Yield

Bottom Line

Alcoa's adjusted free cash flow — reported FCF less stock compensation less the five-year average of acquisition spend — was $526M in FY2025: a 4.45% yield on an $11.81B market cap, against a three-year average of 0.16%. The balance sheet selects the framework's most forgiving reference line, 8–9%; FY2025 sits 355 bps below it. Consensus reaches that line only in FY2028, and the announced $3.1B South32 purchase reinstates a $620M annual deduction that takes it away again.

The Adjustment Line by Line

The framework's yield basis strips two things out of reported free cash flow: stock-based compensation, because it is a real cost paid in shares, and the trailing five-year average of acquisition spend, because serial buyers are not free to hand the cash to shareholders. For Alcoa the first deduction is small and the second, historically, is zero.

The deterministic feature file could not compute this series: fit_features.adjusted_fcf returns null for every year, with not_computable recording "missing SBC for FY 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023, 2024, 2025; no complete consecutive five-year acquisition window with SBC." The structured cash-flow feed carries operating cash flow and capital expenditures but no stock-compensation line. The table below therefore rebuilds the adjustment from the filed Statements of Consolidated Cash Flows, which disclose stock-based compensation directly for all seven years. Reported FCF matches fit_features.adjusted_fcf.series[].fcf exactly, year for year.

Adjusted FCF = reported FCF − SBC − 5-yr average acquisition spend; derived from company filings. Reported FCF and capital expenditures from the FY2025 Statement of Consolidated Cash Flow [1]; SBC for 2019–2021 from the FY2021 statement [2] and for 2021–2023 from the FY2023 statement [3].

Stock compensation runs $25M to $41M a year — under 8% of reported FCF in the good years, and immaterial to the answer [4]. The acquisition deduction is zero for a specific reason: there is no acquisitions line anywhere in Alcoa's investing activities from FY2016 through FY2025, and the one large deal in that decade — the August 2024 buy-in of Alumina Limited — was settled in stock, appearing as a $2,377M credit to additional capital rather than a cash outflow [5]. The zero is a fact about payment form, not about restraint. What the cash-flow statement did not charge, the share count did: 178M shares at the end of FY2023, 261M at FY2025, and 263,909,445 outstanding on July 27, 2026 [6]. That dilution lands in the denominator of every yield below, not in the numerator.

Source: derived from reported financials, FY2021, FY2023 and FY2025 Statements of Consolidated Cash Flow [7] [8] [9].

Yield on Three Bases

Market capitalisation is $11,812.86M — 261M shares at the $45.26 close of July 31, 2026 (fit_features.market_cap). Three readings follow from the same numerator series.

Adjusted FCF ÷ market capitalisation; derived from company filings and the $45.26 close of July 31, 2026. Numerator components from the FY2025 Statement of Consolidated Cash Flow [10].

Current. $526M ÷ $11,812.86M = 4.45%.

Three-year average. Adjusted FCF of −$475M, $6M and $526M across FY2023–FY2025 averages $19.0M, or 0.16%. Widening the window barely helps: the five-year average is $170.0M (1.44%) and the seven-year average $163.3M (1.38%). The single best adjusted-FCF year in seven is FY2025's $526M.

The company's own baseline. fit_features.yield_baseline is not_computable — the feature needs adjusted FCF, which it does not have. Applying the feature's own stated method (adjusted FCF ÷ same-year shares × last close on or before fiscal year-end) to the filed figures gives the series below.

Source: derived from reported financials and year-end closing prices; adjusted FCF per the table above [11].

The median of those seven years is 3.67%, with a range from −7.85% to 6.96%. Today's 4.45% is 1.21 times that median. The fortress signature the framework hunts — a stable 3.5–4% name repricing to 8–9% — requires roughly a doubling; Alcoa has not doubled off its baseline, and the baseline itself is not stable enough to be a baseline. Three of the seven years sit at or near zero.

That last point carries the P2 stability test directly. fit_features.fcf_stability is not_computable for the same missing-SBC reason; computed from the filed series, the rolling five-year average of adjusted FCF is $122M (FY2019–23), $68M (FY2020–24) and $170M (FY2021–25) — the average itself moves by a factor of 2.5 across three consecutive windows, a coefficient of variation of 0.35. The framework tolerates an occasional negative year every five to eight years as the price of 25–30% margins in the good ones. Alcoa's FY2023 was negative, and FY2020 and FY2024 rounded to nothing.

