Dreams AI Associate

←

Aluminium Bahrain B.S.C. (Alba)

Bahrain Bourse: ALBH · Initiation of coverage

Rating

Over-Weight

Fair value

BHD 1.150

Analyst

S. Rahman

Status

published

Stage 3 · Valuation

Generation complete: ALBH_Valuation_Analysis_2026-08-28.md

accepted

Associate summary: Generation complete: ALBH_Valuation_Analysis_2026-08-28.md

# VALUATION ANALYSIS: Aluminium Bahrain B.S.C. (Alba)

**Ticker:** BHB: ALBH (Bahrain Bourse); ALBH:LI (London Stock Exchange, GDR)
**Sector:** Materials / Primary Aluminium
**Peer set (as instructed):** SABIC (2010.SR), Emirates Global Aluminium (EGA, unlisted), Alcoa Corporation (NYSE: AA)
**Valuation date:** 28 August 2026 (market data as at the close of 26 August 2026)
**Workflow stage:** Task 3 of 5, initiating-coverage skill
**Governing contract:** dreams-associate/contracts/ic-report-structure.json v1.0.0
**Model of record:** dreams-associate/reports/ALBH/Task2_Model/ALBH_Financial_Model_2026-08-28.xlsx
**Prepared for:** Al Ramz Investment Research

---

## RECOMMENDATION

| | |
|---|---|
| **Last close** | **BHD 0.878** (26 August 2026) |
| **Fair value** | **BHD 1.150** |
| **Upside** | **+31.0%** |
| **Stock rating** | **OVER-WEIGHT** |
| Rating thresholds at the last close | Over-Weight above BHD 1.010; Under-Weight below BHD 0.746 |
| Market capitalisation | BHD 1,243m / US$3,306m |
| Enterprise value, post-acquisition | US$5,066m |
| Shares outstanding, net of treasury | 1,416.0m |
| Free float | 10.00% |
| Consensus | Strong Buy, 4 analysts, mean target BHD 1.450, range BHD 1.30 to 1.75 |

We initiate on Aluminium Bahrain at Over-Weight with a fair value of BHD 1.150, 31% above the 26 August 2026 close of BHD 0.878. Our fair value is deliberately 21% below the consensus mean of BHD 1.450. We are constructive on this equity, but less constructive than the street, and we think the reasons matter more than the conclusion.

The rating is not a comfortable one. The Over-Weight threshold is BHD 1.010 and three of our six valuation methods sit within 5% of it. Over-Weight survives a WACC of 13.68% and fails above it. It survives a mid-cycle LME of US$2,700/t at restored volumes and fails much below that. In our view the correct posture on Alba today is a constructive one held with a clear view of exactly what would break it.

---

## 1. EXECUTIVE SUMMARY

### 1.1 The valuation question

Alba trades at 3.94x trailing earnings and 0.575x book (Investing.com and TradingView, accessed 28 August 2026; both figures independently reproduced from Alba's own reported EPS of 223 fils trailing and book value per share of BHD 1.5279 at 30 June 2026). It has traded below book in every year since 2021. The shares are down 20.5% year to date against trailing earnings per share up 45%.

That combination usually means one of two things. Either the market has made a mistake, or the market is telling you the earnings are not real. Our work says the market is substantially right about the earnings and substantially wrong about the magnitude of the discount.

Alba's H1 2026 profit attributable to shareholders rose 228% to US$372.8m while Q2 net finished production fell 61% year on year to 155,469 tonnes (Alba Q2 and H1 2026 results release, 4 August 2026). Q2 output annualises to roughly 622kt against a 1.62 Mt nameplate, which is approximately 38% utilisation. The company is earning record profits from a plant running at just over a third of capacity, because LME cash averaged US$3,576/t in Q2 2026, up 46% year on year, and realised premiums went with it.

The whole valuation therefore reduces to two questions. What is the mid-cycle aluminium price, and how much of Alba's 1.62 Mt comes back and when.

### 1.2 What the market is discounting

We ran the base-case model backwards to solve for what the BHD 0.878 close implies. At our 13.27% WACC and 2.0% terminal growth, the market is discounting a perpetual mid-cycle LME of approximately **US$2,644/t**, assuming a full volume recovery to 1,620kt. That is below the bottom of the US$2,650 to US$2,750/t CRU mid-cycle band Alba itself discloses, and materially below the US$2,800 to US$3,000/t structural equilibrium floor that independent sector work now puts on the 2028 to 2030 cost curve (Axis Securities, *Global Aluminium Outlook*, 10 June 2026).

Put the other way, on the unchanged base-case model the market is applying a **15.02% WACC**, an implied cost of equity of roughly 18.4%.

In our view the market is pricing Alba at the bottom of mid-cycle and giving essentially zero credit for the cycle, for the disruption premium, or for the Ma'aden mark of eighteen months ago. That is too harsh, but it is not absurd, which is why our fair value sits closer to the market than to consensus.

### 1.3 The single most important finding

Tasks 1 and 2 both flagged Aluminium Dunkerque as the largest information gap in this coverage. Alba announced a US$2.2bn acquisition of the European Union's largest primary smelter on 2 June 2026 and has disclosed no revenue, no EBITDA and no balance sheet for the target. Neither has American Industrial Partners nor Bpifrance. EFG Hermes' Yousef Husseini put the problem precisely on 5 June 2026: "On a capacity basis, the Dunkerque valuation seems rich, but could be justified depending on its earnings and cash flow, we don't know those figures yet" (AGBI, 5 June 2026).

**We now know those figures.** The holding company being sold, Aluminium Dunkerque Industries France (ADIF, SIREN 834746661), files consolidated statutory accounts with the French registry. The FY2025 *comptes consolidés* were filed on 23 April 2026.

| ADIF consolidated (EUR m) | FY2022 | FY2023 | FY2024 | FY2025 |
|---|---|---|---|---|
| Revenue | 931 | 836 | 834 | 895 |
| **EBITDA (EBE)** | **104** | **169** | **317** | **292** |
| EBIT | 52.4 | 127 | 259 | 227 |
| Net income | not disclosed | not disclosed | 154 | 152 |
| Equity | 197 | 266 | 414 | 533 |
| Net debt | 438 | 217 | 220 | **201** |
| EBITDA margin | 11.1% | 20.2% | **38.0%** | **32.6%** |
| Net debt / EBITDA | 4.2x | 1.3x | 0.7x | **0.7x** |

*Source: Pappers.fr, sourcing INPI / RNE statutory filings; accessed 28 August 2026.*

At the ECB fixing of EUR/USD 1.1641 on 27 August 2026, FY2025 EBITDA of EUR 292m is **US$340m**, which puts the US$2.2bn price at **6.47x trailing EBITDA**, and FY2024's EUR 317m at **5.96x**. That is *below* the 6.15x at which Alba's own shares changed hands in the Ma'aden transaction. The per-tonne optic is rich at US$7,333/t. The cash-flow optic is not.

The bridge between the two optics is the power contract, and it is the reason we do not simply mark Dunkerque up. Aluminium Dunkerque draws 450 MW continuously, roughly 3.94 TWh a year, and is France's largest single industrial electricity consumer. **The ARENH regulated price of EUR 42/MWh expired on 31 December 2025.** A 10-year EDF supply contract signed on 22 July 2025 runs from 1 January 2026 covering "une part substantielle des besoins", with an undisclosed significant upfront advance to EDF. The price is not disclosed, and the successor regime to ARENH is expected to be stable but on average higher (AFP via Connaissance des Énergies, 22 July 2025).

Both the FY2024 and FY2025 EBITDA figures above were earned under ARENH. That is the whole point.

### 1.4 Summary of methods

| Method | Low | **Base** | High | Weight | Contribution |
|---|---|---|---|---|---|
| Discounted cash flow, Task 2 model as built | 0.884 | **1.056** | 1.397 | 30% | 0.317 |
| Discounted cash flow, Dunkerque at its statutory accounts | 1.040 | **1.229** | 1.458 | 15% | 0.184 |
| Scenario probability-weighted DCF | 1.010 | **1.214** | 1.407 | 15% | 0.182 |
| Trading comparables on mid-cycle EBITDA | 0.843 | **1.051** | 1.383 | 20% | 0.210 |
| Alba's own historical multiples | 1.061 | **1.206** | 1.335 | 10% | 0.121 |
| Precedent transactions and asset value | 0.931 | **1.318** | 1.799 | 10% | 0.132 |
| **BLENDED FAIR VALUE** | **0.941** | **1.146** | **1.439** | **100%** | **1.146** |
| **Rounded** | | **1.150** | | | |

*Source: Al Ramz Investment Research, Company financials.*

The base-case range across six methods is BHD 1.051 to BHD 1.318, a spread of only 25%. That is unusually tight for a single-asset cyclical producer whose plant was hit by a missile five months ago, and it is tight because every method is ultimately anchored on the same two facts: Alba trades at 0.575x book, and a strategic buyer with full diligence access paid BHD 1.240 a share for a fifth of the company in February 2025.

