
Prediction markets called the sweep by 8 PM ET
The implied probability of a Democratic sweep rose from 74% on Polymarket and 72% on Kalshi at the US stock market close (4 PM ET on 3 November) to about 98.5% on both venues by 8 PM ET and 99% by 9 PM, well before most races were formally called. Kalshi's contract reached 100% by midnight. Midterm markets on the two venues traded $573M of notional in the final seven days, 4.2 times the previous week, and $794M in the 24 hours after the close on election day. Kalshi accounted for 88% of results-night volume ($698M against $96M on Polymarket), up from 68% in the final week.
Prediction markets data
Request accessStocks rallied on the result before giving back the gains
AI stocks were already weak going into the vote. Earnings reports in late October showed capital spending still rising, and Micron and CoreWeave fell 15% and 22% respectively between 28 October and 2 November, while Meta and Amazon declined after raising their spending plans. Hedge funds also cut risk, with gross exposure recording its largest drawdown since 2016.
On 4 November the S&P 500 rose 1.6% and the Dow 3%, with stocks sorted by their exposure to policy. Suppliers to the AI build-out fell, including Vistra (-13.8%) and CoreWeave (-9.9%), while Microsoft and Amazon gained 6-7% and the solar ETF 5.7%. Eight of the ten stocks that moved most after the 2024 presidential election reversed direction, with GEO Group down 1.9% after a 42% gain following the 2024 presidential election and Tesla down 3.3% after a 15% gain.
The selloff broadened on 5 November, driven mainly by concern over new data-centre construction. According to Cleanview data cited by Bloomberg, about 40% of pending US data-centre capacity sits in states where newly elected governors have promised to stop or pause such projects, and J.P. Morgan's Michael Cembalest estimates that more than half of the capacity planned for 2028 could face pauses or conditions. The S&P 500 fell 4.4% and the Nasdaq-100 6.4%, leaving both below their pre-election levels, and the previous day's winners in solar and electric vehicles gave back their gains. Over the two days, Micron fell 20%, Vistra 19% and CoreWeave 16%.
Traders added to falling AI stocks and sold the winners
Hyperliquid perpetuals provide a view on how traders positioned around the result.
- US AI chips and memory: the number of contracts held rose 60% over the two days while prices fell 12-28%, and funding rose to about five times its resting level. Funding stayed high during US trading hours, so the premium cannot be explained by a stale reference price. Traders were therefore adding long positions and paying a rising premium to hold them through two days of losses. A crowded, losing long position can unwind quickly if prices keep falling, as forced selling adds to the decline.
- Korean chips: contracts rose 56%, although funding follows Korea's trading hours and does not show the direction of the new positions.
- Megacap AI buyers: contracts on the Microsoft, Amazon, Meta and Alphabet perpetuals fell 27% from election day, first as prices rose on 4 November and then as they fell on 5 November, which points to profit-taking followed by de-risking.
- AI infrastructure: contracts declined as prices fell, with Vistra (para:VST) contracts down 29% while its price dropped 19%, suggesting that holders closed positions.
- Musk companies: Tesla (xyz:TSLA) traders appear to have moved from short to long. Open interest rose overnight on 4 November as funding turned negative, and positions were rebuilt with rising funding on 5 November as the price fell. SpaceX (xyz:SPCX) contracts were little changed overnight on 4 November, but ended the two days 14% higher as the price fell 12%, so dip-buying also reached SpaceX on the second day.
- Indices: in the S&P 500 and Nasdaq-100 perpetuals, funding stayed close to its resting level during the 4 November rally, consistent with demand that matched the cash market, before contracts rose 15% and funding tripled on 5 November as traders bought the decline.
Anthropic perps offer the most direct read on demand
Anthropic has no public market, so its Hyperliquid perpetuals (io:ANTH and xyz:ANT) are the only venue where its value is repriced in real time. With no stock to arbitrage against, funding on these contracts moves mainly with demand. On 4 November the price rose 6%, contracts increased and funding reached about 17 times its resting level. Most of that repositioning took place while US markets were closed. In the 24 hours from 8 PM ET on election night, more than three-quarters of io:ANTH's volume traded before the US open and only 15% during US hours, compared with 43% on a typical day. Perpetuals on US-listed stocks behaved differently, as they tend to trade when the underlying stock trades and sets the reference price: the S&P 500 perpetual still did 35% of its volume during US hours over the same 24 hours. Anthropic contracts held have more than quadrupled since 1 October, most of it after the company filed on 26 October for a Nasdaq listing that could be the largest IPO on record. Together with funding well above its October levels, this suggests strong demand for long exposure to a company that investors cannot otherwise buy.
Perps priced the sector moves before the open, while the broad market turned during US hours
On 4 November, by 9 AM ET, half an hour before the US open, stock perpetuals were already moving in the direction of the day's close for 15 of 16 stocks and, for the median stock, had already priced about two-thirds of the day's move. On 5 November they again pointed the right way for most stocks but had priced only about a fifth of the decline, and the S&P 500 perpetual was flat at 9 AM. In both sessions, the broad market move came once US trading began. Activity rose with volatility, as stock perpetuals traded $3.2B on 4 November and $4B on 5 November before the US close, compared with $2.4B on election day.
Our take
While not a substitute for the cash market or for fundamental work, onchain markets improved the speed and specificity with which the midterm result could be read and traded. Prediction markets settled the political outcome by 8 PM ET, more than 13 hours before the US open. Stock perpetuals turned that outcome into sector calls overnight, pointing the right way for 15 of 16 stocks before the open on 4 November. Positioning data then showed how traders responded: crowded dip-buying in the hardest-hit US chip and memory names, de-risking in the megacaps that had rallied, and a clear demand signal in Anthropic. The signals have limits, as perpetuals missed the index's turn on both days, and the clearest policy winners, solar and wind stocks, still have no perpetuals on Hyperliquid.


