Dislocation

What happened to the price

Alcoa fell 46% in eight weeks — $83.79 on 2 June 2026 to $45.26 on 31 July — in three dated legs: an aluminum-price unwind through June, an 8.9% drop on the 30 June South32 acquisition, and a 6.1% drop on record second-quarter results that landed 3.2% short of consensus. Traded volume through the fall measures 1.23x its pre-peak median: repricing, not capitulation. Consensus earnings estimates were cut after the price moved, not before.

The drawdown, quantified

Source: daily closing prices, company price history as reported; the same series that feeds the deterministic capitulation gauge.

The deterministic feature file measures a different episode. Its capitulation_gauge.drawdown records a peak of $47.42 on 26 November 2024, a trough of $22.57 on 8 April 2025, a depth of 52.4% and 133 days from peak to trough, with the current close of $45.26 on 31 July 2026. Those figures are arithmetically correct and describe the tariff-era fall of 2024–25 — an episode that has since healed: at $45.26 the stock sits 4.6% below that November 2024 peak. The gauge appears to select the deepest fall in the price record rather than the live one, and the 2024–25 fall is 3.6 points deeper than the current one. The drawdown a buyer would be entering today is the June–July 2026 fall, computed here from the same daily price file, and it is the one this tab anatomises. The mismatch is recorded as a data gap.

Source: weekly closes drawn from the run's daily price history; the daily peak, trough and current levels reconcile to the deterministic capitulation gauge in fit_features.capitulation_gauge.drawdown. Prices as reported, not adjusted for the pending share issuance.

The chart shows two distinct falls with a 138% advance between them. From $35.26 on 4 November 2025 the stock rose to $83.79 on 2 June 2026 — 143 sessions — as Middle East supply disruption lifted the aluminum price; management put the smelting capacity offline inside the Strait of Hormuz at 3 to 3.5 million metric tons [1]. The fall that followed gave back that advance and rather more of it. On 261 million shares, market capitalisation went from $21.9 billion at the peak to $11.8 billion on 31 July — a fall of $10.1 billion. The 52-week range is $28.40 to $83.79, and the current price sits at 30% of that range.

The trigger

Four of the eight weeks carry an identifiable dated event; the rest is a one-way grind in the metal price.

Sources: closing prices from the run's daily price history; event column from the Form 8-K of 1 July 2026 [2] and the Q2 FY2026 earnings call of 16 July 2026 [3].

Leg one — the metal price, no filing. Between 2 and 30 June the stock lost 37.8% with no company filing of any kind. The explanation came six weeks later, from the company: the LME aluminum price "has returned to pre-Middle East conflict levels following a macro-driven correction" [4]. Asked directly why the price retreated when the disruption had not been resolved, the CEO answered: "The first answer is sentiment. The fundamentals from when the Iran conflict started have not fundamentally changed" — the 3 to 3.5 million tons of Strait of Hormuz capacity remained offline [5]. This leg is the largest of the five and the least company-specific. It is a commodity repricing that a shareholder experienced through the equity, not an event at Alcoa.

Leg two — the acquisition. On 30 June 2026 Alcoa entered an Umbrella Implementation Deed to acquire South32's interests in bauxite mining, alumina refining and aluminum smelting [6]. The upfront consideration is $3.1 billion of cash plus approximately 17 million Alcoa shares carrying an agreed value of about $1 billion — struck on the 10-day volume-weighted average price to 26 June of $58.79 — representing roughly 6% of shares outstanding after issuance, with up to a further $750 million payable in cash if alumina and aluminum prices exceed agreed strike prices over four annual periods [7]. The cash leg is backed by a committed senior unsecured 364-day bridge facility of up to $3.1 billion from a single commitment party, to be refinanced with senior unsecured notes before closing [8]. The market reaction on 1 July was a fall of 8.9% on 16.9 million shares, the heaviest session of the drawdown and the heaviest since 23 October 2025. A share-issuing, debt-funded acquisition that lifts the count by roughly 6% is directly relevant to the share-count test carried in Self-Help; at $45.26 the equity consideration struck at $58.79 is worth 23% less than its agreed value.