---

## 2. DISCOUNTED CASH FLOW

### 2.1 Structure

We value Alba as at **31 December 2026**, so the derived figure is a twelve-month forward fair value from the 26 August 2026 close. The explicit forecast period is FY2027E to FY2030E. FY2026E cash flow is deliberately excluded from the discounting because the bridge uses FY2026E closing net debt, which already contains it; discounting FY2026E cash flow and then deducting FY2026E net debt would double count.

This also handles the Aluminium Dunkerque acquisition cleanly. The US$2,083m acquisition facility drawn to fund it sits inside FY2026E net debt of US$1,759m, and Dunkerque's earnings enter the forecast from FY2027E under the Task 2 consolidation switch.

### 2.2 Cost of capital

Task 2 carried a 6.20% risk-free rate explicitly flagged as a placeholder for re-sourcing at Task 3. We have re-sourced it.

| Component | Value | Basis |
|---|---|---|
| Risk-free rate, Bahrain sovereign 10-year USD | **7.125%** | Reoffer yield on the Kingdom of Bahrain US$1.0bn issue maturing 10 June 2036, priced 10 June 2026 into an order book above US$3.2bn |
| Equity risk premium, mature market | 5.50% | Al Ramz house assumption. Damodaran carries 4.23% at the 5 January 2026 vintage, so the house figure is the more conservative |
| Levered beta | 1.35x | Peer unlevered beta median is 0.58x on vendor betas and 1.09x on five-year weekly regressions against home indices; relevered at 25% target gearing that is a 0.75x to 1.40x range. We take the top |
| Single-asset concentration and liquidity premium | 1.50% | Cut from the 2.00% Task 2 placeholder because the sourced sovereign yield now carries the country component the placeholder did not |
| **Cost of equity** | **16.05%** | |
| Pre-tax cost of debt | 5.80% | Weighted average of the facilities in audited note 13 |
| Marginal tax rate | 15.00% | Bahrain Domestic Minimum Top-up Tax, Decree Law 11 of 2024 |
| Target debt / total capital | 25.0% | Current market gearing is 34.7%, which would give a 12.19% WACC and a higher fair value |
| **WACC** | **13.27%** | |
| Terminal growth | 2.00% | Below long-run global aluminium demand growth of roughly 2.2% |

*Source: Al Ramz Investment Research, Bloomberg, Company financials.*

Three points on this build, because the discount rate is the single largest swing factor in the valuation.

**First, we use the observed Bahrain sovereign yield rather than a ratings-implied spread.** Bahrain is rated B by S&P (stable, 21 November 2025), B2 by Moody's (negative, 18 April 2026) and B by Fitch (stable, 23 February 2026), having been downgraded from B+ by both S&P and Fitch in the last year. Damodaran's adjusted default spread for Bahrain of 4.67% added to the US 10-year Treasury of 4.65% would imply a 9.32% sovereign yield, against the 7.125% Bahrain actually paid in the primary market in June 2026. The ratings-implied spread is materially wider than the observed market yield. We use the market.

**Second, we disclose what the alternative construction gives.** Building the discount rate the Damodaran way instead, on a 7.12% Bahrain country risk premium over a US risk-free rate, produces a 14.34% WACC and a BHD 0.94 DCF value, which would be Equal-Weight. That is the bear end of the cost-of-capital debate and we would rather state it than suppress it.

**Third, our beta is the conservative choice.** Alba's own regression beta is unusable: five-year weekly against MSCI ACWI gives 0.053 with an R-squared of 0.001, which is the honest signature of a 10% free float on a frontier exchange. TradingView publishes 2.81 without disclosing its benchmark. Neither is usable, so we build from peers and take the top of the relevered range.

### 2.3 Free cash flow and terminal value

| US$m | FY2027E | FY2028E | FY2029E | FY2030E |
|---|---|---|---|---|
| EBIT | 1,101 | 1,110 | 851 | 708 |
| NOPAT | 1,046 | 944 | 723 | 602 |
| Add: depreciation and amortisation | 519 | 530 | 543 | 556 |
| Less: capital expenditure | (340) | (360) | (370) | (380) |
| Less: increase in net working capital | (195) | (122) | (61) | (8) |
| **Unlevered free cash flow** | **1,029** | **992** | **835** | **770** |
| Discount factor at 13.27% | 0.883 | 0.779 | 0.688 | 0.607 |
| **Present value** | **909** | **773** | **574** | **468** |

*Source: Al Ramz Investment Research, Company financials. Table binds to the Task 2 DCF Inputs tab, row 15.*

FY2030E is a deliberately mid-cycle terminal year: LME at US$2,750/t, production back at the pre-disruption 1,620kt run rate, and 1,910kt of consolidated sales including Dunkerque. It is a legitimate normalised base rather than a cycle peak, and we would note that US$2,750/t is **below** the CY2030 threshold of US$2,942/t that Alcoa and South32 negotiated into the contingent consideration of their AliGroup transaction on 1 July 2026. Two sophisticated counterparties structuring a US$5.6bn deal put mid-cycle 2030 aluminium US$192/t above where our terminal year sits.

We make one analyst adjustment to the terminal year, and it costs us BHD 0.26 a share. The Task 2 model carries FY2030E capital expenditure at 0.68x depreciation. That is defensible in the years immediately after a curtailment but it is not a perpetuity assumption, so we normalise capital expenditure to equal depreciation in the terminal year.

| Terminal value build | US$m |
|---|---|
| Terminal year NOPAT (FY2030E) | 602 |
| Add: depreciation and amortisation | 556 |
| Less: normalised capital expenditure, set equal to depreciation | (556) |
| Less: terminal working capital investment (2.0% x US$1,594m) | (32) |
| **Normalised terminal free cash flow** | **570** |
| Memo: unadjusted FY2030E unlevered free cash flow | 770 |
| **Terminal value at 31 December 2030** | **5,159** |
| **Present value of terminal value** | **3,134** |

### 2.4 Enterprise to equity bridge

| US$m unless stated | |
|---|---|
| Present value of FY2027E to FY2030E unlevered free cash flow | 2,724 |
| Present value of terminal value | 3,134 |
| **Enterprise value** | **5,858** |
| Terminal value as % of enterprise value | 53.5% |
| Less: net debt at 31 December 2026E | (1,759) |
| Less: non-controlling interest (Bpifrance 6% of the Dunkerque holding company) | (120) |
| **Equity value** | **3,979** |
| Shares outstanding, net of treasury (m) | 1,416.0 |
| **Fair value per share (US$)** | **2.81** |
| **Fair value per share (BHD)** | **1.056** |
| **Upside to the DCF value** | **+20.3%** |

Terminal value at 53.5% of enterprise value is comfortably inside the 70% threshold at which an explicit forecast is judged too short.

### 2.5 What the DCF value implies

| Sanity check at BHD 1.056 | Implied | Reference |
|---|---|---|
| EV / FY2030E mid-cycle EBITDA | 4.64x | Core smelter peer median 6.59x; Alcoa/South32 negotiated 6.80x |
| Terminal exit multiple (TV / FY2030E EBITDA) | 4.08x | Primary aluminium assets transact and trade at 4x to 6x mid-cycle |
| EV per tonne of annual capacity | US$3,046/t | Peer median US$6,201/t; Chalco, the cheapest peer, US$3,527/t |
| EV as % of the Alba Line 6 whole-scope build cost | 54.8% | Line 6: US$3.0bn for 540kt including Power Station 5 |
| P/E on FY2027E earnings | 4.59x | Alba's own six-year band 2.51x to 13.80x, median 5.45x |
| P/E on FY2030E mid-cycle earnings | 7.15x | |
| P / FY2026E book value | 0.67x | Alba's own band 0.575x to 0.96x, median 0.80x |
| FY2030E dividend yield at the stated 35% payout | 5.5% | Alba's own band 2.55% to 10.79%, median 5.53% |

Every one of these sits at or below Alba's own historical median and well below the peer set. We are comfortable that the DCF is not stretching.

