Aluminum analysis — 2026-10-16

Virtually all available LME aluminium at September's close was Russian-origin, Chinese smelters were running at close to full effective capacity, and Chinese domestic ingot stocks fell back below their 2026 low — yet the benchmark price fell 5.2% in the first fortnight of October. Four measures of availability tightened and the price went the other way. The disagreement between them is the question worth asking.

TL;DR

• LME opening stocks stood at 238,625 tonnes on 13 October, below the roughly 250,000 tonnes Reuters described as a 36-year low in mid-August.
• Russian-origin metal accounted for 99.98% of available LME stocks at the end of September, against approximately 95% in August.
• The rise came from alternatives disappearing rather than Russian metal arriving: Indian-origin available stocks fell 12,400 tonnes to 50 tonnes, while Russian tonnage itself fell 3,425 tonnes to 226,625.
• Chinese smelters operated at 99.7% of effective capacity of 45.26 million tonnes in August, with national output running above the 45-million-tonne policy ceiling since March.
• Chinese domestic ingot inventory fell to 636,000 tonnes on 15 October, marginally below the 2026 low of 638,000 tonnes set on 30 September.
• Post-holiday destocking is a recurring seasonal pattern, and ingot stocks had already fallen 158,000 tonnes across the three weeks before the closure.
• The cash bid fell from $3,203/t on 30 September to $3,035.50/t on 13 October, a decline of 5.2% over the fortnight.
• Investment Funds held the largest net long in the LME complex at +124,931.67 lots on 9 October and liquidated longs for a second consecutive week.

Four Measures, One Direction, and a Price Going the Other Way

The aluminium market currently produces four readings that each point towards constrained supply or constrained availability, and a price that has moved against all of them.

Those readings measure different things. Exchange inventory measures metal registered in LME warehouses. Origin data measure what that metal is and who would accept it. Smelter utilisation measures spare operating capacity in the world's largest producer. Chinese social inventory measures physical availability in the largest consuming market. They are not four observations of the same phenomenon, and the fact that all four tightened in the same period does not establish a shortage. A market can be simultaneously fully utilised and adequately supplied.

What the combination does establish is that the conventional reading of exchange inventory as a scarcity signal has stopped working. The 3 September Bloodstone analysis asked whether global supply would recover before exchange stocks did. Six weeks on, exchange stocks have fallen further, Chinese stocks have fallen too, and the price has weakened.

The Warrant Pool Is Now Effectively Single-Origin

The most significant development of the period is in the LME's September country-of-origin data. Russian-origin metal accounted for 99.98% of available aluminium stocks at the end of September, with total available stocks of 226,675 tonnes comprising 226,625 tonnes of Russian material and 50 tonnes of Indian.

The composition of that change matters more than the headline percentage. Russian tonnage fell 3,425 tonnes during September. The share rose because Indian-origin available stocks fell 12,400 tonnes, leaving 50 tonnes on warrant. The warrant pool did not become more Russian because Russian metal arrived; it became more Russian because everything else left. Indian-registered stocks have now fallen from 236,000 tonnes to effectively nothing over roughly fourteen months. The end-September available figure reconciles exactly with the live warrant figure reported for 1, 2 and 5 October, confirming that the origin breakdown and the daily series describe the same pool.

The commercial consequence is that the metal remaining in the system is material many participants do not want. StoneX analyst Natalie Scott-Gray, quoted by Reuters on 14 October, described much of the remaining exchange inventory as unwanted by market participants. The Russian metal on warrant is pre-12 April 2024 production and therefore remains deliverable under LME rules, but deliverability and desirability are different propositions. What this establishes is not that every tonne is inaccessible, but that a falling headline inventory number no longer describes a falling quantity of metal the broad market would willingly take.

Ownership concentration compounds the point, though the evidence is older. Reuters reported in mid-August, citing LME daily position reports, that a single entity held 80 to 90% of all warranted stocks. No corroborated October figure has emerged, so that observation describes the position two months ago rather than today's ownership structure. Its relevance is mechanical: the LME's lending rules, strengthened in March 2025 to cover the whole prompt month where a position exceeds 150% of available stocks, constrain how a dominant holding can translate into spread pressure.