Leg three — the quarter. Q2 2026 revenue rose 24% sequentially to $3,966 million; the realised primary aluminum price rose from $4,209 to $4,752 per metric ton and adjusted EPS from $1.40 to $2.12 [9]. The CFO called that the highest quarterly revenue in Alcoa's almost ten-year history, and set against it a print "modestly below consensus," the variance driven by "lower-than-expected aluminum price realization late in the quarter, as LME prices declined sharply in the final two weeks of June" [10]. Adjusted EPS of $2.12 came in 3.2% under the $2.19 consensus. The same call lowered full-year alumina production guidance to 9.5–9.6 million metric tons and shipments to 11.5–11.6 million, raised full-year other corporate expense to about $180 million and depreciation to about $660 million [11]. The stock fell 6.1% the next session on 11.7 million shares.

Separating the event legs from the drift: of the $38.53 the stock lost between 2 June and 31 July, the two dated event days account for $7.53, or 19.5%. The undated June grind alone accounts for $31.65, or 82%, before the late-July rebound of $2.38. The framework's caution about a 10–20% slide on no event and normal volume does not map cleanly here, because the undated leg is 38% rather than 15% — but it remains true that four fifths of what happened to Alcoa's price happened to the aluminum price, not to Alcoa.

For orientation, the prior fall the feature file measures had a cleaner company-specific trigger: US Section 232 tariffs of 25% on aluminum imported from Canada took effect on 12 March 2025 and cost roughly $20 million in that quarter alone, before the April 2025 macro selloff took the stock to $22.57 [12].

The fear gauge

Source: daily traded volume from the run's price history, 1 June to 31 July 2026; the pre-peak benchmark is the 5.71 million-share median of the 180 calendar days before 2 June 2026. One non-session record dated Saturday 18 July is excluded as a feed artefact.

The deterministic gauge reports a volume spike multiple of 1.23x, defined as the maximum 20-day average volume in the peak-to-trough leg divided by the median daily volume over the 180 days before the peak (fit_features.capitulation_gauge.volume_spike). That figure is computed on the 2024–25 leg. Applying the identical definition to the live June–July 2026 leg gives 1.23x as well: a peak 20-day average of 7.02 million shares (10 June to 9 July) against a pre-peak median of 5.71 million. The two episodes are indistinguishable on this measure.

Individual sessions ran hotter than the 20-day average implies. The heaviest were 1 July at 16.9 million shares — 3.5x the trailing 50-day average — then 15 June at 12.4 million (2.8x) and 17 July at 11.7 million (2.1x). For scale, the ten largest volume days in Alcoa's recorded history run between 4.8x and 6.7x their trailing 50-day averages. No session in this drawdown reaches that band.

The tape has since gone quiet rather than climactic: 20-day average volume on 31 July was 5.20 million shares, below the 5.71 million median that prevailed before the fall began. On the framework's own test — volume must have spiked, peak fear rather than the start of the slide — a 1.23x measured multiple and a post-fall tape running below its pre-fall average describe orderly repricing by holders who could sell, not forced or exhausted selling. Whether the price is nonetheless wrong is a question for Damage Math and Yield; it is not answered by the volume.

Who was selling

The direct evidence is thin, and the thinness is itself a finding.

Short interest is unavailable for this run. Every file under data/short_interest/ is empty of rows; the collection manifest records that FINRA returned no reported short-interest rows for the ticker and that no position rows were staged. No level, no change, no days-to-cover can be stated. This is a gap, not a zero.

Insider transactions stop before the window. The Form 4 and 5 record assembled for this run runs from October 2016 to 27 January 2025 — 579 transactions, none inside the June–July 2026 fall. The last open-market sale on record is a $1.0 million disposal by the Chief Commercial Officer on 22 October 2024, at $42.29. No insider buying or selling into this drawdown can be confirmed or ruled out.