---

## 3. SENSITIVITY ANALYSIS

### 3.1 Table 1: WACC vs terminal growth rate

Fair value per share, BHD. Base case shaded.

| WACC \ g | 1.0% | 1.5% | **2.0%** | 2.5% | 3.0% |
|---|---|---|---|---|---|
| **11.27%** | 1.252 | 1.294 | **1.340** | 1.392 | 1.449 |
| **12.27%** | 1.116 | 1.149 | **1.185** | 1.224 | 1.268 |
| **13.27%** | 1.002 | 1.028 | **1.057** | 1.087 | 1.121 |
| **14.27%** | 0.905 | 0.926 | **0.948** | 0.973 | 0.999 |
| **15.27%** | 0.821 | 0.838 | **0.856** | 0.875 | 0.896 |

*Source: Al Ramz Investment Research. Over-Weight requires above BHD 1.010; Under-Weight below BHD 0.746.*

The Over-Weight rating survives every combination at or below a 13.27% WACC and fails at 14.27% and above at any growth rate below 3.0%. **The break-even WACC for the rating is 13.68%.** Under-Weight is not reached until a WACC of 16.68%, which we regard as implausible.

### 3.2 Table 2: WACC vs exit EV/EBITDA multiple on FY2030E mid-cycle EBITDA

The cyclical convention, testing the same valuation from the multiple side.

| WACC \ exit | 3.5x | 4.0x | 4.5x | 5.0x | 5.5x | 6.0x |
|---|---|---|---|---|---|---|
| **11.27%** | 1.020 | 1.129 | 1.239 | 1.348 | 1.458 | 1.567 |
| **12.27%** | 0.978 | 1.084 | 1.189 | 1.295 | 1.400 | 1.506 |
| **13.27%** | 0.938 | 1.040 | 1.142 | 1.244 | 1.345 | 1.447 |
| **14.27%** | 0.899 | 0.998 | 1.096 | 1.194 | 1.293 | 1.391 |
| **15.27%** | 0.862 | 0.957 | 1.053 | 1.148 | 1.243 | 1.339 |

*Source: Al Ramz Investment Research.*

Our Gordon growth terminal value implies a 4.08x exit. At the peer core-smelter median of 6.59x the fair value would be roughly BHD 1.50. We are not taking that.

### 3.3 Table 3: mid-cycle LME price vs terminal Alba production

This is the table that actually decides the equity. Alba is a price taker with a fixed cost base, so mid-cycle metal price and restored volume between them explain almost all of the valuation dispersion.

Fair value, BHD. Terminal EBITDA is flexed by the change in LME against the US$2,750/t base at a 75% net pass-through, because roughly a quarter of an LME move is given back through the alumina price index over time. Aluminium Dunkerque sales of 295kt are held constant.

| LME (US$/t) \ Alba production (kt) | 1,450 | 1,550 | **1,620** | 1,720 | 1,820 |
|---|---|---|---|---|---|
| **2,400** | 0.275 | 0.330 | 0.368 | 0.423 | 0.478 |
| **2,550** | 0.545 | 0.616 | 0.665 | 0.736 | 0.806 |
| **2,700** | 0.816 | 0.902 | 0.962 | 1.048 | 1.134 |
| **2,750** | 0.906 | 0.997 | **1.061** | 1.152 | 1.243 |
| **2,900** | 1.177 | 1.283 | 1.358 | 1.465 | 1.571 |
| **3,050** | 1.447 | 1.568 | 1.655 | 1.778 | 1.900 |

*Source: Al Ramz Investment Research, LME, Company financials.*

Read across the base row. At the model's US$2,750/t and 1,620kt the fair value is BHD 1.061. At the bottom of the CRU band Alba discloses, US$2,650/t, at the same volume, it is roughly BHD 0.98, which is Equal-Weight. At US$2,900/t, still below the Alcoa/South32 negotiated CY2030 threshold of US$2,942/t, it is BHD 1.358.

The volume axis matters less than the price axis, which is the correct result for a producer whose cash cost is roughly 80% of its realised price. A 170kt shortfall against the 1,620kt base costs roughly BHD 0.09. A US$150/t shortfall in the metal price costs roughly BHD 0.28.

### 3.4 Table 4: Aluminium Dunkerque steady-state EBITDA

The largest unforced band in the valuation.

| Dunkerque steady-state EBITDA (US$m) | US$/t on 300kt | Fair value (BHD) | Upside | vs base case |
|---|---|---|---|---|
| 100 | 333 | 1.037 | +18.2% | (0.058) |
| 150 | 500 | 1.136 | +29.4% | 0.041 |
| **200** | **667** | **1.234** | **+40.5%** | **0.139** |
| 250 | 833 | 1.332 | +51.7% | 0.237 |
| 300 | 1,000 | 1.430 | +62.9% | 0.335 |
| 376 | 1,253 | 1.579 | +79.9% | 0.484 |
| 440 | 1,467 | 1.705 | +94.2% | 0.610 |

*Source: Al Ramz Investment Research.*

The consolidated fair value moves by roughly BHD 0.20 a share for every US$100m of steady-state Dunkerque EBITDA. Section 5 sets out where within this range we think the answer sits and why.

---

## 4. COMPARABLE COMPANIES ANALYSIS

### 4.1 Peer set and the selection problem

The instructed peers are SABIC, Emirates Global Aluminium and Alcoa. Only one of the three is a usable trading comparable.

**SABIC produces no primary aluminium.** It is a diversified petrochemicals company and is in the set only because it held 20.62% of Alba until February 2025. We show its multiples for completeness and exclude them from every derivation.

**EGA is unlisted.** It is by some distance the closest structural comparable to Alba: the same gas-powered Gulf model, similar product mix, similar customer geography, and it was struck in the same 28 March 2026 attacks. It publishes no statutory accounts and no bond prospectus exists. We can nonetheless derive an implied multiple, because EGA has appointed Citi, Goldman Sachs, Emirates NBD Capital and First Abu Dhabi Bank for an ADX listing at a reported US$10bn to US$15bn valuation.

**Alcoa is the only listed pure-play of genuine size**, but 45% of its FY2025 segment EBITDA came from alumina rather than aluminium, so its EV per tonne of primary aluminium overstates a smelting-only multiple.

We therefore extend the set. Of thirteen names, seven are "core smelters" whose value is predominantly primary aluminium smelting, and those seven are our reference set.

### 4.2 Peer table

| Company | Ticker | Core | Mkt cap (US$bn) | Net debt (US$bn) | EV (US$bn) | FY2025 revenue (US$bn) | FY2025 EBITDA (US$bn) | EBITDA margin | **EV/EBITDA FY** | EV/EBITDA LTM | P/E trail | P/E fwd |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Alcoa | AA | Yes | 13.51 | 0.87 | 14.38 | 12.83 | 1.98 | 15.5% | **7.25x** | 6.22x | 10.61x | 8.76x |
| Emirates Global Aluminium | unlisted | Yes | 12.50 | 3.41 | 15.91 | 8.71 | 2.53 | 29.0% | **6.29x** | n/a | n/a | n/a |
| Norsk Hydro | NHY.OL | Yes | 19.39 | 1.04 | 20.43 | 22.30 | 3.10 | 13.9% | **6.59x** | 3.92x | 18.53x | 9.59x |
| Century Aluminum | CENX | Yes | 4.54 | 0.14 | 4.68 | 2.53 | 0.43 | 16.8% | **11.01x** | 9.77x | 7.62x | 3.39x |
| Vedanta | VEDL.NS | Yes | 11.51 | 5.57 | 17.08 | 18.22 | 5.86 | 32.2% | **2.91x** | 4.59x | 16.26x | 8.55x |
| Chalco | 2600.HK | Yes | 23.22 | 5.28 | 28.50 | 35.88 | 5.83 | 16.3% | **4.89x** | 4.13x | 7.18x | 6.04x |
| China Hongqiao | 1378.HK | Yes | 28.98 | 3.18 | 32.16 | 24.16 | 7.05 | 29.0% | **4.56x** | 4.59x | 7.45x | 5.94x |
| Press Metal | PMETAL.KL | Yes | 16.38 | 0.42 | 16.80 | 4.02 | 0.84 | 20.9% | **20.00x** | 15.45x | 25.59x | 26.12x |
| Hindalco | HINDALCO.NS | No | 23.75 | 7.30 | 31.05 | 28.78 | 3.99 | 13.9% | 7.78x | 7.34x | 13.86x | 9.03x |
| Rio Tinto | RIO | No | 178.83 | 14.36 | 193.19 | 57.64 | 25.36 | 44.0% | 7.62x | 8.68x | 14.78x | 12.89x |
| South32 | S32.AX | No | 16.79 | (0.28) | 16.51 | 5.82 | 2.46 | 31.0% | 6.71x | 12.90x | 15.34x | 17.02x |
| Ma'aden | 1211.SR | No | 72.15 | 5.59 | 77.74 | 10.29 | 4.30 | 41.7% | 18.08x | 20.59x | 35.20x | 27.04x |
| SABIC | 2010.SR | No | 40.08 | (0.96) | 39.12 | 31.07 | 4.77 | 15.3% | 8.20x | 11.84x | n.m. | 37.93x |
| **Aluminium Bahrain** | **ALBH** | **Yes** | **3.31** | **0.56** | **3.87** | **4.73** | **1.08** | **22.8%** | **3.59x** | **3.59x** | **3.94x** | **3.82x** |

