Actionable insights — The Company Keeping American Aluminum Home

1. Pick the rung, not the commodity — value-chain position decides who keeps the premium

1. Draw the chain explicitly before choosing a name: miner → refiner → smelter/processor → fabricator → end product. Write one sentence per stage saying what it buys and what it sells.
2. Locate where the price distortion you are betting on is applied. A tariff, premium or subsidy attaches at one specific point in the chain — usually at import of a particular form of the material.
3. The stage that sells into the distorted price without paying it captures the whole distortion. The stage that buys at the distorted price has a cost problem it must try to pass on. Everyone upstream of the distortion is unaffected; everyone downstream is squeezed.
4. Prefer the pure play at that rung. A diversified miner that happens to produce some of the material dilutes the exposure; ask what fraction of revenue is actually the distorted product.
5. Re-run the exercise whenever the distortion moves. The right rung is not a permanent property of a commodity — it changes when the policy changes, which means rotating within a commodity is a legitimate trade, not indecision.

• Which form of the material the tariff or quota names (ore vs concentrate vs refined metal vs semi-finished) — that word decides the rung; the share of a company's revenue that is the distorted product; a publisher or an insider rotating from one rung to another inside a commodity they still like.

2. Do the premium-vs-price margin arithmetic before touching a valuation model

1. Split the realised selling price into its two components: the global benchmark (which the producer does not control and which every competitor also receives) and the local premium (which only some producers receive).
2. Test whether the producer's largest cost input moves with the benchmark. If it does not — energy, labour, long-dated contracts — then an increase in the premium is close to pure margin, and operating leverage is far higher than a revenue-multiple comparison implies.
3. Express the premium as a percentage of the benchmark. A surcharge worth 60% of the world price is not a rounding item; it is the investment case, and it should be modelled as the primary variable rather than as a spread assumption.
4. Compare that premium to its own history, not to the metal price's history. A premium at three times its two-year-ago level, driven by a policy that is still in force, is the thing to underwrite.
5. Then, and only then, ask whether the equity price reflects the premium at all — the mismatch between a premium at a record and a stock well off its highs is where the opportunity lives.

• Named regional premia (Midwest, European duty-paid, Shanghai) quoted separately from the benchmark; whether the producer's power is contracted, captive or spot; EBITDA growing several times faster than volume as the operating-leverage tell; premium series at records while the equity is not.

3. Buy a policy-created premium after a sentiment-driven pullback — and check the operating business moved the other way

1. When a policy-driven stock sells off, classify the cause: a headline about the policy, or a deterioration in the business. These require opposite responses.
2. Test it against the physical evidence: capacity utilisation, realised prices, cash flow, leverage. If those improved over the same period the stock fell, the drawdown is a re-rating of policy risk, not a downgrade of the asset.
3. Check whether the underlying premium also fell, and by how much. If the premium gave back a fraction of what the equity did, the equity is discounting a worse outcome than the physical market is.
4. State the discount as an implied scenario — "at this price the market assumes X" — and then ask whether X is plausible. That converts a valuation judgement into a policy judgement you can actually research.
5. Set a buy-up-to limit rather than a target, so the entry discipline survives a rebound and the position is only added on the terms the analysis supports.

• A single-session drop on a policy news story with no change to guidance; the underlying spread or premium falling far less than the equity; record operating metrics reported during the drawdown; management debt repayment and a crossover to net cash while the stock derates.

4. Stress-test a policy thesis at the reduced setting, not at zero

1. Identify the single policy variable the thesis depends on and name the realistic downside setting — usually a negotiated reduction, not abolition. Most policy risk resolves as a partial climb-down.
2. Model earnings at that reduced setting and compare them to the pre-policy baseline, not to today. The question is whether a halved intervention still leaves economics far better than the world before it.
3. Separate the cyclical justification for the policy from the structural one. A tariff levied for negotiating leverage can vanish; one attached to a critical-minerals designation and a defense-industrial dependency has a floor under it.
4. Then handicap the outcomes honestly and say which you expect and why — including the case where a modest, settled tariff is better for the equity than an unresolved negotiation, because it removes uncertainty.
5. Accept the residual: the position will trade on headlines regardless. Size for that, and treat headline volatility as the cost of owning the premium rather than as new information.

• Reported negotiating ranges (the "50 to 25" style number) as the real downside case; pre-policy margin history as the true floor; critical-mineral / national-security designations that outlive a trade dispute; whether resolution itself — at any level — is a positive catalyst for the equity.

5. Separate the cash-flowing core from the option bolted onto it — and price them differently

1. Split the company into operating assets producing cash today and development projects producing nothing yet. Underwrite the first on current economics; treat the second as an option.
2. For the option, list the gates that must clear before it is real: litigation, permits, a power-supply agreement, engineering, a final investment decision. Each unresolved gate is a reason it is not in the base case.
3. Ask the disqualifying question: if the project dies, does the existing business suffer? If the answer is no, the option is free upside; if the answer is yes (funding commitments, cross-collateral, capex obligations), it is a liability wearing a growth costume.
4. Note who else is paying. Government funding, state incentives, a majority JV partner and a policy-created funding mechanism all shrink the sponsor's capital at risk — check the equity share against the capital share.
5. Track the gates as discrete catalysts rather than waiting for completion; each one that clears re-rates the option without changing the core.

• JV percentage versus funding percentage; government grants and state incentives as the sponsor's de-risking; litigation that has become politically entangled (an election year raises the variance); the FID as the single dated catalyst; language that explicitly ring-fences the project from the operating business.

6. Audit the two quiet items: counterparty concentration, and the quality gradient inside the asset base

1. For any producer, ask who buys the output and who owns the shares. When the same party does both at scale, the relationship is a structural dependency — supportive today, but a single point of failure and a governance consideration.
2. Quantify both: percentage of shares held, percentage of sales taken. Two large numbers on the same name is the flag, not either alone.
3. Inside the asset base, rank plants by cost, age and utilisation rather than counting them. A company running fewer, better plants at capacity is worth more than one running more plants below capacity — divestment of the worst asset is an upgrade, and the sale proceeds are a second benefit.
4. Look for a non-price differentiator in the asset base that a competitor cannot cheaply copy — verified low-carbon output, captive feedstock, a long-dated power contract. These matter when a regulatory regime (a carbon border adjustment, a sourcing rule) starts pricing them.
5. Treat unplanned outages as a base-rate, not an anomaly, in energy-intensive processing — and check whether the most recent ramp is fully de-risked before extrapolating a record quarter.

• A single name appearing as both large shareholder and largest customer; divestments of the highest-cost asset (a margin upgrade and a cash inflow at once); certified/verified product lines tied to a tightening regulatory regime; a plant that only just reached nameplate capacity, whose record quarter is not yet a run rate.