Which Bar Applies

fit_features.balance_sheet_class returns "unknown", with not_computable recording "EBITDA missing for FY 2025." It does supply net debt: $842M, which reconciles to the filed balance sheet as $2,438M of long-term debt plus $1M due within one year less $1,597M of cash [12]. Alcoa's disclosed profit measure is Total Segment Adjusted EBITDA, $1,940M in 2025 and $2,065M in 2024 [13]. Consensus EBITDA for the same year is $1,964.8M on nine estimates, 1.3% away, so the two definitions can be used interchangeably here.

The rule reads: fortress if net debt is nil or net debt / EBITDA is 0.5 or below; levered at 3.0 or above; moderate in between.

Net debt from fit_features.balance_sheet_class.net_debt, tied to the FY2025 balance sheet [14]; EBITDA from segment disclosure FY2025 [15] and FY2023 [16]; the post-acquisition figure is management's own stated ceiling [17].

On FY2025 figures Alcoa lands in the fortress band at 0.43 times, which selects the framework's lowest bar. That is the reading most favourable to the company, and it is the one used here. It is also cycle-dependent: run the same $842M against FY2023's $734M of segment EBITDA and the ratio is 1.15 times, squarely moderate [18]. And management has told the market where the post-deal balance sheet will sit: the cash portion of the South32 consideration was sized so as "not exceed a leverage ratio of 2.0x based on recent pricing" [19]. Two times is moderate, and moderate selects 10%.

Stated in plain arithmetic against the most forgiving line: 4.45% on FY2025 adjusted FCF against the 8% fortress bar — 355 bps short. Against the top of the fortress band at 9%, 455 bps short. Against the 10% default bar the post-deal balance sheet selects, 555 bps short. In dollars, the 8% line asks for $945M of adjusted FCF; FY2025 produced $526M, and the seven-year average produced $163M.

Mid-Cycle Normalization

Alcoa is a commodity cyclical, so the FY2025 figure needs testing against a normalized year. Total Segment Adjusted EBITDA over FY2019–FY2025 ran $1,626M, $1,317M, $3,053M, $2,280M, $734M, $2,065M and $1,940M [20] [21] [22]. The mean is $1,859M and the median $1,940M. FY2025's $1,940M is the median. On the profit line, FY2025 was a mid-cycle year, not a depressed one — which means the low yield is not a cycle artefact waiting to unwind.

The normalization therefore has to run through the cash bridge rather than through the profit line. Assumptions, stated so they can be re-run under alternates:

1. Mid-cycle EBITDA $1,859M — the arithmetic mean of the seven years above. Using the median instead adds $81M.
2. Unallocated costs −$308M — the three-year average of transformation, corporate expenses and other unallocated items ($268M in 2023 [23], $315M in 2024 and $342M in 2025 [24]).
3. Intersegment eliminations +$30M — the seven-year average of a line that swings from −$231M to +$252M [25].
4. Cash interest −$128M — the FY2025 amount paid, net of capitalisation [26].
5. Cash taxes −$250M — between the $167M paid in FY2025 and the $319M paid in FY2023; Alcoa pays essentially no U.S. federal tax, with Australia the largest single jurisdiction [27].
6. Capital expenditure −$750M — the company's own 2026 projection, of which $675M is sustaining [28].
7. Legacy cash outflows −$203M — pension contributions of $20M plus the three-year average $183M reduction in noncurrent liabilities, the cash cost of asset-retirement, environmental and restructuring obligations [29].
8. Working capital $0 — assumed neutral through the cycle. This assumption is generous, and the next section shows why.
9. SBC −$37M — the three-year average.

Source: derived from reported financials, FY2019–FY2025 10-Ks; assumptions listed above [30] [31] [32] [33].

Mid-cycle adjusted FCF of $213M is 1.80% on today's market cap. A second route reaches the same place: reported FCF as a share of Total Segment Adjusted EBITDA across the six positive years averaged 14.3%, and $1,859M at 14.3% less $37M of SBC is $229M, or 1.94%. The seven-year realized average of $163M sits just below both, at 1.38%. Call the mid-cycle band 1.4% to 1.9%.

Run the bridge backwards and the requirement becomes concrete. An 8% yield needs $945M of adjusted FCF, which at the bridge's cost lines needs $2,592M of segment EBITDA. A 10% yield needs $1,181M, which needs $2,828M. Alcoa has cleared $2,592M once in the seven years above, in 2021.

