|
Part I
The Mechanism
|
Tin.
AFP ran the headline last month: Myanmar tin mega-mine restarts operations. The sell-side picked it up within hours. Supply relief. Price ceiling forming. The tin rally is “mature.” If you’re reading this letter, you already know the next sentence: they looked at the headline and stopped reading.
Man Maw is not back. The International Crisis Group called it a “slow restart.” Output is still well below pre-suspension levels. Ore is moving again, sure — in the way a garden hose moves water after a three-year shutoff. The United Wa State Army closed that mine in August 2023, and the infrastructure doesn’t just switch on because someone filed a permit. Meanwhile, combined LME and Shanghai Futures Exchange tin stocks have fallen from 22,600 tonnes in February to 13,100 tonnes. That decline happened while the Man Maw “restart” was being priced as solved.
Here’s what the restart narrative misses entirely: tin’s demand base shifted underneath while everyone was watching the mine. Solder accounts for more than half of all tin consumed globally. Every GPU, every printed circuit board, every server rack in every AI data center Nvidia and AMD keep shipping — soldered with tin. Shanghai Metals Market estimates each gigawatt of installed AI data-center capacity requires 1,200 to 1,500 tonnes of tin. That’s more than three times a traditional data center. There are 190 gigawatts of new hyperscale AI capacity announced as of early 2026.
Nobody on the tin desk talks about AI. Nobody on the AI desk talks about tin. That gap is the whole story.
|
Part II
The Diagram
|
Story off. Diagram on.
Global tin mine production fell to roughly 290,000 tonnes in 2025, down from 294,000 in 2024. Refined production is expected to grow about 3% in 2026. Demand growth: 3.5%. That half-point gap doesn’t sound like much until you realize tin is the smallest of the LME base metals by volume. The buffer between balance and deficit is measured in weeks of consumption, not months.
The supply side is a list of disruptions stacked on top of each other. Myanmar’s Wa State — roughly 10% of global concentrate — shut for three years and now trickling. Indonesia, the world’s largest exporter of refined tin, has been throttling export permits since 2024 while its energy minister publicly confirmed in February 2026 that Jakarta is reviewing a full ban on tin ingot exports. The DRC’s Bisie mine, 6% of global supply, suspended operations for a month in 2025 over M23 rebel advances and only hit 10,039 tonnes in H1 2026. Three of the top five producing jurisdictions disrupted simultaneously.
On the demand side, the math is getting violent. A traditional server uses about 500 grams of tin. An AI server needs 4 to 5 kilograms. A 10,000-card AI computing cluster requires 2.5 to 3.2 tonnes of PCB solder alone. And that’s before you count the power management systems, the cooling infrastructure, the networking equipment. Every hyperscaler capex announcement is a tin purchase order nobody modeled.
This is the smallest LME market trying to absorb the largest infrastructure buildout in a generation. The numbers don’t balance. They haven’t balanced all year. The exchange just hasn’t noticed because tin doesn’t get copper’s headline space.
|
Part III
The Weak Link
|
Indonesia.
In February 2026, Energy Minister Bahlil Lahadalia stood at the Indonesia Economic Outlook in Jakarta and said the quiet part out loud: the government is reviewing a halt to tin exports. “Everything must go into industrial downstreaming.” Indonesia exported 52,416 tonnes of refined tin ingots in 2025. That’s roughly 15% of global refined supply flowing through one policy decision.
If you watched the nickel playbook, you know what comes next. Jakarta banned raw nickel ore exports in 2020, forced downstream investment, and remade the global nickel supply chain in its image. Bauxite went the same way. Tin is the next domino, and the minister named it explicitly. The only question is timing.
Meanwhile, the paper market is doing what paper markets do when they can’t see the loading docks. LME three-month tin hit $54,184 last week. That’s high by historical standards. But BMI just raised its 2026 average forecast to $51,000 from $49,000, and Fastmarkets is modeling AI-driven solder demand more than doubling to 23,640 tonnes by 2030. The price is “high” only if you’re using last cycle’s demand model.
I remember when copper at $10,000 looked expensive. Then the electrification math caught up. Tin at $54,000 looks expensive if you think it’s a packaging metal. It looks cheap if you realize it’s become the metallic glue holding the AI buildout together, and three of its five major supply sources are either broken, throttled, or actively threatening to shut the door.
Tin is the only LME base metal where visible exchange inventory covers less than two weeks of global consumption. Two weeks. Every other base metal has months. That’s not a buffer. That’s a rounding error with a warehouse label on it.
|
Part IV
The Chain Reaction
|
The sequence here is mechanical, and it runs faster in tin than in any other base metal because the market is so thin.
The trigger doesn’t have to be Indonesia. It could be another Wa State suspension — that mine runs on the goodwill of an ethnic army, not a regulatory framework. It could be a DRC flare-up that shuts Bisie’s export corridor again. It could be something as mundane as a single large physical delivery failing on the LME. In a market this thin, the trigger almost doesn’t matter. What matters is the stock level when it hits.
Tin did this in January 2026 — spiked to a record $56,800 on Indonesian export constraints, Myanmar feedstock shortages, and surging AI solder demand. The cash-to-three-month spread, which was sitting at over $400 in contango in July, has already narrowed to $105. That compression is the physical market whispering what the headlines haven’t said yet.
Where does the capital go? Not into the diversified miners where tin is a footnote on page 47 of the annual report. The edge — if there is one — is in pure-play tin exposure outside the disruption zones. Alphamin’s Bisie operation in the DRC is the third-largest industrial tin mine in the world, targeting 20,000 tonnes for 2026, but it sits inside an M23 conflict zone and runs on security agreements that could evaporate overnight. Elementos and its Oropesa project in Spain, or Nathan Trotter’s new Tin Ridge smelting and recycling facility in Virginia — these are the kind of assets that reprice when the market realizes the three traditional supply pillars are all cracked at once.
The financial layer sees a metal at an all-time-high range and calls it expensive. The physical layer sees 13,100 tonnes of visible inventory trying to supply a 290,000-tonne-per-year market that’s growing at 3.5% into an AI buildout nobody put in the model. In my experience, when those two layers disagree, the physical layer wins. It just takes longer than the sell-side’s attention span.
Sources: International Crisis Group, LME, Shanghai Futures Exchange, Shanghai Metals Market, Fastmarkets, Coface, BMI, AFP, International Tin Association, Alphamin Resources