China Has No Operating Headroom Left

Chinese production has exceeded its nominal 45-million-tonne policy ceiling since March. AZ Global estimated that smelters operated at 99.7% of effective capacity of 45.26 million tonnes in August, a figure Reuters published on 9 September. First-half output reached 22.34 million tonnes, up 2.24% from 21.85 million tonnes a year earlier, against full-year 2025 production of 45.02 million tonnes.

The ceiling remains in force as policy and was renewed under the 2025–2027 Action Plan. What has happened is capacity creep rather than policy abandonment: operating improvements allow actual output to exceed nameplate, with Chen Jingmin of Zijin Tianfeng Futures putting the achievable margin at around 3%. New domestic units still require capacity replacement elsewhere.

The correct inference is narrow. Utilisation at 99.7% means there is almost no further upside from Chinese primary production, which removes the mechanism that has absorbed global demand growth for two decades. It does not by itself demonstrate that the market is short. A producer running flat out is supplying the maximum it can, which is a statement about elasticity rather than about adequacy.

Data Centres Add Demand and Compete for Power

Alcoa's third-quarter results, published on 15 October, disclosed definitive agreements to sell portions of its former Massena East site in New York, with potential proceeds of up to $315 million. The payment structure is instructive: approximately $50 million to $65 million at closing depending on the amount of approved data centre capacity, and a further $60 million to $250 million in capacity expansion and instalment payments as approved capacity is commissioned over the following seven years. Closing is expected within six months, subject to permitting, regulatory approvals and construction conditions. Chief executive William Oplinger told CNBC the same day that data centre construction was positive for aluminium demand, reflecting the metal's use in electrical infrastructure, cooling equipment and construction.

The development cuts both ways for primary supply. Smelting is electricity-intensive, and data centres compete for the same power and grid connections — which is precisely why closed and curtailed smelter sites are attractive to developers, since the grid capacity is already in place. For producers that creates a potential source of demand alongside an additional constraint on establishing new primary capacity. Neither effect explains the fortnight's price decline, and the additional aluminium demand attributable to data centres has not been reliably quantified.

Chinese Stocks Fall, But Not Only Because of Demand

SMM data put Chinese primary ingot inventory at 636,000 tonnes on 15 October, down 50,000 tonnes on the week and marginally below the 2026 low of 638,000 tonnes recorded on 30 September. The 6063 billet book fell 19,000 tonnes to 168,500 tonnes, a combined domestic withdrawal of 69,000 tonnes in a single trading week.

That withdrawal needs a seasonal frame. SMM's historical analysis shows post-National Day destocking to be a recurring annual pattern rather than an exceptional 2026 development, and ingot stocks had already fallen 158,000 tonnes across the three weeks preceding the closure, the steepest comparable pre-holiday decline in SMM's series. October's draw is therefore the normal reversal of holiday accumulation, following an unusually steep pre-holiday withdrawal.

A second mechanism affects what the ingot figure measures at all. SMM's June research found that Chinese ingot destocking was not driven solely by consumption: smelters were directing more liquid aluminium straight into downstream processing rather than casting it into ingots, reducing the volume entering the ingot inventory system in the first place. The liquid-aluminium proportion reached 76.5% in May, with ingot-to-billet substitution also contributing to warehouse withdrawals. SMM's October weekly review expects that proportion to decline slightly this month, which would increase ingot casting and add to available ingot supply. The June findings do not establish the cause of October's specific withdrawal, but they establish that the ingot inventory series is partly a measure of production allocation rather than purely of demand — which bears directly on how much scarcity the 636,000-tonne figure represents.

Two further qualifications apply. Billet stocks remain around 21,000 tonnes above the corresponding 2025 level, so the drawdown reduces an elevated position rather than establishing a shortage. And bonded ingot inventory rose 1,300 tonnes to 115,900 tonnes over the same week, meaning imported metal awaiting customs clearance built while domestic metal drew.

The Downstream Recovery Is Narrow

Overall surveyed downstream utilisation reached 61.7%, up 1.2 percentage points. The sector breakdown shows how little of that is broad-based.