The honest counter to that: it is clearing it right now. First-half 2026 Total Segment Adjusted EBITDA was $1,631M against $1,034M a year earlier, an annualized $3,262M [34]. The EBITDA level the bars require is reachable at current aluminum prices. What is not reachable, on the current evidence, is the cash conversion — see the next two sections.

Cash Conversion

Free cash flow as a share of revenue over the last decade: $1,542M of cumulative FCF on $113,972M of cumulative revenue, or 1.35%. Split in halves, FY2016–FY2020 converted at 0.93% and FY2021–FY2025 at 1.74% — an improvement, but both under 2%, and the year-to-year path tracks the aluminum price rather than any operating trend.

Source: derived from reported financials; revenue from the FY2025 segment reconciliation [35], cash flow from the FY2025 statement [36].

The binding constraint sits between EBITDA and cash, and the first half of 2026 shows it working in real time. Cash from operations was $429M against $563M a year earlier, on capital expenditure of $305M — free cash flow of $124M, or $94M after the $30M of stock compensation booked in the period [37]. That is a 7.6% conversion of the half's $1,631M of segment EBITDA, in the strongest six months of revenue in the company's history.

The cause is visible on the same page: receivables absorbed $440M, inventories $128M and payables $101M, $669M in total, driven by higher aluminum pricing [38]. The same pattern ran in FY2024, when receivables consumed $493M [39]. Alcoa's working capital is indexed to the metal price: the better the price, the more cash the balance sheet swallows on the way up. That is why 2021's $3,053M of EBITDA converted to $530M of FCF and a 4.34% yield on that year's market cap — the best profit year of the decade produced a yield barely half the fortress bar.

The counter-fact that matters: a working-capital build is a level effect, not a rate. If aluminum prices plateau rather than keep rising, receivables stop growing and conversion normalizes toward the bridge above. Management reported $422M of free cash flow in the second quarter alone and a $1.4B cash balance, with adjusted net debt of $1.4B inside its target range [40]. A flat-price second half would look materially better than the first.

The Acquisition Line

On June 30, 2026 Alcoa agreed to acquire South32's bauxite, alumina and aluminum interests for $3,100M of cash plus approximately 17 million shares valued at about $1,000M, with a ticking fee of 5% a year on the cash consideration and up to $750M of further cash contingent on alumina and aluminum prices over four annual periods [41]. Bridge financing of up to $3,100M was committed the same day [42]. Management estimates $80M to $100M of ticking fees payable at closing, partly offset by a locked-box mechanism it values at more than $200M as of June 30, and expects South32 shareholder approval in October or November [43].

The framework's acquisition deduction, dormant for a decade, switches on. A $3,100M cash purchase carries a trailing five-year average of $620M a year for the five years it sits in the window. Applied to the numbers already established:

Adjusted FCF = reported FCF − SBC − 5-yr average acquisition spend; derived from company filings, with the acquisition line set at $3,100M ÷ 5 per the announced terms [44]. Yields on the pro-forma market cap of $12,582M — 261M shares plus 17M issued, at $45.26.

The deduction is larger than the entire year of adjusted cash flow it is deducted from. It also raises the denominator: 17 million new shares, about 6% of the post-issuance count, take pro-forma market capitalisation to roughly $12,582M at the current price [45]. The capital-allocation reading of this — whether cash pointed at acquisitions can also be pointed at repurchases — belongs with the buyback record in Self-Help.

Two things cut the other way, and both are real. Management identified approximately $900M of net-present-value synergies with $50M of run-rate cost savings from the first year, and expects the deal to be accretive to earnings and cash flow immediately on closing [46]. And the acquired assets bring roughly 5.2 million tonnes of alumina and 900 thousand tonnes of aluminum capacity — a 53% and 37% pro-forma increase [47]. The framework's deduction charges the price without crediting the earnings it buys; a reader who thinks the assets are worth more than $3.1B should read the line as a timing charge rather than a permanent one. The closing conditions, including South32 shareholder and regulatory approval, are not yet satisfied [48].

