Chapter 4

What the Price Assumes

Alcoa closed 2025 with the largest statutory profit in its history as a standalone company: net income attributable to Alcoa of $1,157 million, against $60 million the year before [1]. In the same year, the metric management actually runs the business on — Total Segment Adjusted EBITDA — fell, to $1,940 million from $2,065 million [1]. A record profit and a lower operating result in the same twelve months is the reconciliation this chapter has to settle, because the multiple the market puts on Alcoa depends entirely on which of those two numbers is treated as the base.

The capital ledger (The Operator's Word) treated FY2025's headline earnings only as the thing that let management reach its net-debt target. Here they are the problem to solve: what part of that profit is repeatable, what the low forward multiple is quietly built on, and why a stock that screens cheap on cash flow has also fallen further than the one priced peer in its set.

The record profit is mostly a one-time gain

The gap between the two numbers is one line. Consolidated pretax income of $1,064 million in FY2025 came almost entirely from "Other (income) expenses, net" of $1,057 million — a line that was a $91 million expense the year before [1]. Two events fill it. On July 1, 2025, Alcoa sold its 25.1% interest in its Saudi Arabia joint venture to Ma'aden for total consideration of $1,350 million — of which only $150 million was cash — and booked a gain of $786 million, net of costs [2]. The Ma'aden shares it received were then marked to market for a further $197 million gain [2]. Together, roughly $983 million — about 92% of pretax income — is a non-recurring gain that was overwhelmingly non-cash.

Source: FY2025 Annual Report (Form 10-K), reconciliation of Total Segment Adjusted EBITDA to consolidated net income [1].

The two bars move in opposite directions because they measure different things. Statutory net income rose nineteen-fold on the asset sale; the operating result fell on weaker segment economics — the same two-speed cycle established in The Price-Taker, with the Alumina segment sliding as the Aluminum segment held. Reported diluted earnings of $4.37 per share [3] carry that one-time gain inside them; a trailing multiple built on that figure is measuring the price against a base the company is unlikely to repeat.

One qualifier keeps this from being a quality flag. The gains never touched operating cash. Both are stripped out in the cash-flow reconciliation — a $784 million removal for asset sales and the $197 million mark-to-market — yet cash from operations of $1,185 million still exceeded net income, supported by $623 million of depreciation and a largely non-cash $918 million restructuring add-back [4]. Over FY2021–FY2025, cumulative operating cash flow of roughly $3.6 billion and free cash flow near $1.0 billion both exceeded cumulative net income of about $0.9 billion, because depreciation runs above capital spending and the loss years were driven by non-cash charges. The cash conversion is genuine over the cycle; it is the headline earnings number, not the cash, that flatters the trailing view.

A footnote to the earnings-quality read, because it recurs. The items Alcoa labels restructuring are not occasional: $918 million in FY2025, $341 million in FY2024, $184 million in FY2023, plus a $144 million goodwill write-off that took goodwill to zero [1]. The audit is otherwise clean — an unqualified opinion, no auditor change, no working-capital build — with a single Critical Audit Matter: the $1,405 million estimate for mine reclamation and bauxite-residue closure that generates those recurring charges as smelters and refineries close [5]. The reporting is trustworthy; the "adjusted" figures simply exclude a real, repeating cost.

What the forward multiple is built on

Measured against next year's earnings rather than last year's, the stock is not expensive. At the July 31, 2026 close of $45.26 and a market value of $11.8 billion, Alcoa trades at roughly 6.8x consensus FY2026 EPS of $6.65 and about 4.2x forward EV/EBITDA, for a forward free-cash-flow yield near 7.4% — versus a trailing multiple close to 12x on normalized earnings.

Source: derived from the 2026-07-31 close and consensus estimates (S&P Capital IQ); market value and net debt per FY2025 reported financials.

The forward multiple sits below the trailing one for a single reason: consensus embeds a step-change in earnings that has not happened yet. Analysts model FY2026 revenue up 15.7% and EBITDA up roughly 52%, with normalized EPS rising 76% year over year. The question is where that jump is supposed to come from, and the broker models are unusually specific about the answer.

Source: Visible Alpha broker models (10 brokers), consensus segment EBITDA; FY2025 actuals per company filings.

Nearly all of the modeled earnings step-up comes from one segment. Aluminum EBITDA is modeled up 245% to about $3.6 billion, while the Alumina segment swings to a modeled loss of roughly $0.3 billion as realized alumina prices near $338 per tonne fall below cash cost near $405 per tonne. And the Aluminum number rests on one price. Broker models put the Section 232-inflated Mid-West premium near $2,342 per tonne in FY2026, close to double its FY2025 level, and that premium alone accounts for the bulk of the segment's per-tonne EBITDA jump.

Source: Visible Alpha broker models (7 brokers), Mid-West premium per tonne; FY2025 actual per company filings.

That is the substance of what the price underwrites. A buyer at $45.26 is not paying for a broad commodity recovery; the aluminum market is already described by management as tight, low on inventory, and in deficit for the year [6]. The buyer is paying for the durability of a tariff-set premium that brokers themselves cannot agree on beyond next year: their FY2028 estimates for that same premium span $838 to $2,340 per tonne. The embedded growth is concentrated in one policy-dependent price, which is why the multiple compresses so quickly the moment the forward year comes into view.