Holder-base filings pre-date the fall. The most recent beneficial-ownership filings are Vanguard Capital Management's Schedule 13G of 29 April 2026 (5.5%) and Allan Gray Australia's amendment of 17 February 2026 (4.7%). Both were filed before 2 June. Orbis Investment Management (6.2%) and Eagle Capital Management (5.98%) last filed in November 2024. Nothing in the corpus documents a holder change inside the drawdown.

The one documented structural seller is prospective, not historical. The acquisition creates a known future supply of stock: at least half of the roughly 17 million shares of consideration is to be distributed in specie to South32's shareholders, and the balance is saleable by South32 in an orderly manner [13]. The deed caps that selling at 20% of Alcoa's average daily trading volume on any one trading day for three months following completion [14]. Closing is expected in the first half of 2027 [15]. That overhang sits ahead of the stock, not behind it — it cannot explain the fall that has already happened, and it is a reason the supply picture over the next four quarters is known rather than unknown.

On the framework's distinction between forced or anchored sellers and informed ones, the honest answer for Alcoa is that the corpus does not identify who sold. What it does show is that the selling was absorbed at volumes barely above normal, which is more consistent with discretionary rotation out of a commodity equity whose commodity had fallen than with a liquidation.

Estimates versus price

Source: consensus revision history from the run's CapIQ estimates feed (data/sp/estimates.json, momentum series at 180-day, 90-day, 30-day and current as-of dates), paired with the last closing price on or before each as-of date from the run's daily price history.

Source: derived from the consensus revision history in the run's CapIQ estimates feed and the daily price history; index base 3 February 2026 = 100.

The sequence is unambiguous. Between 4 May and 2 July the price fell 22.1% while consensus FY2027 EPS rose 12.3%, from $6.70 to $7.52, and FY2028 EPS rose 24.0%. The cut came afterwards: in the month to 2 August, FY2027 EPS was marked down 21.1% to $5.93 and FY2028 down 15.7% to $7.04, while the price fell a further 7.0%. Over the full 90 days to 2 August the price is down 27.5%, FY2027 EPS down 11.4%, and FY2027 revenue up 1.4%; FY2028 EPS is 4.5% above where it stood 90 days ago.

Measured from before the spike began, the gap is wider still. On 3 February 2026 the stock closed at $61.35 with FY2027 consensus EPS at $5.23 — 11.7 times forward earnings. On 31 July it closed at $45.26 with FY2027 consensus at $5.93 — 7.6 times. Six months of falling price against rising estimates.

Two qualifications belong with that. First, consensus for a commodity producer two years out is a wide band, not a point: the eleven FY2027 EPS estimates run from $2.74 to $9.17 with a standard deviation of $1.78, so a 21% move in the mean is well inside the dispersion of the panel. Second, the sell side's price targets have not converged on the current quote — twelve targets average $62.98 with a median of $59.50 against a $45.26 close, and the recommendation split is seven buy, five hold, one underperform. Consensus forward free-cash-flow yields on the current $11.8 billion market capitalisation compute to 7.4% for FY2026, 13.3% for FY2027 and 16.1% for FY2028 (fit_features.consensus_forward_yield); what those figures mean against the framework's bar is the subject of Yield.

What the drawdown amounts to

There is a dislocation here, and it is dated. The stock lost 46% in eight weeks against two identifiable documents — the 30 June acquisition and the 16 July quarter — and one undated commodity leg that management itself calls sentiment-driven. On the framework's timing test, the price moved first and consensus followed: the price fell 22% while FY2027 estimates were still rising, and the eventual 21% estimate cut arrived a month after the fall was largely complete.

Two facts cut the other way and are not softened. The volume signature is not capitulation — 1.23x on the measured definition, no session above 3.5x its 50-day average, and a post-fall tape running below its pre-fall average. And a substantial part of what fell was a spike that had been built in seven months: at $45.26 the stock is still 28% above the $35.26 it traded at on 4 November 2025, before the Strait of Hormuz disruption lifted the metal. What that combination is worth — whether the June repricing destroyed intrinsic value proportionate to the $10.1 billion of market capitalisation it removed — is the arithmetic carried in Damage Math, and the temporary-or-permanent question is settled there, not here.