After election results last Wednesday, it felt like we got some relief from the market grind we’ve experienced since June. But under the hood, the semiconductor index fell 3%, as money rotated out of AI names. On Thursday, major indices dropped with tech-heavy stocks hit the hardest. The Nasdaq fell more than 5%, its worst day since April 2025, the semiconductor index down more than 12% and some data center operators down more than 14%.
Democrats who won their state seats campaigned on pausing data centers and curbing tax incentives. Policies like these could delay when a data center begins generating money, taking all of the beneficiaries of the data center build with them.
While the election results may have lit the match, the fuel for the deeper drop on Thursday was leverage. It appeared in 2 places:
money that had been borrowed to fund the data centers looks like its profits could be further out and
money that hedge funds, margin accounts, and leveraged ETFs had borrowed to own the same trade.
And this was just the latest of several unwinds this year, not the first.
I've been writing about leverage all year. Here's where it stood going into last week:
1. Hedge funds. Goldman Sachs's prime brokerage book, which tracks the borrowing and trading of its hedge fund clients, showed gross leverage at the highest level on record at the end of 2025, up for a third straight year. Net leverage was near three-year highs, and hedge funds' net exposure to semiconductors was at five-year highs.
2. Margin accounts. Margin debt, the money investors borrow from their brokers, hit a record $1.5 trillion in June, up 51.5% from a year earlier. The data comes out with a lag (below is through September), so it predates the last few weeks.
3. Leveraged ETFs. As shared in my previous writings, assets in US leveraged ETFs and single-stock leveraged ETFs have reached about $200 billion in assets. As I wrote in August, leveraged funds manage for the day, not longer, so their moves get amplified. Despite inflows into some of the larger ones, their AUM is still down in November.
Underneath all of this is a different layer: the company-level web of bank-like guarantees, equity stakes, and purchase commitments I wrote about in September.
And the credit market has been telling this story all year as well:
This part of the leverage can't be unwound in a day like what investors took on can.
Goldman's prime brokerage unit noted that between COVID and the end of 2025, fundamental, systematic, and multi-strategy hedge funds had never all lost more than 0.8% on the same day. It has now happened three times this year: February 4, July 28 and October 15.
In my August piece, I described a pattern: leverage builds, then a crowded trade, then a cascade. In the past episodes, the forced-selling phase lasted days or weeks. What came next depended on whether the damage reached credit markets and the broader economy.
Past episodes of leverage all ended in different ways. In 1987, 1998, and 2018 the down moves stayed relatively contained, and the market recovered in months to about two years. In 1929 and 2000-2002, leverage unwound alongside collapsing valuations and a weakening economy, and the recovery took years. In 2020, a fast response from the Fed helped. In 2008, of course, it spread everywhere and recovery followed intervention from both the Fed and Federal Government.
Identifying a market bottom is not easy. Right now it feels fragile. I say that because we still have oil prices, inflation and interest rates to contend with that will keep targeting the debt that has built up in the system.
Based on past episodes, and market-structure work from Jordi Visser, here’s what you can watch:
1. With forced selling, is the leverage actually coming out? After the 2000 peak, margin debt fell by more than half. A real cleanup shows up in lower leverage, not just lower prices. Lower hedge fund gross and net leverage from the prime brokers, monthly FINRA margin debt, and leveraged ETF assets can signal this, on a delay. We’ll be watching for this.
2. Do the panic gauges and trading volume peak together? Fear shows up in the VIX, in options buying and in heavy trading volume, usually together. In 2018, the VIX more than doubled in a single day and the correction was over within about two weeks. It spiked from a very low level and has come down.
That being said, high beta stocks relative to low volatility stocks have repriced back to a more average level.
3. Does credit calm down? Credit default swaps (CDS), or insurance against bond default, eased from Thursday's high. Credit that stops widening is a sign the stress isn't spreading. We’ll be watching this closely. You can view some sideways moves since October:
4. Does the damage stay contained in the leveraged corner? The equal-weight S&P 500 closed at a record on Wednesday, which says the problem so far has been concentrated in AI names. If non-AI stocks and banks start falling too, that's a different story.
5. Who’s still standing? Fund closures and returned capital are part of a washout. The worry is when lenders start taking losses too. That's the point in past episodes when the Fed or banks stepped in, and I don't think we're there.
The patterns from my August piece still apply: know the exact price move that triggers a margin call, remember that a crowded trade has a small exit door, and give a second look to anyone saying there's nothing to worry about while a position is under pressure.
Politics gave investors a reason to sell on Thursday. Leverage is what turned it into a rout. And the end of the rout shows up in leverage too.