1.1B of rehab provisions the seller says Alcoa assumes is booked by Alcoa at $397M of ARO under US GAAP (IFRS carrying value $1,119M) — see Lens 10.

Capex goes up with the assets: +$350–450M a year for AliGroup on top of Alcoa's own $750M (2026) / ~$800M (next three years). Post-close sustaining-plus-growth capex lands at roughly $1.15–1.25B a year.

Asset-monetization target unchanged: $500M–$1B by 2030; Massena East "substantially" negotiated but now exposed to New York's data-centre executive order.

Forensic Red Flags — REFRESHED

Prior findings stand (clean index-priced revenue recognition, the "Other segment items" trading gross-up, the $376M queued power-contract loss, routine asbestos/CERCLA). Five new or newly-quantified items from the deal filings:

1. The ARO gap: seller says ~$1.1B, Alcoa books $397M. AliGroup carries AROs of $1,119M under IFRS; the pro forma remeasures them to $397M at fair value under ASC 805 (a −$722M adjustment). This is legitimate — US GAAP fair-values an acquired ARO at a credit-adjusted rate, IFRS uses a risk-free-type rate and best estimate — but the cash obligation is closer to South32's figure. Lower booked ARO means less accretion expense and flatters pro forma earnings; the cash still has to be spent. Pro forma combined ARO: $1,422M.
2. Goodwill of only $116M on a ~$4.2B purchase price. PP&E is stepped up to $4,282M. Higher D&A follows: pro forma D&A $482M for 6M-26 vs Alcoa's $335M — ~$147M per half-year from AliGroup. Low goodwill means low impairment risk; the cost lands in the depreciation line instead.
3. The purchase-consideration accounting moves with the share price. Preliminary consideration $4,156M = shares $868M (at $51.05 on 2026-09-02) + cash $3,193M + CVR fair value $95M; "a 10.0% increase or decrease in the Alcoa stock price would result in an approximate $87 million" change. At $43.84 the share leg is ~$122M lower than in the pro forma.
4. The CVR is a live, mark-to-market liability with asymmetric strikes. Four years from 2026-07-01, no annual cap, $750M total cap. Aluminum: 22.5% of production × (LME − strike), strikes $3,500 (CY26) → $2,825 (CY27) → $2,847 → $2,870 → $2,942 (CY30). Alumina: 22.5% of 67% of production × (index − strike), strikes $345 (CY26) → $452 (CY27) → $456 → $459 → $471. At today's ~$3,200 LME the aluminum CVR is out of the money for CY26 but in the money from CY27 — on ~900 kt of acquired aluminum capacity, roughly ~$75M a year at a $3,200 LME. The alumina leg is near the money in CY26 (index ~$349–369 vs $345) and far out of the money from CY27 ($452). The seller kept its upside on aluminum, the metal that is currently profitable, and set alumina strikes that only pay in a boom. The $95M fair value will be remeasured through earnings.
5. Notes covenant structure: special mandatory redemption at 100% if the deal fails or misses the end date; 101% change-of-control put; up to 40% equity-clawback redemption before 2029-09-30. The SMR limits the downside of a failed deal to fees and negative carry. That is a genuine de-risking versus the bridge.

Regulatory findings (required, re-verified 2026-10-05):

• SEC Litigation Releases / AAERs: none — total_sec_findings: 0, period 2021-10-05 → 2026-10-05.
• Non-SEC enforcement (web): no 2026 Alcoa Corporation enforcement action surfaced; results returned only older Alcoa Inc. Clean Air Act matters and unrelated entities. Not exhaustive.
• 10-K Item 3 (Legal Proceedings): no new 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes. since the boundary; see the previous dossier and filings/10-k-2025-q4.md (unchanged).
• Merger-control approvals in progress: South Africa, US, Australia, EU, Brazil "and a couple other filings"; CFO: "So far, it's going very well". The ACCC acquisitions register carries an Alcoa–South32 entry.
• Conclusion: no fraud or securities findings. The live forensic issues are deal accounting: an ARO remeasured to roughly a third of the seller's figure, a CVR that will run through earnings, and a period of interest paid on financing for assets not yet owned.

Phase D — Project & stress-test

Forward Projection — REFRESHED

Model integrity first. our model re-ran and now sees 10 quarters (2024-Q1 → 2026-Q2), up from 2, so the seeded revenue growth is no longer a pure default. But our model still reports "No computed values — 51 formulas and no cached results"; the opening balance sheet is partly sourced (8/10 lines — Other current assets and Other current liabilities missing); the share price is 0 on the Assumptions sheet; and our figures shows seed values only. No figure below comes from the workbook. Every projection is with the arithmetic shown.