*Source: Al Ramz Investment Research, Company financials, Bloomberg. Market data at 27 to 28 August 2026. EGA market capitalisation is the midpoint of the reported US$10bn to US$15bn ADX listing valuation and is an indication, not a traded price.*

### 4.3 Statistical summary

| Statistic | EV/EBITDA FY | EV/EBITDA LTM | P/E trailing | P/E forward |
|---|---|---|---|---|
| **Maximum**, core smelters (7) | 20.00x | 15.45x | 25.59x | 26.12x |
| **75th percentile** | 9.13x | 7.99x | 17.40x | 9.18x |
| **MEDIAN, CORE SMELTERS** | **6.59x** | **4.59x** | **10.61x** | **8.55x** |
| **25th percentile** | 4.72x | 4.36x | 7.54x | 5.99x |
| **Minimum** | 2.91x | 3.92x | 7.18x | 3.39x |
| Median, all 12 listed peers | 7.44x | 8.01x | 14.78x | 9.31x |
| *Memo: Alcoa / South32 AliGroup, disclosed through-the-cycle multiple* | *6.80x* | | | |
| *Memo: EGA implied at the reported ADX listing valuation range* | *5.30x to 7.30x* | | | |
| *Memo: Aluminium Bahrain at the 26 August 2026 close* | *3.59x* | *3.59x* | *3.94x* | *3.82x* |

*Source: Al Ramz Investment Research, Company financials.*

We do not use the LTM column. LME cash was US$3,212.50/t on 27 August 2026 against a 2025 average of US$2,632/t, so every peer's trailing EBITDA is measured on a cyclical peak. The entire gap between a 6.59x FY median and a 4.59x LTM median is that timing effect.

The strongest single data point in this section is not a trading multiple at all. On 1 July 2026 South32 announced the sale of its bauxite, alumina and aluminium interests to Alcoa and stated verbatim that the implied enterprise value of up to US$5.6bn "implies a through-the-cycle EBITDA multiple of ~6.8x", footnoted to average consolidated underlying EBITDA of about US$0.8bn over FY21 to FY25. That is a completed arms-length negotiation on a directly comparable asset package, measured explicitly through the cycle. In our view it is better evidence than any broker multiple table, and our review of every primary aluminium transaction from 2015 to 2026 found it to be the **only** deal in the entire set that publishes an EBITDA multiple at all.

The same disclosure gives a second, independent reading. South32 discloses CY2025 segment EBITDA of roughly US$0.9bn for the assets sold, being Worsley and Boddington at US$0.6bn, Hillside at US$0.2bn and Brazil at US$0.1bn. Against the same enterprise value that is approximately **6.2x spot-year EBITDA** against 6.8x through the cycle. The two readings bracket our 6.60x reference multiple from either side, which is about as good a corroboration as this sector offers.

### 4.4 Applying the multiple

Alba cannot be valued on a peer multiple applied to FY2026E or FY2027E EBITDA. FY2026E is distorted by a 44% collapse in production and FY2027E by a metal price still carrying a disruption premium. Applying a full-cycle multiple to an above-mid-cycle EBITDA double counts the cycle, which is the standard trap in cyclical comparables. We therefore apply the multiple to mid-cycle earnings power: FY2030E consolidated EBITDA of US$1,264m.

We then apply an explicit, itemised discount rather than asserting one.

| | Adjustment |
|---|---|
| **Reference multiple**: core smelter median 6.59x and the Alcoa/South32 negotiated 6.80x | **6.60x** |
| Single-asset concentration: one site, struck by a missile on 28 March 2026 | (15%) |
| Strait of Hormuz dependence for inbound alumina and outbound metal | (10%) |
| Sovereign: Bahrain at B2 negative and B stable, downgraded by S&P and Fitch in the last year | (10%) |
| Liquidity: 10% free float, average daily traded value of about US$0.4m | (10%) |
| Offset: cost position, gas contracted at US$4.50/MMBtu, 74% value-added mix, single-site scale | +10% |
| Offset: FY2025 EBITDA margin of 22.8% vs Alcoa 15.5%, Norsk Hydro 13.9%, Century 16.8%, Chalco 16.3% | +5% |
| **Net discount** | **(30%)** |
| **APPLIED MULTIPLE** | **4.62x** |

| Applied to mid-cycle EBITDA of US$1,264m | Multiple | Implied EV (US$m) | Equity (US$m) | Fair value (BHD) |
|---|---|---|---|---|
| **Applied multiple** | **4.62x** | **5,838** | **3,959** | **1.051** |
| Low case, near the peer floor | 4.00x | 5,054 | 3,176 | 0.843 |
| High case, a 15% discount | 5.61x | 7,089 | 5,210 | 1.383 |
| Undiscounted peer median | 6.60x | 8,340 | 6,461 | 1.716 |

*Source: Al Ramz Investment Research, Company financials.*

The two offsets are worth dwelling on. Alba's FY2025 EBITDA margin of 22.8% is higher than every Western peer in the set: Alcoa 15.5%, Norsk Hydro 13.9%, Century Aluminum 16.8%, Chalco 16.3%. Alba out-earns them on margin despite buying all of its alumina, because its gas is contracted at US$4.50/MMBtu against a World Bank European benchmark of US$15.00/MMBtu for 2026, and because 74% of its shipments are value-added products. **The discount we apply is about risk and liquidity, not about asset quality.** We think that distinction is the single most important thing to understand about this equity.

### 4.5 Cross-checks

| Metric | Core smelter median | Alba at BHD 0.878 | Alba at our BHD 1.150 |
|---|---|---|---|
| EV / mid-cycle EBITDA | 6.59x | 3.85x | **4.82x** |
| P/E, trailing | 10.61x | 3.94x | **5.16x** |
| P/E, one year forward | 8.55x | 3.82x | **5.00x** |
| P / book | 2.08x | 0.575x | **0.73x** |
| Dividend yield | 3.30% | 6.16% | **5.2%** |
| EV per tonne of capacity | US$6,201/t | US$2,634/t | **US$3,167/t** |
| FY2025 EBITDA margin | 15.5% to 16.8% (Western) | 22.8% | 22.8% |

*Source: Al Ramz Investment Research, Company financials.*

At our fair value Alba would still trade at a 27% discount to the peer core-smelter median on normalised earnings, at half the peer P/E, and at roughly a third of the peer P/B.

---

## 5. ALUMINIUM DUNKERQUE: WHAT THE PRICE REQUIRES

This section exists because the acquisition is 42% of Alba's current market capitalisation and because nobody else has priced it against disclosed numbers.

### 5.1 The bridge confirms US$2.2bn is an enterprise value

Bpifrance is paying EUR 100m for 6% of the holding company, which implies 100% of the equity at EUR 1,667m. Add ADIF consolidated net debt of EUR 201m at 31 December 2025 and the total is EUR 1,868m, or **US$2,174m** at 1.1641. That is within 1.2% of the disclosed US$2,200m. The triangulation confirms the headline is an enterprise value, which is the basis every multiple below is computed on.