Cheap on cash flow, and de-rated harder than the peer

Two of the sharpest signals in the numbers point in opposite directions at once, and both belong here.

The first is cheapness. Alcoa's trailing free-cash-flow yield of 4.8% is 2.41 times Century Aluminum's 2.0% — the market pays a lower price for each dollar of Alcoa's trailing cash flow than for its one priced peer's (anomaly A2). The second is stress. Over the same window, Alcoa's shares sat 46.0% below their three-year high, against 34.9% for Century — an 1,109-basis-point deeper drawdown (anomaly A3).

Source: reported financials and daily price history to 2026-07-31, as reported; peer set is a single priced comparator (see below).

The two coexist because the cash-flow yield is measured on a trailing year that included a cyclically strong Aluminum result, while the drawdown is measured against a forward tape that has already begun to roll over. The de-rating is not a company-specific stumble. From a three-year low of $22.57 in April 2025 the stock ran 271% to a peak of $83.79 on June 2, 2026, then gave back 46% over roughly six weeks — a near-continuous slide rather than a single earnings gap.

Source: daily price history to 2026-07-31, as reported.

Management dates the reversal to the metal, not the franchise: on the Q2 2026 call it attributed the quarter's soft price realization to LME prices that "declined sharply in the final two weeks of June" [7], and described the LME as having "returned to pre-Middle East conflict levels following a macro-driven correction" while regional premiums held [6]. The expectation reset tracks the aluminum tape rolling over, which is why the cheapness and the stress can both be true.

The peer comparison carries a real limit worth stating plainly: of the six intended comparators, only Century has usable price history loaded. Rio Tinto and South32 have none, and the others lack income statements, so both the "cheaper multiple" and the "deeper drawdown" reads rest on a single name rather than a peer median in any meaningful sense.

The base the multiple is measured against is unstable

A low multiple is only as reliable as the earnings underneath it, and the consensus anchoring Alcoa's is unusually loose. The company has missed both revenue and normalized EPS for two straight quarters. FY2027 normalized-EPS consensus was cut about 18% in the prior 30 days. And the forward numbers carry extreme dispersion: FY2027 EPS estimates span $2.74 to $9.17 across eleven analysts, or 108% of the mean, with FY2028 resting on just five.

Source: consensus estimates (S&P Capital IQ), FY2027 estimate dispersion.

That dispersion is the operating leverage of The Price-Taker showing up in the estimate sheet: because a few-percent move in the aluminum price swings hundreds of millions of dollars of EBITDA, small differences in analysts' price decks produce enormous differences in modeled earnings. The forward multiple is precise; the earnings it divides into are not.

A second instability sits in the cash-flow forecast. Broker models show free cash flow building toward roughly $1.5 billion in FY2027 and net debt swinging to a net-cash position, with that capacity "largely unallocated" and per-share dividends still near zero. Yet the company is simultaneously committing $3.1 billion of cash and about $1 billion of equity to the South32 acquisition [8]. The modeled net-cash build and the deal's outflow point in opposite directions — the consensus cash-flow scalar almost certainly excludes the acquisition — so the "swing to net cash" and the "$4 billion deal" cannot both be read at face value.

The equity leg of that deal compounds a dilution problem the capital ledger already established. Diluted shares rose 22% in FY2025 alone, against a peer group essentially flat, and the pending transaction adds roughly $1 billion more Alcoa stock on top [8]. Absolute free cash flow and per-share free cash flow diverge sharply here; the modeled cash build accrues to a share count that keeps rising. On the balance-sheet side, the two figures a reader has to hold are the FY2025 reported net debt of $842 million and the adjusted net debt of $1.4 billion management cited at the end of June 2026 — the top of its own target range [9], and before any of the South32 cash goes out the door.

What would resolve the price

The market's own read is more constructive than the tape: brokers stand at seven buy, five hold, one underperform, with twelve price targets averaging about $63 and a median of $59.50 — every target above the $45.26 close. The estimate feed carries no live share price, so that gap is best read as sentiment and implied upside rather than a verified quote. It frames the price as discounting the premium's durability more than the analysts' central case does.

What the price is measured on comes down to a short list of checkable items, each falsifiable against a named line in a future filing.

Sources: Q2 FY2026 earnings call, acquisition terms [8]; consensus estimates (S&P Capital IQ) and Visible Alpha broker models.

Reconciled, the picture is coherent rather than contradictory. FY2025's record earnings overstate the base, but the cash conversion under them is real. The forward multiple is genuinely low, but it is low because it is built on a 76% earnings jump concentrated in a single tariff-set premium that even its own modelers cannot agree on past next year. The stock is cheaper on trailing cash flow than its one priced peer, and it has fallen harder — because the cheapness looks backward at a strong year and the drawdown looks forward at a rolling-over price. The two anomalies that opened this report's arithmetic are not a puzzle once the base is sorted; they are the same commodity leverage, read once through last year's cash and once through next year's price. What is left genuinely unresolved is narrow and nameable: whether the Mid-West premium holds near double its historical level, whether alumina climbs back above its own cash cost, and how much of the per-share story the South32 stock consideration resets when the deal closes.