Macro marks used: LME ~$3,200–3,270 (late Sept); alumina FOB WA $349–369; fuel oil per guide $90/bbl.

Q3 2026. Aluminum segment: Q2 $1,073M − ~$230M LME (≈ 690 kt × ~$335/t lower lagged LME) + ~$30M partial power-derivative relief ± premiums ≈ $830–880M. Alumina segment: −$96M + $10M guided + ~$25M price (~1.6 Mt third-party × ~$15/t higher realized) ≈ −$60M. Total segment ≈ $780–820M (vs $977M Q2). Adjusted net income ≈ $562M − ~$175M after-tax ≈ ~$390M → adjusted EPS ~$1.45–1.55 on ~266M diluted shares. That sits near the low compiler's $1.50, so a Q3 "beat" or "miss" will turn on realization timing, not operations.

FY2026: H1 segment EBITDA $1,631M + Q3 ~$800M + Q4 ~$800M (LME ~$3,200 roughly flat sequentially; alumina improves toward breakeven on a ~$355–365 realized price, offset by caustic and the new notes' interest below the EBITDA line) ≈ ~$3.2B segment EBITDA, essentially unchanged from the August base ($3.15B) by a different route: weaker aluminum, better alumina. Adjusted EPS ≈ $1.40 + $2.12 + ~$1.50 + ~$1.45 ≈ ~$6.45; Q4 carries ~$44M of new coupon less ~$25M interest income on parked cash. GAAP will differ by the Ma'aden marks.

What the pro forma says about accretion — the most useful new arithmetic:

• 6M-26 (filed): pro forma basic EPS $3.64 vs standalone $3.15 (diluted $3.61 vs $3.13) → +15.5%.
• LTM to 6/30/26 (estimate, period-end shares): $1,482M ÷ 280.9M = $5.28 vs $1,277M ÷ 264M = $4.84 → ~+8–9%.
• FY2025 (filed) is DILUTIVE: pro forma basic EPS $4.20 vs $4.40 standalone (diluted $4.18 vs $4.37), ~−4.5%, because the as-if-1-January-2025 convention loads ~$249M of one-offs into that year — $93M inventory fair-value step-up, $100M Australian transaction taxes, $56M transaction costs incl. the bridge commitment fee. Ex those, FY2025 would be roughly neutral-to-accretive.
• These already include the financing cost ($83M after tax on 6M; $174M on FY25) and the ARO/PP&E remeasurement. They exclude the ticking fee (in purchase consideration), inventory step-up, CVR remeasurement, integration costs and synergies. Management's "accretive immediately" claim checks out at H1-2026 commodity prices on a run-rate basis; the first reported year will carry the one-offs above, as the filed FY2025 column shows. H1-2026 included the war-premium LME and a trough alumina price, so the accretion is not price-neutral.

FY2027 base — first stub year, close assumed ~mid-year: standalone segment EBITDA ~$2.9B (aluminum ~$730M a quarter at ~$3,100 LME; alumina ~breakeven at ~$360) − corporate/transformation ~$300M − D&A $660M − interest ~$270M (legacy ~$145M + new notes $175M − ~$50M H1 interest income) → pre-tax ~$1.67B; tax ~20% → ~$1.34B. Plus AliGroup H2: EBITDA ~$430M − D&A ~$150M − tax 25% ≈ +$210M, + ~$20M synergies after tax. ≈ $1.57B ÷ ~272.5M weighted shares ≈ ~$5.75. Less ~$75M a year of CVR accrual if LME holds ~$3,200 (from CY27; ~$37M in H2, after tax ~$30M) → ~$5.65.

FY2028 base — first full combined year: standalone ~$1.40B + AliGroup ~$425M (≈ $860M EBITDA − ~$295M D&A, 25% tax) + synergies ~$40M − CVR ~$60M after tax ≈ $1.80B ÷ 281M ≈ ~$6.40.

Bull / bear: Bull — LME $3,500, alumina $420 (CY27 alumina CVR still out of the money below $452): FY27 ~$8.00, FY28 ~$9.00. Bear — LME $2,700 (CVR pays nothing), alumina $320: standalone segment EBITDA ~$1.9B, AliGroup ~$450M: FY27 ~$2.90, FY28 ~$3.20; pro forma net debt ~$4.7B against ~$2.4B combined EBITDA ≈ ~2.0× — right at the ceiling management set.