### 5.2 What the price implies

| Basis | EBITDA (EURm) | EBITDA (US$m) | US$/t on 300kt | Implied EV/EBITDA | vs Alba mid-cycle US$662/t |
|---|---|---|---|---|---|
| FY2025 outturn | 292 | 340 | 1,133 | **6.47x** | 1.71x |
| FY2024 outturn | 317 | 369 | 1,230 | **5.96x** | 1.86x |
| 4-year average | 220 | 257 | 856 | **8.57x** | 1.29x |
| FY2023 | 169 | 197 | 656 | 11.18x | 0.99x |
| FY2022 energy-crisis trough | 104 | 121 | 404 | 18.17x | 0.61x |

*Source: Al Ramz Investment Research, Pappers.fr / INPI statutory filings.*

**The value-neutrality threshold is US$385m a year, or US$1,283 per tonne**, solved so that the present value of Dunkerque's unlevered free cash flow equals the US$2.2bn paid, at our 13.27% WACC and 2.0% terminal growth.

| Earnings basis | Implied Dunkerque EV (US$m) | vs the US$2,200m paid |
|---|---|---|
| FY2024 outturn, US$369m (ARENH pricing) | 2,086 | (5%) |
| FY2025 outturn, US$340m (ARENH pricing) | 1,877 | (15%) |
| 4-year statutory average, US$257m | 1,279 | (42%) |
| Our central post-reset estimate, US$220m | 1,015 | (54%) |
| FY2022 energy-crisis trough, US$121m | 304 | (86%) |

### 5.3 The power reset

Aluminium Dunkerque draws 450 MW continuously, roughly 3.94 TWh a year, or 13.1 MWh per tonne on 300kt, which is consistent with smelter norms. ARENH at EUR 42/MWh expired on 31 December 2025.

| New EDF price (EUR/MWh) | Delta vs ARENH | Annual cost delta (EURm) | FY2025 EBITDA rebased (EURm) | US$m | US$/t |
|---|---|---|---|---|---|
| 42 (ARENH, expired) | 0 | 0 | 292 | 340 | 1,133 |
| 55 | 13 | 51 | 241 | 280 | 934 |
| **65** | **23** | **91** | **201** | **234** | **781** |
| **75** | **33** | **130** | **162** | **188** | **628** |
| 85 | 43 | 170 | 122 | 143 | 475 |
| 95 | 53 | 209 | 83 | 97 | 322 |

*Source: Al Ramz Investment Research. Upper bound on the impact: the EDF contract covers "une part substantielle des besoins", not the whole load, and the undisclosed upfront advance to EDF lowers the ongoing unit price.*

**Our conclusion.** At a EUR 65 to 75/MWh reset on the full load, Dunkerque's rebased EBITDA of US$188m to US$234m brackets the US$212m the Task 2 model already assumes for FY2027E. The Task 2 estimate, which was built with no disclosure at all, turns out to have been well calibrated. We adopt **US$220m as the steady-state case**, which values Dunkerque at roughly US$1.0bn against US$2.2bn paid.

We should be plain about what that means and about how confident we are. On the FY2024 run rate the price was value neutral. On any plausible post-ARENH run rate it is not. Alba is paying a full price for geographic diversification, EU-domestic non-CBAM-liable production and a nuclear-powered low-carbon product, at a moment when it has just learned what single-site concentration costs. That is a defensible strategic decision and an expensive financial one, and the single number that would settle it, the EDF contract price, has not been disclosed by anyone.

We would also note the internal comparison. Alba is paying **US$7,333 per annual tonne** for Dunkerque while its own shares change hands at **US$2,634 per annual tonne**. EFG Hermes made the same point independently, calling it "nearly double Alba's own average historic per-tonne enterprise value".

---

## 6. SCENARIO ANALYSIS

The Task 2 model carries three fully specified scenarios. We discount each at the same 13.27% WACC and apply a 5.0x exit multiple to FY2030E EBITDA rather than a perpetuity, because a perpetuity on a trough NOPAT is not an equilibrium: no smelting industry runs below the incentive price forever, since supply exits.

| | **Bull** | **Base** | **Bear** |
|---|---|---|---|
| Narrative | "The disruption holds" | "Orderly normalisation" | "Hormuz reopens, volumes do not" |
| FY2027E LME (US$/t) | 3,500 | 3,050 | 2,700 |
| FY2030E LME (US$/t) | 3,100 | 2,750 | 2,500 |
| FY2027E Alba production (kt) | 1,420 | 1,300 | 1,150 |
| FY2030E Alba production (kt) | 1,620 | 1,620 | 1,550 |
| FY2027E EBITDA (US$m) | 2,579 | 1,619 | 579 |
| FY2030E EBITDA (US$m) | 2,024 | 1,264 | 648 |
| FY2030E net debt / EBITDA | (1.39x) | (0.18x) | 2.66x |
| Implied EV per tonne | US$5,995/t | US$3,338/t | US$1,255/t |
| **Fair value (BHD)** | **2.602** | **1.195** | **0.145** |
| **Probability** | **20%** | **55%** | **25%** |

**Probability-weighted fair value: BHD 1.214 (+38.2%).**

*Source: Al Ramz Investment Research, Company financials.*

Three observations.

**The bear case is genuinely severe and we have not softened it.** At BHD 0.145 the bear implies an enterprise value of US$1,255 per annual tonne, a deep distress mark below the Rio Tinto disposal of Dunkerque in 2018. It gets there because the bear runs LME at US$2,500/t in perpetuity, which is below the industry's incentive price, with net debt at 2.66x EBITDA after the Dunkerque drawdown. We would not underwrite that as an equilibrium, but a 25% weight on it is the discipline the current price deserves.

**The probability weighting is where reasonable people will disagree.** At 15/55/30 the weighted value falls to BHD 1.091 and at 25/55/20 it rises to BHD 1.336. We chose 20/55/25 because the evidence points both ways with unusual force. Against the bull: Goldman Sachs cut its 2027 forecast to US$2,700/t on 6 July 2026 while raising its 2027 surplus estimate to roughly 1.5 Mt; Morgan Stanley cut 2027 by 13% to US$2,850/t; and J.P. Morgan's published path falls US$1,050/t from Q3 2026 to Q4 2027. Against the bear: the Strait of Hormuz was still effectively closed on 27 August 2026, on day 179, with three transits recorded on 23 August against a pre-crisis baseline of 85 a day; LME inventories are at 246,825 tonnes, down 51.5% year to date; and the market's own forward curve prices out only about US$46/t by the long-dated tenor.

**That last gap is the crux.** The curve and the sell-side consensus are roughly US$1,000/t apart on where aluminium goes. A valuation anchored on the curve and one anchored on J.P. Morgan would differ by about a third of Alba's revenue line. We sit between them, closer to the sell side.

We should also correct a figure carried into the Task 2 bear-case narrative. The claim that roughly 700,000 tonnes of held metal would return on a Hormuz reopening does not survive verification: the string appears only in an AL Circle headline of 18 June 2026 and nowhere in the article body, with no attribution or derivation, and a separate AL Circle item uses a similar figure for the entirely unrelated year-to-date fall in China's net aluminium exports. **The better-supported number is larger, not smaller.** Axis Securities models GCC production recovering from 3.40 Mt in 2026 to 5.50 Mt in 2027 and 6.40 Mt in 2028, which is +2.10 Mt returning in 2027 alone. That makes the bear case more credible, not less, and is one reason we hold the bear weight at 25%.

---

## 7. HISTORICAL MULTIPLES

Alba's own trading history is a legitimate valuation anchor and, for a stock with four analysts and 10% free float, arguably a better one than a global peer median.

| Period | Price (BHD) | EPS (fils) | BVPS (BHD) | DPS (fils) | **Trailing P/E** | **P/B** | **Dividend yield** |
|---|---|---|---|---|---|---|---|
| 31 Dec 2021 | 0.800 | 319 | 1.0614 | 79.71 | 2.51x | 0.754x | 9.96% |
| 31 Dec 2022 | 1.090 | 294 | 1.2871 | 117.58 | 3.71x | 0.847x | 10.79% |
| 31 Dec 2023 | 1.145 | 83 | 1.2636 | 29.18 | 13.80x | 0.906x | 2.55% |
| 31 Dec 2024 | 1.300 | 130 | 1.3587 | 45.17 | 10.00x | 0.957x | 3.47% |
| 31 Dec 2025 | 1.105 | 154 | 1.4729 | 54.06 | 7.18x | 0.750x | 4.89% |
| **Mean 2021-2025** | | | | | **7.44x** | **0.843x** | **6.33%** |
| **Median 2021-2025** | | | | | **7.18x** | **0.847x** | **4.89%** |
| **26 Aug 2026** | **0.878** | **223 (TTM)** | **1.5279** | **54.06** | **3.94x** | **0.575x** | **6.16%** |

*Source: Al Ramz Investment Research, Company financials. Derived from Alba's own reported profit, equity and declared dividends against year-end closes. The 2026 row independently reproduces Investing.com's published P/E of 3.94 and TradingView's published dividend yield of 6.16% to the second decimal.*

Two things stand out.

