Copper's AI Trade Just Cracked. The $127 Billion Wipeout Is a Positioning Story.
Comex copper futures dropped more than 9% this week, the metal's worst stretch of 2026, and eleven major miners gave up roughly $127 billion in market value. The AI data center thesis didn't break. The crowd around it did.
I spent Wednesday afternoon staring at the copper tape, and the thing that struck me wasn't the size of the drop. It was the silence. Bids didn't thin out. They disappeared.
Comex copper futures closed the week down more than 9%. That's the worst week of 2026, and it lands roughly three months after the metal printed a record high on a very clean story: AI data centers need wire.
Let me break this down.
What Actually Broke
The September rally had one engine. Every gigawatt of AI compute gets built out of copper, from the busbars to the transformers to the miles of cable running behind the racks. That's a real demand curve, and I don't think anyone serious disputes it.
But the positioning around that curve got ugly. Hedge funds piled into the long side until copper stopped trading like an industrial metal and started trading like a momentum name. You could see it in the Comex premium over London, in the options skew, in the way every dip got bought inside of two sessions.
This week, they cut most of their bets. Not some. Most.
The numbers tell the story. Eleven of the biggest copper miners shed about $127 billion in market value in a handful of days. That's not a supply story. Nobody discovered a new deposit or found a substitute for wire. That's a crowd trying to leave through the same door at the same time.
Here's what matters: the metal fell because conviction was borrowed, not owned.
The Wider Read-Through
Pull the camera back and this stops being about copper. It's the second AI-adjacent trade this year that unwound on flow rather than fundamentals, and that pattern deserves attention.
Copper is a physical market first and a futures market second. A 9% paper move doesn't change how much ore comes out of Chile next quarter. What it does change is the cost of capital for the next expansion. Miners price new projects off long-run price assumptions and the debt behind them. Wobble the futures curve, and boards get cautious about greenfield spend right when supply needs it most.
There's a macro read too. Copper is a growth proxy before it's anything else. When it breaks this hard, people start asking what it says about industrial demand outside AI. Frankly, some of that question is fair. Chinese property isn't fixed. European manufacturing isn't humming. The AI wire story was carrying the whole bid, and now it's carrying a lot less.
My Honest Take
I think the AI wiring demand is real. I also think this selloff is mostly a positioning event, not a thesis event. Both can be true, and the second one matters more for your P&L.
The reality is that a record high built on a story you can't verify quarter to quarter is a record high built on flow. When the flow stops, price doesn't care how good your narrative is. So if copper is down 9% in a week while data center construction is still running, what does that tell you about how much of the September rally was actually about the buildout?
That's the question I'd sit with over the weekend.
What I'm watching next: miner guidance for 2027 capex, exchange inventory prints, and whether the Comex premium normalizes or stays wide. If inventories stay tight through the next reporting cycle, the thesis is intact and this was a washout. If they build, the crowd was wrong about more than timing.
From a risk perspective, keep your exposure small enough that a 9% week doesn't force a decision. That's the whole lesson.
Related Articles
Key Terms Explained
A marketplace where cryptocurrencies are bought and sold.
Contracts to buy or sell an asset at a specific price on a future date.
Taking a position that offsets potential losses in another investment.
Contracts giving the right, but not obligation, to buy (call) or sell (put) an asset at a set price before expiration.