Why Copper Is Crashing Despite a Historic Supply Squeeze
Copper plummeted over 5 percent even as warehouse inventories hit multi-year lows—a signal that the AI infrastructure boom may be running out of steam before it gets going.
The supply squeeze was supposed to lift everything.
Warehouse inventories in London and Shanghai are evaporating. Copper and aluminum were deep inside historic squeezes—the kind of market conditions that usually force bears to cover and send prices higher. Instead, copper fell 5.6 percent on the day, erasing April gains and leaving the metal well below its recent peak. Tin, another AI-critical component used in solder for semiconductor manufacturing, dropped 8.2 percent. The rally built on promises of unlimited data-center demand turned out to be far less durable than anyone priced in at the beginning of the year.
This is the rarest form of breakdown in commodity markets: prices falling while physical scarcity climbs. When the chart shows one direction and the warehouse receipts show the other, it means the dominant buyers have stopped believing their own story. And that disconnect has sent shockwaves through portfolios that had long treated the AI-copper thesis as near-certain.
Who got burned—and who got nervous
The sharpest pivot came from Stan Druckenmiller, the legendary sovereign-wealth investor whose long copper stance during the AI capital-expenditure cycle became shorthand for the bull case. On March 17 he signaled he had turned bearish, a move that echoed across desks from New York to Singapore. His reversal carried outsized weight because it forced a reassessment among other momentum-driven investors who had piled into copper on the same narrative. Within days, speculative positioning that had sat near record bullish levels started unwinding, and in some cases flipped outright short.
That kind of cascade is dangerous precisely because it moves faster than fundamentals can justify. Trend followers amplify moves—they bought when the narrative held, and they sold when it fractured. Their exits did not come from a revision of the physics of supply and demand. They came from a shift in conviction about whether AI data-center buildouts would arrive fast enough to sustain the premium. When conviction breaks, it does so faster than the physical market can prove anyone wrong.
Broader hedge-fund flows told the same story. CFTC data showed commercial shorts increasing their net positions while managed-money longs declined sharply. Even fund managers who had not participated in the initial copper rally found themselves caught in margin calls triggered by correlated selling across metals. The tin selloff, which accelerated after the copper move, pulled in traders who had been quietly accumulating base-metal exposure on the assumption that AI capex would eventually lift all conductive metals together.
Why the AI story lost grip
Ryan McKay at TD Securities summarized the pressure on three fronts. First, Federal Reserve tightening concerns remain a drag on near-term demand forecasts, particularly for rate-sensitive sectors such as construction and automotive that also consume copper heavily. Second, risk appetite around AI spending has cooled since the midterms displaced the most optimistic projections for corporate capex commitments. Major technology firms that had publicly pledged hundreds of billions toward data-center expansion began signaling delays or scaling back ambitious timelines. And third, trader bets on U.S. tariffs on refined copper—once a widely traded scenario that inflated cross-exchange differentials—have largely faded, compressing the spread between COMEX and LME prices and removing a key speculative engine that had propped up the front end of the curve.
Each factor individually would weigh on sentiment. Together they form a complete exit ramp. The tariff play was arguably the most speculative leg of the entire copper-AI thesis. It relied on a policy bet that neither Congress nor global trading partners have shown enthusiasm for enacting in a form that would meaningfully raise landed costs for U.S. buyers. When that assumption dissolved, the marginal bullish position vanished first, leaving the rest of the trade exposed without its thinnest but most leveraged support.
Beyond the macro drivers, a quieter question emerged in research notes from Bloomberg Intelligence and S&P Global Commodity Insights: how much of the projected copper demand actually rests on installations that exist only on PowerPoint slides? Several utility companies in the Southwest United States reported that grid interconnection queues remain clogged, pushing data-center power-connection timelines well beyond what earlier analyst models assumed. Without firm transmission capacity, even funded projects stall. That realization undermined the velocity argument that had justified buying copper at elevated premiums—because if the demand arrives slower than expected, the physical shortage matters less in the near term regardless of how low inventories sit today.
The contradiction that matters
Supply crunches punish sellers more than buyers—if they know the crunch exists. The paradox this week is that everyone knew the crunch existed. LME and Shanghai registered inventories dropped to levels that normally trigger defensive buying, margin calls on shorts, and volatile upward spikes. Spot premiums ran ahead of futures curves. Yet selling accelerated anyway.
That tells you two things. One, the bulk of the open interest was never anchored in the physical market; it was placed in expectations about a future that looked increasingly distant. Two, once the first wave of liquidation began, hedgers and programmatic desks amplified each other’s exits regardless of warehouse counts. Futures become decoupled from spot when leverage dominates position sizing. What followed was a textbook deleveraging sequence, not a re-rating of the long-term demand curve.
nThe behavior resembled previous commodity crashes where financialization overwhelmed physical realities—notably the crude-oil collapse of early 2020 and the nickel implosion at LME in 2022. In both cases, prices diverged sharply from available inventory precisely because the market had accumulated far more paper than metal. Traders who understand those episodes see red flags immediately. Those who did not simply rode the train until the tracks ran out.
What happens next
The immediate damage is concentrated in the leveraged longs. Funds that entered near the peak have already absorbed losses exceeding 10 percent from those levels. If the next round of economic data reinforces fears of Fed restraint or softens Chinese industrial demand further—Beijing’s property sector remains structurally weak and a significant copper sink—there will be no floor until the speculative overextension fully clears. But the physical squeeze has not disappeared. Inventory draws in London and Shanghai remain visible, and miners such as Codelco in Chile and BHP in Australia still face genuine constraints on volumes they can bring to market this year, from operational disruptions to labor negotiations in Peru.
Which path wins depends on speed. If AI-driven demand materializes within the next six to twelve months—as project pipelines suggest and several hyperscalers continue defending privately—it will eventually collide with these low inventories and restart the bullish setup. If it stalls or slows, the current selloff becomes a lasting repricing rather than a brief capitulation. The market cannot simultaneously hold the story intact and refuse to pay for it. Somewhere along the timeline, cash must precede conviction, and right now cash is exiting the copper complex.
Second-order effects are already visible. Refiners in China, which rely heavily on imported concentrate, are reporting tighter treatment and refining charge negotiations for the second half of the year. Smelters facing potential shortfall-driven curtailments could tighten downstream supply for wire and cable manufacturers, creating a feedback loop that eventually reaches end users. Mining equities are pricing in the possibility that even a recovered metal price fails to translate into higher production or better margins if off-take agreements renegotiate downward under the weight of weaker near-term demand. Investors should watch those refiner charges closely—they are an early indicator of whether the physical market is catching up to the paper collapse.
The broader implication
Every mining stock with significant copper exposure—Glencore, Freeport-McMoRan, Southern Copper, First Quantum—now faces a simpler test. The thesis that powered their rallies over the past two years required two assumptions: that copper stays scarce and that AI demand arrives on schedule. This week proved neither assumption alone is enough to defend speculative excess. Scarcity without timely demand sustains pain for overleveraged bulls but does not guarantee price recovery. Timely demand without scarcity invites disappointment rather than vindication.
The infrastructure supercycle narrative has not died, but it has aged overnight. What the market discovered this week is that narratives which accelerate too quickly tend to deflate just as fast. Investors betting on copper as a direct proxy for AI proliferation assumed the story would sustain itself long enough for the fundamentals to catch up. That bet has not died—it simply needs time to prove or disprove. What died this week was the impatience of traders who mistook a conviction for a certainty. The physical market will ultimately settle the question. The financial market, having learned an expensive lesson, will wait patiently on the sidelines.