The Consensus Check

fit_features.consensus_forward_yield draws on a direct free-cash-flow consensus from the vendor estimate file — not a proxy, though the count of contributing estimates is not disclosed for that item, unlike EBITDA (nine to ten estimates through FY2027, seven in FY2028) and revenue (twelve to thirteen through FY2027, nine in FY2028). Vintage: the estimate file was compiled on August 3, 2026, with per-metric momentum stamps dated August 2, 2026.

Source: consensus estimates per fit_features.consensus_forward_yield, compiled August 3, 2026; "as reported" yields on $11,812.86M, fully adjusted yields on the pro-forma $12,582M. Acquisition deduction of $620M per the announced terms [49].

Taken at face value, consensus clears both bars: 13.25% for FY2027 and 16.09% for FY2028. Subtract stock compensation and the picture barely moves. Subtract the acquisition line the framework requires, and FY2027 falls to 7.18% and FY2028 to 9.85% — below the 8% fortress line in the first year and 15 bps short of the 10% line the post-deal balance sheet selects in the second. On the framework's own basis, the sell side does not agree; it comes close in one year, four years out, and then recedes.

Three checks on the consensus itself, in fairness to both directions:

It was roughly right on the last completed year. The FY2025 consensus FCF of $533.8M compares with $567M actual — 6% light, not systematically optimistic.

It is running ahead of the current year. FY2026 consensus operating cash flow is $1,495M; the first half delivered $429M [50]. FY2026 consensus capital expenditure is $709M against the company's own $750M projection [51]. Hitting the $880M FCF mean requires $756M in the second half.

Its conversion assumptions have no precedent here. Consensus FCF as a share of consensus EBITDA runs 29.1% in FY2026, 49.8% in FY2027 and 63.6% in FY2028. Alcoa's realized conversion of FCF to Total Segment Adjusted EBITDA over FY2019–FY2025 was 18.9%, 3.1%, 17.4%, 15.0%, −59.9%, 2.0% and 29.2%. The FY2026 assumption matches the best year on record; the FY2027 and FY2028 assumptions are roughly double and quadruple it. Note that the FY2027 EBITDA estimates themselves span $2,221M to $4,428M across ten contributors, and the FY2028 mean of $2,989M sits below the FY2027 mean with a low of $1,818M — the dispersion is as wide as the forecast.

The mean-reversion underwrite

Consensus sits below the applicable bar in FY2026 on any adjustment basis, so the path back above it has to be written out rather than assumed.

The mechanism is price and volume, not cost. Aluminum realized $3,376 per tonne in 2025 against $2,841 in 2024 [52], and the Aluminum segment posted record quarterly EBITDA of $1.1B in Q2 2026 at a 32.3% margin [53]. The Alumina segment is the drag: it produced −$136M of EBITDA in the first half of 2026 [54], against $882M for the full year 2025 on an average realized alumina price of $415 per tonne [55]. Alumina recovery plus the acquired capacity is the whole of the step-up consensus is underwriting.

What consensus would have to concede to be wrong: that a step-up in EBITDA does not carry through to cash at 50–64%, because working capital, $750M of capital expenditure, $250M-plus of cash taxes and roughly $200M a year of legacy pension, asset-retirement and environmental outflows stand in the way — as they did in 2021, when $3,053M of EBITDA became $530M of free cash flow.

Probability. On the post-acquisition basis — the 10% bar, $1,258M of adjusted FCF required on a $12,582M pro-forma market cap — the estimate here is roughly 15% (range 10–20%) that Alcoa clears it in any single year of FY2027–FY2029. The basis: consensus at its FY2028 mean produces $1,240M, already 15 bps short, and to clear requires FCF above $1,919M in a year whose EBITDA estimates span $1,818M to $3,793M across seven contributors; Alcoa's highest reported FCF in ten fiscal years is $819M, in 2017. If the transaction does not close, the fortress line applies and the deduction does not, and the probability rises to roughly 30%: the FY2027 consensus of $1,524M after SBC clears $945M comfortably, but still asks for nearly double the best free-cash-flow year in company history.

What would change this read, in either direction: two consecutive halves in which EBITDA-to-FCF conversion holds above 30% with prices flat; the Alumina segment returning to a positive EBITDA run rate near its 2025 level; or capital expenditure settling below the $675M sustaining figure. In the other direction, a second year of working-capital absorption at the H1 2026 rate, or the contingent $750M becoming payable, would push the fully adjusted yield further from both lines. Whether the fall in price that produced today's entry point reflects damage of the same size is settled in Damage Math.