Why Did Wall Street Just Have Its Worst Session Since Covid?

Why did Wall Street just have its worst session since 2020?

On Wednesday, Wall Street was welcoming the end of election uncertainty with a solid rally, but twenty-four hours later, the US market went through one of the most violent sessions in its recent history.

The S&P 500 lost 4.35%, its worst day since Liberation Day in April 2025, and the Nasdaq 100 plunged 6.36%, something not seen since March 2020. Semiconductors were hit the hardest, with SOXX down 12% and some stocks falling as much as 20%.

Drops of this magnitude are very rare, and they usually happen during panic moves like Covid or the tariff shock.

So how could the market swing from euphoria to panic so quickly? To understand it, we first need to look at the state it was in even before Thursday’s session.

A market that was already fragile

For several weeks, doubts had been building around AI stocks. As we explained in our Alpha on Big Tech earnings, investors had started to wonder whether the hundreds of billions of dollars invested in data centers would really pay off.

October had in fact already been the worst month for tech stocks since March 2025, while investors retreated into more defensive sectors like healthcare and consumer staples.

In the space of a few months, the market narrative even flipped completely, moving from a shortage of computing power to the opposite risk, that of having built too much capacity.

The day before the drop, disappointing outlooks from Qualcomm and Arm had weakened the sector even further. In such a tense market, all it took was a spark to trigger panic, and it came from the ballot box.

Data centers at the center of the debate

To understand why the election results worried the market so much, it’s worth remembering the role of data centers. These huge buildings are where Nvidia’s chips, Micron’s memory and the hyperscalers’ servers run, and they’re at the heart of the entire AI thesis.

Anything that threatens their construction therefore directly threatens the future revenue of the entire value chain.

Over the past few months, however, building data centers has become a major political issue in the United States. The largest ones can consume as much electricity as an entire city, so power companies have to build new plants and new lines to supply them. Part of these costs ends up on household bills, at the very moment many Americans are already struggling with the rising cost of living.

Water raises a similar problem, since these facilities use large amounts of it to cool their servers. In regions already hit by drought, like parts of Texas, farmers and residents feel they have to share a scarce resource with tech giants. And once built, these data centers create relatively few jobs, which reinforces the feeling that local communities bear the downsides without really benefiting from them.

Candidates naturally seized on this discontent, and it was in Texas, historically Republican and one of the industry’s main hubs thanks to its land and long-cheap electricity, that it showed up at the ballot box.

Gina Hinojosa won the governor’s race there after campaigning on a moratorium, while across the rest of the country, many towns, notably in Ohio, voted to restrict new projects.

For the market, Texas obviously remains the most sensitive point, since according to The Daily Upside, the freeze ordered by Greg Abbott in September already threatened to delay nearly a fifth of the data center projects in development across the United States. Gina Hinojosa’s win should therefore translate into stricter rules on who pays for grid infrastructure, water use and the tax incentives granted to the industry.

These fears nonetheless look largely overblown to us, first because the United States has no interest in slowing down. AI has become a geopolitical issue in the rivalry with China, just like semiconductors or energy, and no US administration will risk giving up its lead by slowing the construction of its infrastructure.

The restrictions that were voted are also local or specific to certain states, and they could be challenged, eased or worked around. In Texas, Gina Hinojosa won’t take office until January 19, and a lasting moratorium will have to go through a law that the Texas Senate, which remains Republican, will be able to block.

The second reason is that projects can simply move if a state becomes too hostile. Hyperscalers will then build elsewhere, first in other, more welcoming US states, then abroad. Countries like the United Arab Emirates or Malaysia already host major data center campuses, and demand for compute won’t disappear because a county in Ohio voted against a project.

These relocations have their limits, though, since Washington tightly controls chip exports and has every interest in keeping most of its computing power on its own soil. They will also take time and cost more, which could delay some projects and push back orders for some suppliers.

In other words, data center construction should slow down in some places without coming to a halt, which didn’t stop the market from reacting as if the entire chain were under threat.

When sellers no longer have a choice

These concerns explain why the market fell, but not why it fell so sharply. For that, we need to look at who was selling, and above all at the fact that many of them didn’t really have a choice.

