|
Part I
The Mechanism
|
Aluminium.
Bloomberg's commodities desk is running stories about aluminium's "retreat." CNBC had a guy last week saying the Gulf supply shock was "largely priced out." If you only look at the LME three-month price — sitting right around $3,000 per tonne, exactly where it was in early February before the first missiles hit — you'd think nothing happened. Three million tonnes of annual capacity vaporized, and the price yawned.
That's the headline. Here's the engine room.
Read that last number again. Nearly every tonne of aluminium sitting in an LME warehouse right now is Russian. And most Western buyers won't touch it. The non-Russian, deliverable stock on the LME — the metal that actually functions as available supply for the buyers who need it — is roughly 50 tonnes. That's not a typo. Fifty tonnes. In a market that consumes 70 million a year.
The shortage didn't disappear. It migrated. It left the LME screen and moved into regional premiums, where the terminals aren't watching and the headlines don't reach. The price is lying. The warehouses are telling the truth.
Just a few years ago, they laughed at him…
And said that what Elon was planning to do was impossible.
In 2018, the New York Post even called Elon “a total fraud”...
And it proves the media was 100% wrong about Elon.
|
Part II
The Diagram
|
Story off. Numbers on.
Iranian strikes on March 28 hit EGA's Al Taweelah smelter in Abu Dhabi and Alba's plant in Bahrain. Qatalum lost its gas supply when Ras Laffan went dark. Between February and May, Gulf primary aluminium output fell by an annualized 2.8 million tonnes, according to the International Aluminium Institute. The GCC had produced 6.16 million tonnes in 2025 — over 80% of it exported. That pipe got crimped in a single week.
China and Indonesia stepped in. Chinese alloy exports doubled year-on-year to 349,000 tonnes in the first eight months. Semi-finished product exports rose 774,000 tonnes to 4.46 million tonnes. Indonesia shipped a record 114,000 tonnes of primary metal in July alone, with operating capacity jumping from 870,000 to roughly 2.5 million tonnes during 2026.
So why are LME stocks still draining? Because the replacement metal isn't exchange-grade ingot. It's alloy. It's semi-finished product. It can go into a rolling mill, but it can't go on an LME warrant. Consumers who need ingot bid through regional premiums instead. That's the migration.
Here's what the premium structure actually looks like:
A US manufacturer sees $5,370. A Japanese buyer sees $3,310. The LME screen shows $3,055. Same metal. Three completely different prices. The Midwest premium alone — $2,312 per tonne — is 76% of the LME base price. I've never seen a regional delivery premium that large relative to the underlying. Anyone quoting "LME aluminium" as though it describes the cost of buying the metal in any specific geography is reading the wrong number.
|
Part III
The Weak Link
|
Everyone is watching the Gulf restarts. Nobody is watching Guinea.
Guinea produces 34% of global bauxite — 183 million tonnes in 2025. China sources nearly 80% of its bauxite imports from Guinean mines. The squeeze started in 2025, when Conakry raised export tariffs to 10%, mandated in July that 50% of cargo move on state-owned vessels, and revoked the mining license of Guinea Alumina Corporation — an EGA subsidiary — for failing to build a domestic refinery. Then on June 24 this year, they capped annual bauxite exports at 150 million tonnes, an 18% cut from the prior year. The loading docks in Boké are getting quieter.
I've watched enough supply "recovery" timelines in metals to know how this works. EGA says Q1 2027 for full restart — they're removing solidified metal from individual cells, one at a time, and each restarted cell takes a week to stabilize. I'd take the over on that timeline. Alba's lines 1 through 3 are still offline with no public restart date. Qatalum is running on whatever gas Qatar Energy feels like sending.
Indonesia is the wildcard, and people are overweighting it. On paper, Indonesian nameplate capacity jumps from 870,000 to 2.5 million tonnes in 2026. In practice, aluminium smelting requires 13 to 16 megawatt-hours per tonne of uninterrupted power. Indonesia's grid doesn't deliver that reliably. One analyst called 2027 a "power commissioning forecast" rather than an aluminium production forecast. That's exactly right. The metal doesn't flow until the electricity does. Nameplate capacity is a line on a slide deck. Power delivery is the constraint that actually binds.
Meanwhile, China's own production ceiling — 45 million tonnes, imposed in 2017 — is already being breached, with 2026 output projected around 46 million tonnes. Beijing doesn't have another gear to shift into. And Guinea just tightened the bauxite valve feeding the refineries that feed the smelters that are already running over capacity.
This is the gap. The financial layer reads the LME headline and models a return to normalcy. The physical layer — the buyer in Rotterdam paying $3,614 all-in, the manufacturer in Chicago paying $5,370 — is living in a different market entirely. I've been on the wrong side of exactly this kind of paper-physical divergence before, betting on the screen price when the dock price was the one that mattered. The dock wins. It just takes time.
|
Part IV
The Chain Reaction
|
Two paths. The trigger for each is known. Only the timing isn't.
The reversal signal is simple: LME registered stocks climbing back above 300,000 tonnes. That level was last held as a monthly average in June. If stocks rebuild, premiums compress, and the trade dies. If they don't — and the mechanics right now suggest they won't, not with Russian metal comprising the entire buffer and Western buyers refusing to draw on it — the premium dislocation deepens.
Alcoa's Aluminum segment just posted a record 32.3% adjusted EBITDA margin in Q2. That margin is built on cheap alumina — their own refineries feeding their own smelters — and the elevated premium environment that has persisted since March. They're also closing the South32 AliGroup acquisition, approved by shareholders on October 15, adding more refining tonnage to their integrated chain. A smelter that buys alumina on the spot market and sells metal at the LME flat price is a different business than one that mines its own bauxite, refines its own alumina, and sells into a $559 European premium. The vertical integration advantage hasn't been this wide since the early 2000s.
Where capital doesn't go: broad industrial metal ETFs that average the LME price across a dozen base metals and carry no premium exposure. Those instruments don't participate in the dislocation. They track a number that, as of this writing, understates the physical cost of aluminium in every Western market by 18% to 76%.
The screen says $3,000. The dock says $5,370. The market of last resort is full of metal nobody wants. The 2027 surplus everyone is modelling requires three miracles to arrive on time. In my experience, banking on three simultaneous miracles in metals is a strategy with a known outcome. I wouldn't call it a good one.
Sources: LME warehouse data, International Aluminium Institute, Reuters, Westmetall, SMM, Macquarie, Fastmarkets, Al Circle, Guinea Ministry of Mines and Geology
