Tearing apart the tin solder supply chain

Tin.

Everyone with a microphone is talking about the AI supply chain. GPUs. Power. Cooling. Land rights for data centers. CNBC runs three segments a day on semiconductor capacity. Bloomberg has a live tracker for Nvidia's backlog. Not one of them, to my knowledge, has said the word "solder" on air this quarter.

Here's what they're missing. Every GPU, every server rack, every networking switch, every power module in every data center on earth is held together by tin-based solder. Roughly half of all refined tin consumed globally goes into solder. And a single AI server uses more than three times the tin of a conventional one. The chip gets the headline. The metal that physically connects it to the board doesn't exist in the narrative.

Tin is up nearly 58% year-over-year and nobody on the mainstream side can tell you why. The sell-side models still bucket it under "minor metals." The AI analysts don't look below the chip. The commodity desks treat it as too illiquid to matter. That's a lot of people not watching a $53,000-a-tonne metal that just hit an all-time high five months ago.

I've traded around small metals before. The pattern is always the same: they're boring until they're not, and by the time they're not, you're already late. This one feels like that moment.

Story off. Here's the schematic.

Global refined tin consumption runs around 378,900 tonnes per year. Refined production growth is tracking roughly 3% annually. Demand growth is running at 3.5% and accelerating. Coface projects the market shifts into outright deficit in 2026 — the first since 2021. The International Tin Association projects a 13,000-tonne structural deficit by 2030 without significant new mine investment. No significant new mine investment is happening.

The demand side isn't guesswork. CITIC Securities estimates AI data servers alone add approximately 2,500 tonnes of new tin consumption this year. By 2030, that figure scales to 22,000–25,000 tonnes annually — a nearly tenfold increase. Every rack, every PCB layer, every high-density GPU interconnect needs lead-free solder joints that only tin-based alloys deliver at scale. RoHS mandates across the EU, Japan, and China drove substitution toward tin-based lead-free solder. There's no Plan B metal.

The supply side is where the diagram turns red. Three countries — Indonesia, Myanmar, and the DRC — collectively control a chokepoint that feeds half the world's smelting capacity. Here's the flow:

China refines half the world's tin but imports roughly two-thirds of the ore it needs to do it. Myanmar and the DRC together supply 60% of those Chinese ore imports and together represent about 20% of global mine production. Both are active conflict zones. This is the AI supply chain's most fragile link, and it doesn't show up in any of the earnings call decks.

The pipeline for new mines? Critically limited. Few large-scale developments are expected before 2028 to 2030. This is not a problem you can throw capex at and fix next quarter. The geology doesn't care about your buildout timeline.

Three supply valves. All three are stuck.

Indonesia — the world's largest tin exporter — is dismantling itself on purpose. President Subianto ordered the closure of 1,000 illegal mines on Bangka and Belitung islands. A military-backed task force has been sweeping the region: offshore dredging vessels seized, unlicensed smelters shut down, 500 tonnes of tin confiscated. The crackdown could remove up to 80% of Bangka-Belitung's output from global supply. That's not a disruption. That's an amputation.

Myanmar's Man Maw mine — once a cornerstone of Chinese smelter feedstock — has technically reopened after its 2023 resource audit shutdown. Wa State authorities are accepting license applications again. But exports are averaging only about 1,300 tonnes per month during recovery, well below historical levels, and the country's civil conflict makes reliable throughput a coin flip. The dewatering cost-sharing arrangement announced in February tells you everything: they're still pumping water out of the shafts. This is not a mine running at capacity.

And in the DRC, M23 rebel forces continue to disrupt operations in North Kivu, including at the Bisie complex — Alphamin's high-grade tin mine producing near 20,000 tonnes annually. That's roughly 6% of global output sitting in an active warzone.

Meanwhile, the speculative layer is doing what it always does — misreading the room. Chinese investors drove SHFE tin volumes to daily records earlier this year. Prices spiked to $56,800 in January, pulled back on profit-taking, and now most of the momentum chasers think the move is over. LME inventories sit around 8,000 tonnes — down slightly over the last three months and stuck near historic lows. The headline says "stocks stable." But stability at a historically thin base doesn't mean the shelves are full. It means you're still at almost nothing.

I've watched this pattern in other metals. The low-inventory equilibrium creates a narrative window — "supply is normalizing" — that lasts exactly until the next physical delivery contract fails. The paper market exhales. The physical market hasn't taken a breath.

The entire AI buildout assumption — 190 GW of announced data center capacity — is priced as if the physical materials just show up. Copper gets discussed. Lithium gets discussed. Tin? It's like building a skyscraper and forgetting to budget for the rivets.

The sequence from here is mechanical. Not guaranteed — but mechanical. The same way a ratchet turns one way.

The first domino is feedstock. China's smelters import two-thirds of their ore. If Indonesia tightens further — and Subianto's resource nationalism shows no sign of softening — or if M23 advances disrupt Bisie again, ore premia spike. Chinese smelters can't just flip a switch and source elsewhere. There is no elsewhere. Not at this volume. Not at this grade.

Second domino: refined output slows. LME nearby spreads — already elevated — blow wider. The inventory "rebuild" everyone pointed to as evidence of normalization reverses. It doesn't take much. Tin is the smallest of the LME-traded base metals by volume. Thin markets move violently.

Third domino: it hits the AI supply chain directly. Solder alloy producers who supply the electronics OEMs building servers start competing for a shrinking pool of physical metal. Component costs inflate. Data center buildout timelines slip — not because of GPUs or power, but because the solder didn't show up. Nobody on the capex calls is modeling for this. Nobody ever models for the thing they haven't heard of.

Where does capital go? Not into the broad mining ETFs — tin is too small a share of their holdings to move the needle. The edge, if there is one, sits with pure-play and near-pure-play tin producers operating in stable jurisdictions with high-grade assets. Companies like Alphamin (DRC risk but exceptional geology, ~20,000t output) or junior developers in politically stable corridors with offtake deals already signed to electronics OEMs and hyperscalers. The ones positioning for traceable, non-Chinese refined supply. That's the premium the market hasn't priced yet.

The secondary play is recycling and urban mining, currently under-scaled but improving as economics tighten. It's a bridge, not a solution. But bridges have value when the river is rising.

The structural case here is not a price spike. It's a re-rating. Tin spent decades as a "minor metal." The AI buildout just made it a critical material — only nobody updated the label. The narrative will catch up. It always does. The question, as with every structural trade, is whether you want to be positioned before or after the Financial Times runs the feature. In my experience, "after" is more comfortable and a lot less profitable.