The first amplifying factor is leverage, since up until the elections, many investors had kept buying every dip with borrowed money. According to The Kobeissi Letter, US leveraged ETFs attracted around $15 billion in October, their largest inflow in at least two years.

The problem is that these ETFs, which aim to deliver two or three times the performance of an index, have to sell automatically when that index drops sharply. According to Nomura, they may have had to unload up to $47 billion worth of stock near the close, mostly AI-related names.

Hedge funds then amplified the move, because these funds often borrow from big banks to increase the size of their positions. As long as the market rises, everything is fine, but when it falls, banks demand additional collateral, and a fund that’s short on cash then has to sell whatever it can, regardless of price.

According to Goldman Sachs data, hedge funds had already sold around $20 billion of tech stocks over two weeks in October, without really lightening their load, since tech still accounted for nearly a quarter of their exposure.

On top of this forced selling came a lack of buyers, since retail investors, who had been used to buying every tech correction, this time went through their worst day of selling since the pandemic, before changing course in the last ten minutes of the session.

At the very moment funds had to sell en masse, there were far fewer people on the other side to buy, and the VIX, the fear index, logically jumped 63% on the session.

Bonds as a safe haven

As often happens during panic moves, the money leaving equities largely took refuge in US government bonds. The 10-year yield, which was still above 5.3% at the end of September, its highest in nearly twenty years, fell back to around 4.8% yesterday.

This move says a lot about the current situation, because part of the rise in rates over recent months came precisely from the race to build AI infrastructure, whose massive investments were fueling inflation.

For rates to come down, we needed either a soft landing for the economy, or an equity correction strong enough to deflate valuations and cool demand. For now, it’s the second scenario that seems to be taking shape.

The relief on rates could be short-lived, however, as the jobs data released today came in well above expectations, with 165,000 jobs created versus 59,000 expected. Prediction markets now see around a 79% chance of a Fed rate hike in December, and for the most indebted AI companies, starting with some neoclouds, the cost of financing therefore remains a real issue.

Is the worst behind us?

After a session like this, everyone is wondering whether the worst is behind us, and the answer is nuanced. Based on our conversations with several institutional players, hedge funds have reportedly already cut a large part of their positions, which is rather good news since it was their forced selling that made the session so violent.

Other categories of investors could take over, though, starting with CTAs, the funds that follow trends automatically. They buy when the market rises and sell when it falls, without asking any questions about what companies are worth, so such a clear downtrend could trigger new mechanical selling in the coming days.

Retail investors, who are only just starting to sell after buying every dip in recent years, and more traditional funds, which have been relatively spared so far, could also trim their positions if the drop continues.

The calendar doesn’t help either, as we’re approaching the end of the year. In the United States, many investors sell their losing positions at this time of year to offset their capital gains and reduce their taxes, and the AI stocks that have fallen the most are the first ones affected.

The market could therefore still fall in the coming weeks, but for other reasons, and the most violent phase, when the whole market dropped indiscriminately, seems to be coming to an end.

Our take

A drop this violent doesn’t mean companies’ prospects have deteriorated as much as their share prices. A good part of the decline comes from forced selling, and markets can rebound just as violently once that selling is over.

We still need to stay cautious, though, because the market remains very tense, starting valuations were high, and it no longer forgives anything from AI companies. Averaging down without taking the time to reassess can be costly, as some stocks that look cheap are actually far riskier than they appear.

A session like this does have the advantage of acting as a filter, showing very clearly which companies held up despite the panic and which were liquidated first. And it’s often among the solid companies that were wrongly punished, rather than among those that simply fell the most, that the best opportunities are hiding.

Premium members only

In our latest Premium Alpha, we reviewed the stocks hit hardest by this drop to separate those that were wrongly punished from those whose decline is justified, and to identify the best opportunities for our members.

Because after a session like this, our goal isn’t to predict the market bottom, but to figure out which companies deserve your attention once the drop is over.

Want to discover our selection? Head over to OAK Premium and enjoy a 7-day free trial or 30 off with the code 1ANOAK!