**The P/E series is a textbook cyclical inversion and should not be averaged naively.** Alba traded at 2.51x in 2021 and 3.71x in 2022 on peak earnings, and at 13.80x in 2023 on trough earnings. The market has consistently refused to capitalise peak aluminium earnings at more than about 4x. Today's 3.94x is not an anomaly; it is what this market always does with a peak-earnings year. **That is the single strongest piece of evidence for the bears, and it is the main reason our fair value sits well below consensus.**

**The P/B series is far more stable and far more informative.** It ranges from 0.750x to 0.957x across 2021 to 2025, with a mean of 0.843x. Alba has traded below book in every year of the window. Today's 0.575x is 32% below the five-year mean and is the lowest reading in the series.

| Application | Value (BHD) |
|---|---|
| P/B at the five-year mean 0.843x on FY2026E book value of BHD 1.5765 | **1.329** |
| P/B at the five-year low 0.750x | 1.183 |
| P/E at the five-year median 7.18x on FY2030E mid-cycle EPS of 147.8 fils | **1.061** |
| Dividend yield at the five-year median 4.89% on FY2030E DPS of 60.1 fils | **1.228** |
| **Average of the three** | **1.206** |

*Source: Al Ramz Investment Research.*

As a cross-check we ran the justified P/B, (ROE - g) / (Ke - g), at our 16.05% cost of equity. At the five-year average ROE of 15.9% it gives 0.989x and BHD 1.559. At the five-year median ROE of 10.5% it gives 0.602x and BHD 0.949. At the FY2027E ROE of 13.8% it gives 0.840x and BHD 1.324. Our BHD 1.206 sits inside that band and close to its centre.

---

## 8. PRECEDENT TRANSACTIONS AND ASSET VALUE

### 8.1 The transaction set is thin

We reviewed every primary aluminium and alumina transaction we could locate from 2015 to 2026. Across that period there are only **three** completed deals with both a disclosed price and an unambiguous smelter capacity denominator on an enterprise value basis.

| Announced | Target | Acquirer | Consideration | Capacity | **EV or equity per tonne** | Implied EV/EBITDA | LME at announcement |
|---|---|---|---|---|---|---|---|
| Jun 2026 | Aluminium Dunkerque (100%) | Alba, Bpifrance 6% | US$2,200m | 300kt | **US$7,333/t** | 6.47x FY2025 | US$3,855/t |
| Jun 2026 | South32 aluminium value chain | Alcoa | US$4.1bn upfront, up to US$5.6bn | bundled | not meaningful | **6.8x through-cycle, 6.2x CY2025** | US$3,439/t |
| Jan 2026 | Companhia Brasileira de Aluminio (68.6%) | Chalco / Rio Tinto | US$903m | ~400kt | US$3,290/t (equity, bundled) | not disclosed | US$3,142/t |
| Jul 2025 | Ma'aden Aluminium and MBAC (25.1%) | Ma'aden, from Alcoa | US$1,350m | 202kt attributable | US$6,690/t (equity, bundled) | not disclosed | US$2,450/t |
| Feb 2025 | Alba (20.62% block) | Ma'aden, from SABIC | US$966m | 1,623kt (100% basis) | **US$3,557/t** | 6.15x FY2024 | US$2,660/t |
| Oct 2021 | Aluminium Dunkerque (control) | American Industrial Partners | never disclosed | 300kt | n/a | not disclosed | US$2,934/t |
| Feb 2018 | Rio Tinto ISAL, Iceland | Norsk Hydro (aborted) | US$345m | 205kt | US$1,683/t | not disclosed | US$2,182/t |
| Jan 2018 | Dunkerque smelter | Liberty House (GFG) | US$500m | 282kt | **US$1,773/t** | not disclosed | US$2,210/t |
| Nov 2016 | Rio Tinto Lochaber | SIMEC / Liberty House | US$410m | 47kt | US$8,723/t (mostly hydro and land) | not disclosed | US$1,737/t |
| 2016 | New Madrid, Chapter 11 | Magnitude 7 Metals | US$14m | 263kt | US$52/t (distressed) | not disclosed | US$1,604/t |
| **Statistical summary, the three clean EV-basis deals** | | | | **Median** | **US$3,557/t** | | |
| | | | | 75th percentile | US$5,445/t | | |
| | | | | 25th percentile | US$2,665/t | | |
| | | | | Mean | US$4,221/t | | |

*Source: Al Ramz Investment Research, company announcements, SEC EDGAR, HKEXnews, Westmetall LME official series for the Dunkerque and Alba marks and the World Bank Pink Sheet monthly average elsewhere.*

Three features of this set matter.

**The dispersion is about the cycle, not the assets.** Between the minimum and maximum the per-tonne value moves 4.1x while the metal price moves only 2.0x. Two of the three clean observations are the same asset: Aluminium Dunkerque has changed hands four times since 2018 and the price was disclosed only twice.

**Almost nothing in this sector trades on a disclosed multiple.** The Alcoa and South32 transaction of 30 June 2026 is the only deal in the entire eleven-year set that publishes an EBITDA multiple for primary aluminium and alumina assets. That is why we lean on it as hard as we do in section 4.

**Distressed smelters clear at or near zero.** Portovesme, Aldel twice, Aluminij Mostar, Ormet and New Madrid all transacted with either no disclosed consideration or a nominal one, each accompanied by a state or sponsor commitment rather than a price, and Aldel and Ormet subsequently failed outright. That is the tail risk in any smelter without a locked-in power contract, and it is precisely what distinguishes Aluminium Dunkerque, which has a ten-year EDF supply agreement, from the failures. It is also the reason we do not treat replacement cost as a hard floor.

**Per-tonne comparables are a range check here, not a primary method**, which is why we weight this section at 10%.

### 8.2 The Ma'aden mark is the most useful external evidence on Alba

On 17 February 2025 Ma'aden completed the purchase of SABIC's entire 20.62% stake, 292.8m shares, for BHD 363.08m.

| | |
|---|---|
| Implied price per share | **BHD 1.240** |
| Implied 100% equity value | BHD 1,756m / US$4,670m |
| Implied enterprise value (FY2024 net debt of US$1,103m) | **US$5,773m** |
| Implied EV / FY2024 EBITDA of US$939m | **6.15x** |
| Implied EV per tonne of Alba capacity | **US$3,559/t** |
| Implied P/E on FY2024 EPS of 130 fils | 9.54x |
| Implied P/B on FY2024 book value of BHD 1.3587 | 0.91x |
| **Premium to the 26 August 2026 close** | **+41.2%** |

*Source: Al Ramz Investment Research, Argaam, Bahrain Bourse major shareholder register.*

This is the single most useful external mark on Alba equity. It was struck by a strategic buyer with full diligence access, and Ma'aden subsequently recognised a **one-off bargain purchase gain of US$0.2bn in Q4 2025**, meaning the buyer's own assessment of the fair value of Alba's net assets exceeded the price it paid.

Two qualifications keep us from leaning on it harder. It was a non-controlling 20.62% block with no control premium, and it was struck eighteen months ago at an LME of US$2,660/t, before the March 2026 disruption changed both the earnings and the risk profile.

Applying the transaction median of US$3,559/t to Alba's 1,923kt of consolidated capacity gives an enterprise value of US$6,844m and a **fair value of BHD 1.318**.

### 8.3 Replacement cost

| Benchmark | US$/t | Implied Alba fair value (BHD) |
|---|---|---|
| CRU, Indonesian greenfield benchmark | 2,800 | 0.931 |
| Aluminum Association, best case | 4,500 | 1.799 |
| **Alba Line 6, whole scope, US$3.0bn for 540kt** | **5,556** | **2.338** |
| CRU, world ex-China upper bound | 6,000 | 2.565 |
| Aluminum Association / AGBI, conservative US greenfield | 8,000 | 3.586 |

*Source: Al Ramz Investment Research, Aluminum Association "Powering Up American Aluminum" (May 2025), CRU, Alba Bahrain Bourse announcement of 29 May 2016.*

The industry body's own framing is worth quoting directly: "In a best-case scenario, analysts believe **US$4,000 to 5,000 per metric ton** of installed primary aluminum capacity is achievable. Recent projects in Russia, Indonesia and Iran have come in at **US$5,000 to 6,000 per ton**. The most conservative estimates are **US$8,000 or higher**." CRU separately puts Chinese domestic capital intensity slightly under US$1,400 per annual tonne, Indonesia at around US$2,800 and much of the world ex-China at up to US$6,000. The spread from US$1,400 to US$8,000 is a real economic range driven by power cost, labour and permitting rather than a data error. Alba sits nearer the low end on a Bahrain cost basis and Dunkerque nearer the high end on an EU basis, which is itself part of the argument for why Alba paid what it did.

