A November 6 study by Dune, the on-chain analytics platform, tracked Hyperliquid's stock perpetuals through November 4-5, the two trading sessions after the US midterm vote, and found traders building chip exposure while the underlying stocks fell and funding rates climbed. Across US chip and memory perpetuals, contracts held — the venue's open interest — rose 60% over the two days while the underlying stocks lost between 12% and 28%. Funding hit roughly five times its resting level, and contracts held in megacap technology names dropped 27%. Stock perpetuals put equity exposure on a crypto trading venue, so a Wall Street selloff leaves an hour-by-hour record in the venue's funding and open-interest data.

A political trigger for the selloff

The selloff had a political trigger. Democrats' sweeping midterm wins raised concern on Wall Street that the AI data-center buildout could slow, and chip and data-center stocks slipped on November 4 even as the wider market rallied. On November 5, CNBC reported that the Nasdaq fell 5.5%, its worst day since April 2025. Nvidia and Qualcomm each lost more than 9%, and AMD, Micron and Applied Materials posted double-digit declines. Qualcomm also carried the weight of its own disappointing outlook.

Dune's figures show where Hyperliquid traders stood during that drop. The measure counts contracts rather than dollar value, so falling prices cannot explain the increase: traders opened new positions into the decline.

The buying was not spread evenly across AI-linked names. Contracts held in megacap technology names fell 27%, even though the large cloud providers' shares lost only 2-4% on November 5. Vistra, a power producer whose shares trade on expectations of data-center electricity demand, saw its contracts drop 29%. Traders cut positions where prices held up best and built them where the losses were deepest.

Funding shows who was paying

A rising contract count shows more positions, but not which side opened them, because every perpetual has a long and a short. Funding fills that gap. Under Hyperliquid's funding rules, the rate turns positive when the contract trades above the oracle price taken from the underlying market, and longs then pay shorts every hour; when it trades below, the rate turns negative and shorts pay longs. That hourly payment is the tether that keeps a perpetual aligned with the share it tracks.

Funding at about five times its resting level during a 12-28% decline means buyers were bidding the perpetuals above the reference price as the stocks fell. That signals long demand outrunning the shorts willing to meet it — dip buying by another name.

The strongest objection is that some of those longs were not rebound bets. A trader short Nvidia shares elsewhere, or short a Nasdaq contract, could hold a chip perpetual as a hedge. Dune's aggregate figures cannot separate those accounts, and they do not show entry prices, collateral or positions on other venues. The reading holds for the market as a whole, not for every trader in it.

What waiting for a rebound costs

A perpetual tracks a share price with no expiry date and no ownership of the company; Coindoo's earlier look at stocks and commodities on Hyperliquid covers how the venue supplies that exposure. In place of an expiry, the contract settles funding hourly between traders. Hyperliquid takes no fee on those payments. Each payment equals position size times the oracle price times the rate, so the bill scales with the exposure held, not with the margin posted.

Hyperliquid's default interest component is 0.01% every eight hours, or 0.00125% an hour, about 11.6% a year. Markets launched by outside builders can set their own interest rate and funding multiplier, and most of Hyperliquid's equity exposure sits in those markets. If a chip market's resting level matched the default, five times it would be roughly 0.006% an hour, close to 55% annualized.

What that rate costs a 10x long on a $10,000 position:

The table assumes $1,000 of margin, funding held at the rate above, and no trading fees or price change.

None of that cost depends on the stock falling further. A position that goes sideways still pays every hour, and a falling price shrinks the margin that absorbs those payments. For an unhedged long, the rebound had to be large enough to cover the funding bill and arrive soon enough that the account still had the margin to stay open.

Early buyers faced most of the fall

Dune also measured how much of each day's move the perpetuals had priced by 14:00 UTC. On November 4, they already pointed toward the eventual close for 15 of 16 stocks and had absorbed about two-thirds of the median daily move. On November 5, the day of the steepest losses, they had absorbed only about a fifth of that day's decline.

That checkpoint sits half an hour before the regular US open, which moved to 14:30 UTC when US clocks changed on November 1. Premarket trading was already running, so the perpetuals were not the only market adjusting, and only a side-by-side comparison with premarket quotes could show whether they led.

For the rebound trade, the November 5 figure carries the weight. Dune's data does not show when the new contracts were opened. Any opened around that hour, though, still had roughly four-fifths of the day's drop ahead of them and paid elevated funding throughout. An early price is not the same as a good entry.

Two signals that separate holding from unwinding

Two patterns in the hourly funding and open interest data for the chip and memory markets would show which way the trade is resolving over the coming week.

Longs holding. Funding falls back toward its resting level while contracts held stay near their new high. Longs are keeping their exposure at a lower hourly cost, either because shorts have stepped in or because the premium over the oracle price has faded.

Longs unwinding. Contracts held fall as liquidations rise in the same markets, a sign that margin is running out before any rebound. Liquidation figures are the cleaner check, since new positions can mask forced closures in the aggregate contract count.

Until one of those patterns appears, the 60% increase records a bet that traders placed and paid to keep — not one that has paid them.