EXPLAINED | $2.5 Trillion Wiped Off the Nasdaq 100. Why Did the Democrats’ ‘Blue Tsunami’ Spook AI Investors?

A company can buy the chips, borrow the money and sign up a customer. It still needs somewhere to plug in.

That became a far more expensive question on Wall Street on Thursday, 5 November. Two days after Democrats won both chambers of the US Congress, the Nasdaq 100 fell 6.4%, wiping about $2.5 trillion off the combined market value of its companies, according to Reuters. The broader Nasdaq Composite fell 5.5%. When South Korea opened on Friday, its main index dropped so quickly that the exchange stopped trading for 20 minutes.

The election had given new weight to demands for limits on data-centre construction. That landed in a market already asking how long it would take for the billions pouring into artificial intelligence to earn their keep. The sell-off brought those questions together.

What exactly did investors lose?

The $2.5 trillion is Reuters’ estimate of how much the Nasdaq 100 companies’ combined market value fell in one session. Think of a company with a billion shares. If each share closes $10 lower, the company’s market value falls by $10 billion. That measures what investors now value the shares at. The company’s cash balance is a separate figure.

The damage extended well beyond a handful of names. The S&P 500 fell 4.4%, and the Dow lost more than 1,800 points. Chipmakers took an especially hard hit. Nvidia fell about 9%, AMD about 17% and the Philadelphia Semiconductor Index about 12%. The broader Nasdaq Composite finished nearly 12% below its 6 October record close. A fall of 10% or more from a recent high is what investors call a correction.

What did the election change for a data centre waiting to open?

It increased the chance that some projects could face a longer wait or tougher conditions. Texas is where that risk becomes concrete. Republican Governor Greg Abbott had already directed the state’s environmental agency to pause permits for data centres during a review. His Democratic successor, governor-elect Gina Hinojosa, campaigned on halting new projects until safeguards were in place. After winning, she told CNN that rural communities needed protection. At the federal level, Senator Bernie Sanders has proposed a moratorium on new AI data centres. That proposal would still have to become law.

Take two projects. One has the permits it needs and a date for getting electricity. The other still needs approvals and a grid connection. A tougher permitting regime matters much more to the second. Its developer may have to wait longer, pay for additional infrastructure or revise the site plan. Each choice can change when the equipment is delivered and when the customer starts paying.

The fight was already forcing companies to address who picks up the bill. On 5 November, OpenAI said that for future campuses it owns and operates, it would cover the electricity and infrastructure costs it creates and provide financial assurances before major spending starts. The commitment goes straight to a question local residents and utilities can ask. If a power company builds a substation for a data centre that opens late or never opens, who pays for it?

Why did companies that rent out AI computing get hit so hard?

They have to pay for a lot of equipment before customers can use it. Specialist cloud companies such as CoreWeave borrow to build and equip facilities, then rent out access to the machines. The borrowing has a payment schedule. A delayed opening can leave the company paying interest while it waits to start collecting revenue from that capacity.

CoreWeave shares fell 10% on Thursday. At the same time, the market cost of buying protection against a default on its debt reached a record, Reuters reported, citing S&P Global Market Intelligence. That shows how much more investors were willing to pay for protection. The stock and debt markets were both reacting to the risk that a business built on expensive equipment and borrowed money might have to wait longer for its returns.

IREN, another company that rents out AI computing, supplied a real example the same day. It moved its target for reaching a revenue pace of more than $4 billion a year from December to January 2027, with February possible. It cited a customer-requested design change and supplier timing. IREN said the capacity behind the target was already contracted and funded. It recorded $200 million in AI cloud revenue in the September quarter. The larger figure describes the pace the company expects once more machines are running. The smaller figure is the revenue it recorded over those three months.

A delay and a cancellation have very different consequences. If a customer takes the same capacity two months later, the supplier has to carry the cost for longer. If the customer takes less capacity, the supplier may also have to find another buyer. Thursday’s selling showed how much investors cared about that distinction. The next earnings calls may tell them which problem they actually have.

How did a fight over US data centres stop trading in Seoul?

Follow one AI server. Its processor may be designed in the US, its high-speed memory supplied by a South Korean company and its chips manufactured in Taiwan. Nvidia and SK Hynix are working together on future generations of AI memory. If buyers in the US slow orders or move delivery dates, companies across that supply chain can feel it.

Japan’s Nikkei fell more than 6% in early Friday trading, and Taiwan’s chip stocks came under pressure. In Seoul, Samsung Electronics and SK Hynix were among the biggest losers. At 9:03 a.m. local time, the Korea Exchange activated its circuit breaker. The first stage halts trading for 20 minutes when the KOSPI falls at least 8% from its previous close and stays there for one minute. At 9:26 a.m., after trading resumed, the index was still down 8.6%. The pause gave traders time to reassess. The sellers came back.

When would India hit pause, and where else does this matter here?

In India, the first nationwide threshold is a 10% move in either direction in the Sensex or Nifty 50, whichever reaches it first. If that happens before 1 p.m., trading pauses for 45 minutes, followed by a 15-minute reopening session. An 8% fall of the kind that triggered Seoul’s halt would fall short of India’s first market-wide threshold.

There’s also an Indian project through which to follow the same questions about power, costs and timing. TCS’s HyperVault and OpenAI have announced an initial 100 MW of AI infrastructure in India, with an option to expand. The milestones to watch are when the site gets power, when the machines go live and when customers start using them. The US debate makes another question worth asking here. Who pays for grid upgrades if the project takes longer than planned?

What happens next?

CoreWeave reports on 10 November and Nvidia on 17 November. CoreWeave can tell investors whether it is bringing computing capacity online on schedule. Nvidia can give a wider view of whether customers are moving delivery dates or cutting orders. In Texas, watch the permitting decisions. A project that opens later can still buy the same chips. A project that gets scaled back may buy fewer. For a market that just took $2.5 trillion off the Nasdaq 100, the difference is what happens to the next order.