Managed money just delivered its biggest copper sale since April 2023. The Market Ear, citing Goldman Sachs positioning work associated with analyst Quinn, called it a 3.5-year selling extreme. Funds sold about $2.6 billion of high-grade copper futures in the week of Sept. 8 to 15. Liquidation accounted for roughly $2.2 billion of that. The December contract fell 5.6% from record highs.
Then the metal sat on its long-term trend line and its 100-day moving average. That is not how a broken bull market usually looks on the first pass. That is how a crowded long looks when it meets a tariff headline and an AI wobble on the same calendar.
The one theme for investors is simple. The puke was positioning. It was not, on the evidence so far, a funeral for copper demand. As long as data centers still get built and grids still get wired, a forced sale at support is a research event. It is not a signal to abandon the metal.
What the crowd actually did
Futures are a leverage machine. When the tape drops, margin and risk desks do the thinking for people who did not size the trade. A $2.2 billion liquidation print is that machine at work. It is not a new mine coming online. It is not Chile adding concentrate. It is funds getting smaller.
The reasons listed in the note were real enough to scare a desk. The White House had not decided whether to tariff refined copper. The pace of the AI buildout was under debate. Broader financial conditions tightened. Calls for a slower AI spend raised questions about data-center construction. Data centers are a structural source of copper demand. If that sleeve blinked, the long book blinked with it.
Goldman’s data-center equity basket fell about 6.9% in that unwind. Managed-money copper longs have tracked that basket. The copper sale was therefore, in part, an AI-positioning event. It was not, by itself, proof that kilowatts stopped needing cable.
By Monday, Sept. 21, the metal had already started to argue back. COMEX copper traded near $6.76 a pound, up about 1% on the session after a mid-month dip toward the mid-$6.30s. London quotes sat near $14,600 a tonne. Chinese physical buying and talk of refinery maintenance put a bid under the same market that funds had just lightened. The puke and the bounce can live in the same month. They often do.
Support is a fact. “The trend is dead” is an opinion
The second chart in the note is the one that matters for process. Price tagged the rising long-term line and the 100-day average at the same time the selling hit a multi-year extreme. Strong trends shake people off at those lines. Weak trends lose those lines and keep going.
Investors do not need a religion about the 100-day. They need a rule. If copper holds that band and the physical market stays tight, the mid-September dump was a washout. If copper loses the line and inventories swell, the dump was the first chapter of something worse. As of this week, the line is still there. Visible tightness is still in the conversation. That is the setup.
A 5.6% drop from a record is not small. On a $6.80 handle it is enough to wreck an over-levered futures book. It is not enough to rewrite a decade of mine-grade decline, grid buildout, and data-center power. Those are slow facts. Positioning is a fast fact. The opportunity is refusing to let the fast fact cancel the slow one without evidence.
AI did not stop using copper
Data centers need copper. So do transformers, interconnects, and the lines that move power from a gas plant or a hydro dam to a rack of chips. A 6.9% slide in a data-center equity basket can mark a valuation reset. It does not un-pour a slab or un-trench a cable already in the ground. It may delay the next campus. Delay is not zero. It is also not “the metal is done.”
The Market Ear’s close was blunt. As long as the AI trade remains alive, do not get too bearish on copper. That sentence is a filter, not a forecast. If capex guidance from the large cloud names keeps a multi-year power build in the model, copper’s demand story still has a sponsor. If those names cancel campuses in bunches, the sponsor leaves. Watch the build, not the tweet.
Tariff noise sits in the same inbox. A U.S. duty on refined copper can pull metal into American warehouses and starve other regions. That happened earlier in this rally. When the White House hesitates, the trade unwinds and price gives some of it back. That is a location squeeze. It is not a new pile of concentrate from a new pit. Investors who bought the squeeze as if it were geology learned a lesson in September. Investors who sold the geology because the squeeze unwound may learn the next one.
What the opportunity looks like in practice
The opportunity is not “buy every copper ticker at the open.” The opportunity is a sequence.
First, treat the mid-September sale as a positioning extreme until inventories and the trend line say otherwise. Extremes do not guarantee a bounce. They do tell you the easy long has already left.
Second, separate producers from stories. A mine that sells pounds this year still works if $6.70 holds. A junior that needs $8 and a bought deal works only if the tape is kind. Canadian copper stocks and other TSX names will move with the metal. They will move more than the metal. That torque is the feature after a flush. It is also the fine print if the line breaks.
Third, keep the AI link in the file, not on a shrine. Copper longs and data-center stocks have traveled together. When the basket heals, copper often heals with it. When the basket dies for a quarter, copper can look orphaned even if the smelter queue is real. Size for that correlation. Do not pretend it is destiny.
Fourth, watch the physical tells that funds cannot fake. Warehouse stocks, the nearby-versus-three-month spread, Chinese buying reports, and mine disruptions still settle arguments that a CFTC table only starts. A liquidation week plus rising inventories is bearish. A liquidation week plus tight stocks is how strong trends humble the late long and then continue.
How this week can prove the theme right or wrong
The proof is dull. If COMEX holds the mid-$6.60s to $6.70s while Fed-speak and tariff headlines land, the puke looks like a sale of the lows. If price loses the 100-day and the rising line with size, the crowd was early, not wrong, and the opportunity waits lower. Chinese demand into month-end and any White House word on refined-copper tariffs will do more than another opinion piece.
LME tightness and COMEX location remain two markets that can tell different stories on the same day. That split is how September already confused people. It will confuse them again. The investor who writes both prices in the notebook will be less surprised than the investor who only watches the fund-flow chart.
None of this is a promise that $7 returns this month. Records invite gravity. Gravity invited the $2.6 billion sale. The sale invited a question: was that the end of the move, or the bill for getting too long at the highs? The Market Ear’s charts argue for the second reading while AI capex is intact. That is the opportunity. Stay with the metal that still has to run the machines. Let the crowd keep the memory of the puke.
Disclaimer
Market commentary as of Sept. 21, 2026, based on The Market Ear’s positioning note and publicly reported futures and equity moves. Copper prices vary by venue and minute. This is not investment advice and not a recommendation to buy or sell copper futures, copper mining stocks, Canadian copper stocks, or any related ETF. Futures liquidation can reverse. Trends break. Do your own work. Past performance is not a guide to future results.