Secondary aluminium accounts for almost the entire improvement. Three of six segments were flat or lower, including plate, sheet and strip. SMM attributes the aggregate improvement to post-holiday restarts and order backlogs rather than a recovery in end demand, and the distribution supports that reading: a synchronised demand recovery would show in more than one segment.

The Price and the Curve

The cash bid fell from $3,203/t on 30 September to $3,035.50/t on 13 October, a decline of 5.2% across the fortnight. That period straddles the Chinese National Day closure on the physical side but represents a continuous series on the LME.

The curve requires care. Bid-to-bid on 13 October, cash at $3,035.50 against three-month at $3,036 implies a contango of $0.50/t, having narrowed from roughly $10 to $14 in the first week of October. Reuters reported on 14 October that the spread had moved back into a modest backwardation. These are different dates and potentially different price conventions, and both are reported here as published rather than reconciled. What is consistent is that the spread is close to flat and has spent October in that region, against the $8/t backwardation Reuters cited in mid-August.

The global picture offers a mechanism for the apparent contradiction. Reuters reported on 14 October, citing the International Aluminium Institute, that Gulf primary production fell by an annualised 2.8 million tonnes between February and May. The market absorbed that disruption through increased Indonesian primary output and higher Chinese semi-finished exports. Lost primary supply in one region can therefore be met from elsewhere, through production growth and changes in trade flows, without the global balance tightening in the way a single-region disruption would imply.

A producer's realised pricing corroborates the weakness independently of exchange quotations. Alcoa reported an average realised third-party aluminium price of $4,527 per tonne in the third quarter, down from $4,752 in the second, with Aluminum segment third-party revenue falling 9% sequentially on lower shipments and realised prices. Realised prices include regional premiums and product value-add and are therefore not comparable in level to the LME cash bid, but the direction is the same.

Fund positioning adds the other side. Investment Funds held +124,931.67 lots net long on 9 October, by some distance the largest net position in the LME complex, and reduced longs by 11,202.87 lots for a second consecutive week of comparable liquidation. A large long being unwound is itself a source of price pressure independent of physical balance, and the scale of the remaining position means that pressure has further to run if liquidation continues.

Guinea Remains a Proposal

Guinea's bauxite export restriction has been discussed since March without becoming an enacted quota. Mines Minister Bouna Sylla told Reuters on 18 March that it was not really a quota but that export volumes would be reduced, and Reuters reporting on 28 April still described restrictions as forthcoming. A figure of 150 million tonnes has circulated since, against 2025 exports reported at 178 million tonnes on Oceanbolt data and 182.8 million tonnes by the Guinean Ministry of Mines, while shipbroker Arrow has put 170 to 190 million tonnes as a more realistic outcome. This belongs in the analysis as unresolved alumina cost risk rather than as a supply event.

Outlook

The central question for the next several weeks is whether the price weakness is positioning or fundamentals. Funds hold the complex's largest net long and have liquidated at similar pace twice running; if that continues, further price pressure would tell us little about physical balance. If liquidation stops and the price stays weak, the physical reading becomes harder to dismiss.

In China, the test is whether domestic ingot stocks keep falling once post-holiday restarts have worked through, and whether the liquid-aluminium proportion declines as SMM expects — which would add ingot supply independently of any change in demand. With smelters at 99.7% of effective capacity, any sustained domestic drawdown cannot be met by raising Chinese primary output.

Key Risks

Exchange inventory and metal availability have come apart. LME stock levels no longer provide a reliable read on metal the broad market would accept.

Chinese ingot inventory is partly a measure of production allocation. The proportion of liquid aluminium sent directly downstream affects how much metal enters the ingot series at all, independently of consumption.

Post-holiday destocking is seasonal. October's withdrawal follows a recurring annual pattern and an unusually steep pre-holiday draw.

Ownership concentration data are two months old. The 80 to 90% single-entity holding was a mid-August observation and is not established as the current position.

High smelter utilisation measures elasticity, not adequacy. Chinese output at 99.7% of effective capacity removes supply upside without demonstrating a deficit.

Guinea's export restriction is a proposal, not policy. No enacted quota has been established, and informed estimates of the eventual volume differ materially.