One correction to a figure carried in earlier stages. Alba's US$3.5bn Line 6 number was the **original bankable feasibility estimate for the same scope**, later cut to approximately US$3.0bn. It is not US$3.0bn plus Power Station 5. Both figures include the power station (Alba Bahrain Bourse announcement, 29 May 2016).

**At the 26 August 2026 close Alba trades at US$2,634 per annual tonne, which is 47% of the whole-scope cost of its own Line 6 and 36% of what it has just agreed to pay for Aluminium Dunkerque.** A buyer of the shares is acquiring existing capacity at roughly a third of the price the company is paying for new capacity.

That gap is not automatically a valuation argument. Replacement cost caps what an asset is worth only when returns justify rebuilding, and a smelter earning below its cost of capital is worth less than its steel. It does put a floor under the downside case, and it explains why a strategic buyer paid a 41% premium to today's price for a fifth of the company.

---

## 9. VALUATION SUMMARY AND FOOTBALL FIELD

### 9.1 Blended fair value

| Method | Low | **Base** | High | Weight | Contribution | Upside at base |
|---|---|---|---|---|---|---|
| Discounted cash flow, Task 2 model as built | 0.884 | **1.056** | 1.397 | 30% | 0.317 | +20.3% |
| Discounted cash flow, Dunkerque at its statutory accounts | 1.040 | **1.229** | 1.458 | 15% | 0.184 | +40.0% |
| Scenario probability-weighted DCF | 1.010 | **1.214** | 1.407 | 15% | 0.182 | +38.3% |
| Trading comparables on mid-cycle EBITDA | 0.843 | **1.051** | 1.383 | 20% | 0.210 | +19.7% |
| Alba's own historical multiples | 1.061 | **1.206** | 1.335 | 10% | 0.121 | +37.4% |
| Precedent transactions and asset value | 0.931 | **1.318** | 1.799 | 10% | 0.132 | +50.1% |
| **BLENDED FAIR VALUE** | **0.941** | **1.146** | **1.439** | **100%** | **1.146** | **+30.5%** |
| **ROUNDED FAIR VALUE** | | **1.150** | | | | **+31.0%** |

*Source: Al Ramz Investment Research, Company financials.*

**Weighting rationale.** The DCF carries the highest combined weight at 45% across its two variants because it is the only method that prices the four-year volume recovery rather than assuming it away, and because Alba's near-term earnings are too distorted for any multiple to be applied honestly. Trading comparables take 20%: the peer set is real but only four of thirteen names are structurally comparable and none shares Alba's single-site or Hormuz exposure. The scenario DCF takes 15% because the outcome here is genuinely bimodal and a point estimate understates that. Historical multiples and precedent transactions take 10% each: both are informative and both rest on samples too thin or too cycle-dependent to bear more weight.

### 9.2 Football field

Fair value per share, BHD. Last close BHD 0.878; blended fair value BHD 1.150.

| Method | Range |
|---|---|
| Discounted cash flow | 0.884 to 1.397 |
| DCF, Dunkerque at statutory accounts | 1.040 to 1.458 |
| Scenario weighted DCF | 1.010 to 1.407 |
| Trading comparables | 0.843 to 1.383 |
| Historical own multiples | 1.061 to 1.335 |
| Precedent transactions and asset value | 0.931 to 1.799 |
| Sell-side consensus target range | 1.300 to 1.750 |
| 52-week trading range | 0.715 to 1.150 |

*Source: Al Ramz Investment Research, Bloomberg. The 52-week low is shown as the Investing.com and TradingView figure of 0.715; the Yahoo daily series over the same window gives 0.755. Both readings are preserved.*

Note that the entire sell-side consensus range sits above the top of our DCF base case and above five of our six method midpoints. We are not disputing that Alba is cheap. We are disputing how cheap.

### 9.3 Implied valuation at the fair value

| At BHD 1.150 | |
|---|---|
| Market capitalisation | US$4,331m |
| Enterprise value | US$6,090m |
| EV / FY2030E mid-cycle EBITDA | 4.82x |
| EV per tonne of annual capacity | US$3,167/t |
| EV as % of the Line 6 whole-scope build cost | 57% |
| P/E on FY2027E earnings | 5.00x |
| P/E on FY2030E mid-cycle earnings | 7.78x |
| P / FY2026E book value | 0.73x |
| FY2030E dividend yield at the stated 35% payout | 5.2% |
| Discount to the Ma'aden transaction price of BHD 1.240 | (7.3%) |
| Discount to the consensus mean target of BHD 1.450 | (20.7%) |

*Source: Al Ramz Investment Research, Company financials.*

Our fair value still leaves Alba below its own historical median P/E of 7.18x on mid-cycle earnings only marginally above it, below its own historical mean P/B of 0.843x, below the Ma'aden transaction price, and at a 27% discount to the peer core-smelter EV/EBITDA median. We do not think we are being generous.

---

## 10. INVESTMENT RECOMMENDATION

**OVER-WEIGHT. Fair value BHD 1.150. Upside +31.0% from the BHD 0.878 close of 26 August 2026.**

The house rule places Over-Weight above +15% upside and Under-Weight below -15%. At the last close that means a fair value above BHD 1.010 or below BHD 0.746. Our BHD 1.150 clears the Over-Weight threshold by 14%.

Our thesis in one paragraph. Alba is a first-or-second-quartile cost producer with contracted gas at US$4.50/MMBtu, a 74% value-added mix and the highest EBITDA margin of any Western peer, trading at 0.575x book, 3.9x trailing earnings and 47% of the replacement cost of its own newest potline. The market is discounting a perpetual mid-cycle aluminium price of about US$2,644/t, below the bottom of the band the company's own consultant publishes and below the level at which new capacity gets built. It is doing so because the 2026 earnings are unmistakably a war premium earned at 38% utilisation, because the sovereign has been downgraded twice in a year, and because 10% free float and US$0.4m of daily turnover make the stock uninvestable for most institutions. Those are real reasons for a discount. They are not, in our view, reasons for a 42% discount to global peers on normalised earnings when a strategic buyer paid a 41% premium to today's price eighteen months ago and booked a bargain purchase gain on the trade.

**What would change our mind.** The rating fails if the WACC is above 13.68%; if mid-cycle LME settles below roughly US$2,650/t with volumes restored to 1,620kt; or if Aluminium Dunkerque completes and earns less than roughly US$100m of EBITDA a year. Each is plausible. None is our base case.

### 10.1 Key catalysts

**1. Disclosure of the Aluminium Dunkerque EDF contract price, or of Dunkerque earnings on completion (Q4 2026).** This is the highest-value catalyst because it is the largest unpriced item. Our sensitivity puts roughly BHD 0.20 of fair value on every US$100m of steady-state Dunkerque EBITDA. Completion requires French foreign direct investment clearance and EU Foreign Subsidies Regulation approval; EU antitrust clearance was obtained in May 2026.

**2. A restart schedule for Reduction Lines 1 to 3.** Alba has given no guidance in either its Q1 or Q2 2026 release. EGA has guided its own Al Taweelah recovery to pre-war capacity by Q1 2027 and was at 25% of reduction cells restarted on 26 August 2026. Alba's Q2 utilisation of roughly 38% is constrained by inbound alumina logistics as much as by the announced curtailment, so this catalyst is gated on Hormuz rather than on an Alba decision.

**3. Normalisation of the Strait of Hormuz.** Oman's and Iran's foreign ministers met on 25 August 2026 to discuss a phased framework for a temporary joint navigation corridor alongside a joint mine-clearing project. This is genuinely two-sided: it restores Alba's volumes and removes the price premium at the same time. Our scenarios put the net effect at negative for the bear and positive for the base.

**4. A New Replacement Line decision.** We ascribe zero value to Line 7. Alba has disclosed neither its capacity, where the 380kt and 540kt readings conflict, nor its capital cost, and no board decision is public. Independent sector work already assumes a 540kt Alba potline 7 in GCC supply from 2028. Any credit is upside to our number.