EPS path: 2026 ~$6.45 adj · 2027 ~$5.65 · 2028 ~$6.40 (August: $6.20 GAAP / $5.30 / $5.75). FY27/28 rise because AliGroup's disclosed EBITDA sits at the top of the inferred range and alumina recovered; they are capped by interest accruing since 23 September 2026 and by the CVR. Treat the error bars as wide: the operative variable is LME, ±$300/t ≈ ±$1.50 of EPS. Sell-side FY27/FY28 consensus: n/a. The only market cross-check is the forward P/E of 7.45 at $43.84 → implied next-year EPS ~$5.88, which sits between my FY27 and FY28.

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

Bull vs Bear — REFRESHED

Bull case. At $43.84 you buy a business that just earned $2.29B of LTM adjusted EBITDA for ~4.8×, and the deal it is closing is accretive by ~8–15% on the GAAP-converted pro forma at H1-2026 commodity prices (the filed FY2025 column is ~4.5% dilutive because it carries ~$249M of one-offs), financed with fixed-rate 2034/2036 paper at ~6.7% with no bridge refinancing risk left, and with a special mandatory redemption that caps the cost of failure. The variable that made August look dangerous has turned: alumina is ~$350–370, at or above Alcoa's Q2 cash cost, which matters doubly because the acquired portfolio is alumina-weighted. The share component has already deflated by ~$250M in Alcoa's favour. The largest synergy block (WA mine-plan integration, ~40% of $900M NPV) is one only Alcoa can capture. Rehab liabilities that cost South32 its patience are booked by Alcoa at $397M. The CVR's alumina strikes ($452+ from CY27) mean Alcoa keeps nearly all of an alumina recovery. Gallium, ELYSIS and the Section 232 onshoring program remain unpriced options. Analysts' targets sit ~43% above spot.

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

1. Alcoa is paying a higher multiple than its own stock, with debt, ahead of the asset. ~6.5–6.8× EBITDA for AliGroup (the seller's expert: top of a $4.5–5.0B fair range) versus ~5× for AA itself, with ~$175M a year of coupon accruing from 23 September 2026 and the assets arriving H1 2027 at the earliest. If LME drifts to $2,700–2,800, the combined company sits at ~2.0× leverage — the ceiling — before synergies, with $1.15–1.25B a year of capex to fund.
2. The seller kept the aluminum upside. The aluminum CVR strike drops to $2,825 from CY27, below spot, so ~22.5% of the acquired smelters' price upside above ~$2,850 goes back to South32 until mid-2030 (within a $750M cap). The acquired smelting exposure is therefore less levered to aluminum upside than Alcoa's own, and fully exposed to the downside.
3. Hillside is a binary asset with a 2031 cliff. The seller's expert values it at $1.9–2.1B — roughly 40% of the whole Aluminium Business — and the NPV-scenario chart in that report spans roughly $1.1B to $2.9B (read off the chart's axis ticks; the bars are Low/High commodity cases across power scenarios, and individual values are not tabulated). Grant Samuel says the NPV is "extremely sensitive to aluminium prices … as well as its ability to access reliable and affordable power" — so that span mixes metal-price and power risk. Grant Samuel describes the operating life as "largely tied to its ability to secure new and reliable power supply agreements" and calls the status quo "increasingly problematic to sustain" as ~15 GW of coal capacity retires by FY35. Alcoa's answer on the July call was process, not numbers.

Pre-mortem (18 months out — April 2028). LME never recovered above $3,000 after the Gulf restarts; China crept to 46 Mt; the Midwest premium compressed. The deal closed in Q2 2027 on schedule. That was the problem: Alcoa had paid $90–130M of coupon before owning the assets ($180–220M including the $90M ticking fee), issued 17M shares that South32 distributed and its holders sold, and inherited Hillside with four years of power certainty left. Alumina stayed ~$340–360, enough to stop the bleeding and too little to make the 5.2 Mt earn its multiple. The CVR paid out on aluminum in 2027 and 2028 even though Alcoa's own smelting margins were mediocre, because the strike sat below spot. WA approvals slipped two quarters and the refineries ran at reduced flow. Leverage touched 2.2×, Moody's moved the outlook to negative, and the stock re-rated to trough multiples on trough earnings. No single error — a fair deal at a full price, financed early, into a flat tape.

Are multiples too high? No. At ~4.8× LTM EV/EBITDA and ~7.5× forward earnings the stock prices in a meaningful LME decline already. The question is the same as in August — is the E about to be spent? — and it now has a numerical answer: yes, at ~6.5×, partly with the E's own debt capacity, and accretively at H1-26 prices.