Intelligence Monitoring Points

• LME origin data: whether October's breakdown shows any non-Russian metal returning to warrant, from a base of 50 tonnes.
• Warrant concentration: whether a current single-entity holding percentage is published or reported.
• Spreads: whether the cash-to-three-month relationship establishes a direction, having spent October close to flat.
• Fund positioning: whether a third consecutive week of long liquidation arrives from +124,931.67 lots.
• Liquid aluminium proportion: whether October's expected decline materialises and lifts ingot casting.
• Chinese downstream: whether the recovery broadens beyond secondary aluminium.
• Guinea: whether a formal export decree is issued, and at what volume.

FAQ

Are LME aluminium stocks at a record low? They are low by historical standards. Reuters described stocks of roughly 250,000 tonnes in mid-August as the lowest since 1990, and opening stocks stood at 238,625 tonnes on 13 October. The more informative figure is that available stocks at end-September were 226,675 tonnes, of which 50 tonnes were not Russian-origin.

Why has the Russian share risen to almost 100%? Because other origins left rather than because Russian metal arrived. Russian tonnage fell 3,425 tonnes during September while Indian-origin available stocks fell 12,400 tonnes to 50 tonnes.

Is Russian metal on the LME unusable? It remains deliverable under exchange rules if produced before 12 April 2024. Market participants quoted by Reuters describe much of the remaining inventory as unwanted, which is a statement about commercial preference rather than eligibility.

Do falling Chinese inventories mean Chinese demand is strong? Not necessarily. Post-holiday destocking is seasonal, and SMM research has found that a high proportion of liquid aluminium going directly to downstream processing reduces the volume entering ingot inventories regardless of consumption.

Why is the price falling if supply is constrained? The analysis does not establish that it should not be. Fund liquidation from the complex's largest net long is one candidate, and Reuters reports that lost Gulf primary output has been absorbed through Indonesian production and Chinese semi-finished exports.

Data and source note: LME price, stock and warrant data are daily observations as reported for the dates stated, with the most recent being 13 October 2026; cash and three-month figures are bid quotations. LME country-of-origin figures are end-September 2026 available stocks from the exchange's monthly country-of-origin report, and describe on-warrant available metal rather than total warehouse inventory. The mid-August references to a 36-year low and single-entity warrant concentration are from Reuters reporting dated 14 August 2026 and describe conditions at that date. Chinese smelter utilisation is AZ Global's estimate for August 2026 as reported by Reuters on 9 September 2026. Chinese inventory, downstream utilisation and liquid-aluminium data are from Shanghai Metals Market, with weekly changes measured against 8 October and the liquid-aluminium proportion of 76.5% dated May 2026. Gulf production and global absorption figures are from Reuters, 14 October 2026, citing the International Aluminium Institute. Investment Fund positioning is from the LME MiFID II Weekly COTR, Investment Funds "Other" series, positions as at 9 October 2026. The cash-to-three-month spread observations for 13 and 14 October come from different sources on different dates and are reported as published rather than reconciled. Alcoa figures are from the company's third-quarter 2026 results released on 15 October 2026; the Massena East transaction terms and realised price figures are as disclosed, and no megawatt capacity figure appears in that disclosure. Guinean export figures are attributed to their respective sources, which differ; no enacted export quota has been established.

Sources

• London Metal Exchange — Country of Origin Stock Data, September 2026
• London Metal Exchange — Commitments of Traders, positions as at 9 October 2026
• Reuters — Aluminium on thin ice as LME stocks hit 36-year low, 14 August 2026
• Reuters — Russian aluminium's share of available LME stocks almost 100% in September (via Kitco), 14 October 2026
• Reuters — Aluminium's calm belies lingering supply-chain stress (via Mining Weekly), 14 October 2026
• Reuters — China's aluminium capacity ceiling and overseas expansion, 9 September 2026
• Reuters — Guinea bauxite export policy, 18 March and 28 April 2026
• Alcoa Corporation — Third Quarter 2026 Results, SEC Exhibit 99.1, 15 October 2026
• CNBC — William Oplinger interview, 15 October 2026
• Shanghai Metals Market — Chinese aluminium inventory, downstream utilisation survey and liquid aluminium research, June and 15–16 October 2026
• Bloodstone Research — Aluminum analysis, 3 September 2026