The selling in semiconductor shares wiped about a trillion dollars from the sector’s market value, and the money moving into single-purpose funds that magnify the same sector’s daily moves has not slowed down.
The VanEck Semiconductor exchange-traded fund, known by its ticker SMH and holding about 72 billion dollars, took in more than 430 million dollars over three trading days, according to data on the flows. That pace put the fund on course for its strongest week of inflows since the start of September.
Buyers went further out the risk curve as well. A daily three-times fund tracking semiconductors, the Direxion one, absorbed roughly 3.8 billion dollars even as its price fell nearly 21 percent during the week. That volume is enough to make the 19-billion-dollar fund’s weekly inflow a record.
The pattern is unusual because the funds in question are not designed to be held through a decline. Products that promise a multiple of an index’s daily return reset every day, which means the compounding works against holders in a choppy market. A fund that tracks three times the daily move of an index will lose more than three times as much over a period when the index falls and then rebounds only partway.
Buyers of these products generally know that. They are used for short holding periods, often days, sometimes hours, and the flow data captures entries and exits in the same product rather than a stable base of long-term holders. A record week can reflect conviction, or it can reflect a large number of traders taking positions they intend to close quickly.
Why chips specifically drew this response is a question about what investors think they are buying. Semiconductors have become the closest listed proxy for spending on artificial intelligence, and the equities have re-rated around expectations for data center construction. When the shares fall, the buyers who wanted exposure but missed the earlier run treat the decline as an entry point.
There is a case for that view. Demand for accelerators and the memory that goes with them has outrun supply for more than two years, and the companies that make them have reported record results in consecutive quarters. Capacity being built now will come online over the next several years, and the orders behind it are contracted.
There is also a case against assuming that the equity decline is a pricing error. Semiconductor shares had risen far enough that a modest change in expectations about future spending moves them a great deal, and a falling price is not by itself evidence that the underlying business has deteriorated or that it has not.
The mechanics of fund flows matter for what happens next. Inflow into an index fund produces purchases of the underlying shares, which supports the price. Inflow into a fund that uses derivatives to multiply exposure produces a different set of trades, directed at swap counterparties rather than at the shares themselves, and the fund’s own rebalancing adds to moves at the close.
That rebalancing is the part market participants watch. A triple-exposure fund must buy more exposure when the index rises and sell when it falls, in order to keep its daily multiple intact. In a falling market the process adds pressure, and in a sharp rebound it adds demand, which can make the last hour of trading in these names busier than the rest of the day.
Retail participation in single-sector funds has grown with the popularity of thematic investing. The largest technology and semiconductor products now rank among the most actively traded exchange-traded funds in the United States, measured by turnover rather than assets, and their holders skew toward accounts that trade frequently.
Concentration is the other feature of the trade. The semiconductor index is dominated by a handful of companies, and several of them are also among the largest positions in broad market funds. A decline in a few stocks therefore registers in portfolios that were not intended as bets on chips.
Analysts who cover fund flows said the inflows are best read as a measure of how crowded the long side of the trade had become, not as a forecast. Crowded positions unwind quickly when the buyers run out, and the money that entered in a record week can leave in a record week.
What the sector’s fundamentals will not do is settle quickly. New fabrication capacity takes years to build and longer to justify, and the buyers who are ordering equipment and memory today are working from demand forecasts that will be revised many times. The shares will trade on those revisions, and the funds that magnify the moves will keep magnifying them.
The suppliers themselves have lengthened the horizon on which they ask to be judged. Equipment makers book orders a year or more ahead, memory producers sign multi-quarter contracts, and foundries publish capacity plans that stretch past the current demand cycle. That makes the shares sensitive to any signal that a customer is slowing a buildout, and it makes the buyers of the dip dependent on a schedule they cannot observe directly.
None of this settles whether the sector’s shares are cheap or expensive after the decline. What the flows do show is that a large group of investors is willing to take the same exposure through an instrument designed for days rather than years, at a moment when the underlying names are falling and the demand they depend on is being reassessed.