Contrarian view. The market is treating AA purely as an LME derivative — −18.9% in ~3.5 weeks, no company news. It is ignoring that alumina has quietly moved from below Alcoa's cash cost to above it. That move is worth more to the post-deal company than to the pre-deal one, and it does not appear in a share price that follows aluminium tick for tick. Against that, the market is also ignoring a risk: Section 338 duties now stack on Section 232 for Canadian aluminum derivative products, and the Q2 call shows nobody — management or sell-side — discussing tariffs any more.

Devil's Advocate (short-seller) — REFRESHED

• The "below replacement cost" line meets the seller's fairness opinion. The seller's own independent expert says Alcoa is paying the top of a $4.5–5.0B value range. "Below replacement cost" and "top of fair value" are both true only if nobody should replace this capacity. That was the August critique, and it now has the seller's signature on it.
• Accretion is a financing artefact. Pro forma EPS rises because ~75% of the price is debt at ~6.7% and the target's EBITDA yield (~15% on $5.6B) exceeds it at H1-2026 prices. Rerun at LME $2,700 and alumina $320 (AliGroup EBITDA ~$450M) and the target yields ~8% on $5.6B before D&A and capex: accretion collapses toward zero while the leverage stays.
• Negative carry is now certain, not a risk. ~$175M a year of coupon accruing from 23 September 2026; ~6–9 months of it before close, partly offset by interest on the cash — plus the $90M ticking fee from mid-October. Roughly $150–200M of pre-close cost, about 4.8–6.5% of the $3.1B cash price, is spent before Alcoa earns a dollar from Worsley.
• The ARO arithmetic flatters the deal. $1,119M of IFRS rehab obligations becomes $397M of US GAAP ARO. The cash cost of closing and rehabilitating Worsley's mine footprint does not fall by $722M because the discount rate changed.
• The CVR is a quiet short on Alcoa's own upside. From CY27 the aluminum strike (~$2,825–2,942) sits below spot. Every rally above ~$2,850 pays South32 22.5% of the acquired smelters' upside, with no annual cap, until mid-2030.
• Hillside's value is wide-banded. The expert selected $1.9–2.1B; its NPV-scenario chart spans roughly $1.1–2.9B (read off the axis ticks; bars are Low/High commodity cases across power scenarios, values not tabulated), and the expert ties the result to both aluminium prices and reliable, affordable power. A combined weak-metal, poor-power outcome near the bottom of that chart would leave Hillside worth materially less than its selected range — not quantifiable more precisely from the text — in a jurisdiction where Alcoa has never operated.
• Single scenario that permanently impairs: LME ≤$2,800 through 2027 + alumina ≤$340 + WA approval delay beyond the six-month contingency + Hillside power renegotiated near the low case. Plausibility: moderate-to-low — four conditions, each individually plausible, partly correlated through China.
• If results disappoint 20–30%: a 25% haircut to FY27E (~$5.65 → ~$4.25) at an unchanged ~7.5× ≈ ~$32 (−27%).

Management Questions — CARRIED (re-ordered)

The 15 questions in the previous dossier Lens 14 still apply. The August dossier's six deal questions are re-ranked for the 15 October call. (c) permanent financing is answered (6.625%/6.875% notes) and drops off; (a) is partly answered (LTM EBITDA ~$861M), but cost per tonne at spot alumina is not. In information-value order:

1. At today's ~$3,200 LME and ~$360 alumina, what is AliGroup's run-rate EBITDA, and how much has accrued in the locked box since 2026-04-01 (the ">$200M at 6/30" estimate)?
2. What is the net pre-close carry — coupon on the $2.6B notes less interest income — per quarter, and where do the bridge commitment fees land in Q3?
3. What is the CVR fair value at 9/30, and how much do you expect to pay in CY27 at current strikes?
4. Hillside: what is the base-case power-price assumption in your valuation for post-2031, and what is the EBITDA sensitivity to a 2× reset? (Restated from August (d), now with the seller-expert selected value of $1.9–2.1B and its NPV chart's ~$1.1–2.9B axis span — Low/High commodity cases across power scenarios — to answer against.)
5. WA mining approvals: is ministerial approval still expected by end-2026, and how much of the six-month contingency has been used?
6. Does any of Alcoa's Canadian product (billet, rod, foundry) fall under the Section 338 stacked duties, effective 2026-09-15?
7. Has Alcoa filed for the July 2026 Section 232 onshoring program (Warrick/Massena), and what is the duty saving? (August (e).)
8. At what share price does the buyback beat M&A on your own hurdle rate? (August (f), sharper at $42–44 with zero insider buying.)