**5. An increase in free float or a Saudi cross-listing.** Asked directly on 18 February 2026, CEO Ali Al Baqali said "this is a shareholders' matter, actually. But what we are doing, it's already in our radar." A meaningful float increase would attack the liquidity discount we carry at 10% of the peer multiple, worth roughly BHD 0.11 a share on our comparables method alone.

**6. Insurance recoveries.** The December 2025 rectifier fire was confirmed as fully insured with recoveries expected during 2026. Nothing has been disclosed on the 28 March 2026 attack. We model no recovery for the attack.

### 10.2 Key risks to the valuation

**1. Mid-cycle aluminium price (high probability of dispersion, up to -45% impact).** The 2027 sell-side range runs from Goldman Sachs at US$2,700/t to CRU above US$4,000/t, a spread of 52% of the low, with no basis in the public record for preferring one. At US$2,550/t mid-cycle and 1,620kt our fair value is BHD 0.665.

**2. Slow volume recovery (medium probability, -10% impact).** Q2 2026 output of 155,469 tonnes annualises to 38% of nameplate against an announced curtailment of only 19% of capacity. The residual is alumina availability. If Alba reaches only 1,450kt by 2030 rather than 1,620kt, fair value falls roughly BHD 0.09.

**3. Aluminium Dunkerque (medium probability, -6% to +15% impact).** Alba is paying US$7,333 per annual tonne, nearly three times its own market EV per tonne, for an asset whose FY2024 and FY2025 margins were earned under an expired power regime. The deal also takes FY2026E net debt to US$1,759m and 1.56x EBITDA from 0.52x at end-2025.

**4. Cost of capital (medium probability, -13% impact).** Bahrain was downgraded to B by S&P in November 2025 and to B by Fitch in February 2026, with Moody's outlook moved to negative in April 2026. A further downgrade would widen the sovereign yield we use as the risk-free rate. At a 14.27% WACC our fair value falls to BHD 0.948, which is Equal-Weight.

**5. Liquidity and governance (structural, not event-driven).** A 10% free float, 69.38% held by Mumtalakat and 20.62% by Ma'aden, with average daily traded value of roughly US$0.4m, means a position of any institutional size cannot be exited quickly. This is a permanent feature of the equity and is why we carry an explicit liquidity discount rather than treating it as a risk footnote.

**6. Renewed conflict (low probability, high impact).** The Askar facility was struck once. It can be struck again. We carry a 1.50% single-asset concentration and liquidity premium in the cost of equity for this, which is a judgement and not a measurement.

**7. Taxation (low probability, -3% impact).** Bahrain's Domestic Minimum Top-up Tax under Decree Law 11 of 2024 is modelled at 1.5% in FY2026E, 5% in FY2027E and 15% from FY2028E. Management stated on 18 February 2026 that it foresees no impact in 2026. An earlier or wider application would reduce our number.

---

## 11. DATA CONFLICTS AND LIMITATIONS

Per the contract's requirement to surface conflicts rather than smooth them, the following bear directly on this valuation.

**Resolved from Task 1.** Conflict C9, the ALBH quote date, is resolved. BHD 0.878 is the close on 26 August 2026, confirmed five ways: the final Yahoo daily bar; the Investing.com JSON payload reading `"last":0.878, "lastUpdate":"2026-08-27T00:00:00.000Z"`; companiesmarketcap's US$3.30bn market capitalisation; TradingView's P/E of 3.932 times EPS of 0.2233; and a published consensus upside of 65.15% to a BHD 1.450 target that arithmetically requires a 0.878 base. Decisively, the trailing EPS of 223 fils cannot have existed before the H1 2026 results of 4 August 2026.

**Unresolved and material.**

1. **The CRU mid-cycle band of US$2,650 to US$2,750/t could not be corroborated.** It is disclosed by Alba but no public CRU document stating it was located; CRU's own reachable 2026 statement points the other way, above US$4,000/t through Q2 2027. Our terminal LME of US$2,750/t rests on a company-sourced figure. It is, however, supported from a different direction by the US$2,800 to US$3,000/t incentive-price floor in independent sector work and by the CY2030 threshold of US$2,942/t negotiated into the Alcoa/South32 contingent consideration, both of which sit above it.

2. **Peer betas differ by roughly a factor of two depending on source.** Vendor betas for non-US listings give a peer-median unlevered beta of 0.584; five-year weekly regressions against home indices give 1.088. Relevered, that is a 0.75x to 1.40x range and roughly BHD 0.37 of fair value. We take 1.35x, the conservative end.

3. **The 2026 market balance ranges from a 592kt surplus to a 1.7 Mt deficit**, a 2.3 Mt spread on a 74 Mt market. Alba's own Q1 2026 release stated a surplus including China and its Q2 release stated a 934kt deficit, a reversal of 1.53 Mt in one quarter without comment. Goldman Sachs at -0.10 Mt and J.P. Morgan at -1.7 Mt differ by a factor of 17.

4. **Aluminium Dunkerque's accounts are French GAAP statutory, not IFRS.** EBE is not directly comparable to Alba's own EBITDA definition, which is profit before tax plus finance cost plus depreciation and amortisation plus change in fair value of derivatives less the realised gain on interest rate swap settlement. The underlying *comptes consolidés* PDFs are paywalled; our figures are the registry's extraction of those filings.

5. **The Line 6 capital cost per tonne carries an unresolved internal conflict.** US$3.0bn for 540kt is US$5,556/t on the whole scope including Power Station 5, but Alba's own Annual Report 2019 claims "less than US$4,000 per metric tonne". The only coherent reading is that the sub-US$4,000/t figure excludes the power station. That split is inferred and was never disclosed, so we use both readings and do not average them.

6. **Forward EV/EBITDA is not available for any peer.** Consensus EBITDA estimates for FY2026 and FY2027 sit behind paywalls on every service reached, so forward P/E is the only consistent forward multiple in the comparables section. This is recorded as a gap rather than filled with an estimate.

7. **The 52-week low is either BHD 0.715 or BHD 0.755**, and three-month average daily volume is either 162,260 or 190,819 shares, depending on vendor. Neither affects the valuation.

**Deliberately excluded from the fair value.** No value for the New Replacement Line. No insurance recovery on the 28 March 2026 attack. No credit for a Saudi cross-listing or free-float increase. No control premium. Each is a real call option on the equity and each is upside to our number.

---

## 12. HANDOFF TO TASK 4

The following are now available in `ALBH_Financial_Model_2026-08-28.xlsx` for the chart slots the report contract requires.

| Contract requirement | Location |
|---|---|
| `dcf_sensitivity_heatmap` (chart 28, mandatory) | Sensitivity tab, Tables 1 to 4, with conditional colour scales applied |
| `comps_with_stats_summary` (charts 30, 31) | Comps tab, sections A and B, including max / 75th / median / 25th / min |
| `historical_multiples` line with average (chart 34) | Section 7 of this document; series 2021 to 2026 for P/E, P/B and dividend yield |
| `blended_fair_value` table and football field (chart 32, mandatory) | Valuation Summary tab, sections A and C |
| DCF components (chart 29) | DCF tab, sections 2 to 4 |
| Scenario comparison (charts 13, 14) | Scenarios tab and section 6 of this document |
| Share price 12 months (chart 01) | `Sources/W1_ALBH_daily_12m_2025-2026_YahooFinance.csv` |

**New tabs added to the Task 2 workbook:** DCF, Sensitivity, Comps, Precedents, Valuation Summary. All five are live-formula tabs linked to the Task 2 statements, so changing any Task 2 driver reflows the entire valuation and the rating, which is computed rather than asserted.

**Source cache.** All primary documents obtained at Task 3 are in `dreams-associate/reports/ALBH/Sources`, with per-workstream notes in `SOURCE_NOTES.md`. Task 4 should not re-fetch any of them.

---

*Al Ramz Investment Research. This document is the Task 3 deliverable of the initiating-coverage workflow and is an input to the final report, not a published research note. Numbers bind to the model of record at `ALBH_Financial_Model_2026-08-28.xlsx` and to the sources cached at `reports/ALBH/Sources`.*

Cancel

Deliverables

ALBH_Valuation_Analysis_2026-08-28.md

Open

Gate rule: the next stage unlocks only when you accept this one. Every acceptance, edit, and rework note is written to the audit trail.