MegacapearningsonHyperliquidandPolymarket

Figures as of 31 October 2026 unless stated.

Five of the seven largest US tech companies have reported Q3 results, each after the 4pm ET close, so the stocks’ main reaction came at the next morning’s open. Their perpetual futures on Hyperliquid traded through the night. Across the 26 megacap reports this year, the perp had priced a median of about 70% of the next opening gap by 8pm ET, when US after-hours trading ends, and pointed the right way in 19 of the 21 reports with a gap of more than 2%. It changed little between 8pm and 4am and covered most of the remaining gap after US premarket trading opened, reaching about 90% by 9am.

The Q3 reports mostly followed this pattern. Alphabet’s perp was 11.4% lower at 9am on 29 October, and the stock opened 11.7% lower. Amazon’s perp was 7.4% lower at 9am the following day, while the stock opened 4.6% lower and closed down 1.4%, the largest overshoot this year. Meta’s perp was still 2.1% higher at 8pm on 21 October, and the stock opened 3.0% lower. Across all 26 reports, the perp pointed the wrong way an hour after the release in 8 cases.

In the week before each Q3 report, volume on Hyperliquid stayed within its usual range, with the weekday and weekend pattern of any other week and a modest rise on the day before. Tesla was the exception, with two days of heavy trading that appear unrelated to its report. The perp’s price over that week gave no better hint of the result: across all 26 reports, it drifted in the direction of the eventual move in 12 cases, about what chance would produce. In the 24 hours after the release, volume reached 7–11 times the average day of two to three weeks earlier.

Hyperliquid positions are public, so we can track how traders of different sizes respond to a report. After 21 of the 26 reports, accounts with positions under $250k bet that the stock would reverse: they bought when it fell and sold when it rose. Up to the open, these bets mostly lost. In 17 of the 21 large moves, a stock that had fallen overnight opened lower still, and one that had risen opened higher still. Larger accounts, by contrast, bought and sold in similar amounts.

Alphabet’s report followed this pattern. In the five hours after the release, smaller accounts bought about three times as much as they sold, larger accounts sold slightly more than they bought, and $15.8M of positions were liquidated in the first hour, the most after any megacap report this year. The stock opened 11.7% lower. Amazon was one of the exceptions: smaller accounts sold as the perp fell, betting on a further drop, and the stock opened above the perp’s overnight price, so that bet lost too.

A day before the release, Polymarket’s contracts leaned the right way on four of the five Q3 reports, with Amazon the exception. The last trades before the release then moved Alphabet from 25% to 75% before it missed and Tesla from 56% to 39% before it beat. At the release, the contracts had Meta and Microsoft right, Alphabet and Tesla wrong, and Amazon at 50%. With so little traded, a few orders were enough to move these prices.

Even a correct forecast of the result would have been a poor guide to the share price. Of the 20 reports that beat estimates, as resolved on Polymarket, 12 were followed by a lower share price at the next close; all 6 misses were followed by a fall.

Polymarket’s contracts traded between a few hundred and a few thousand dollars a day, picking up in the final two days before the release. Each contract settled about two hours after the release, and those two hours were the busiest of its life measured per hour, with up to $6,800 traded for Meta. By then the result was public, and trades took place at 0–1¢ or 99–100¢ as holders cashed out and buyers collected the last cent before settlement.

With the largest US tech companies spending record amounts on AI, each earnings report is a test of whether that spending justifies their valuations. Of the two onchain markets, Hyperliquid’s perps offer the more direct signal: they price the stock itself and trade hundreds of millions of dollars after each report, while Polymarket’s contracts trade a few thousand dollars a day, and even a correct forecast of the result says little about the share price. On the perps, the price at 8pm and at 9am is worth watching: both have been close to where the stock opened, though usually a little short of it. The position data, which US markets do not publish, suggest why: smaller traders tend to bet on a reversal straight after the release, which may hold the perp back, and the stock then opens further along. Apple reported on 2 November, after our cut-off, with Polymarket pricing a beat at 82% on 31 October, and Nvidia reports later in November.