$1 Trillion Overnight Loss Hits US Tech Stocks on Black Thursday as Markets Reassess AI’s Political Costs and Jensen Huang Exits the Top Ten Richest

Falls of 3.44% in the Dow, 4.36% in the S&P 500 and 5.52% in the Nasdaq marked a sharp retreat across the three major US stock indices on 5 November. Technology suffered the most pronounced selling pressure. A 12.25% decline left the Philadelphia Semiconductor Index sharply lower at the close, while SanDisk lost more than 22%. Losses exceeded 17% for AMD, Micron and Western Digital, over 13% for SK Hynix and more than 10% for Broadcom. Nvidia fell 9.26%.

Behind the immediate market reaction was the midterm elections’ “blue wave”. Having recaptured both the House and Senate, Democrats also achieved breakthroughs in important centres of data centre construction, among them Texas and Georgia. Investors worried that the party would tighten conditions governing the physical expansion of AI infrastructure. Environmental approvals, connections to power supplies and land-use permits could all face stricter requirements, with even temporary moratoriums a concern. In UBS’s estimate, approximately 30% of planned additional US data centre capacity could be affected.

Valuations now rest on a different premise. Before the vote, AI pricing reflected expectations of unlimited demand and supply that would eventually catch up; political factors effectively counted for nothing. Since 5 November, excluding those factors from the calculation has no longer been possible.

Data Centre Approvals Put Business Models Under Pressure

For Nvidia, the closing decline of 9.26% meant a one-day reduction of $447 billion in market value. The stock was officially in a technical bear market, standing 24.1% below its record closing high reached on 6 October. A $15.1 billion reduction took Jensen Huang’s net worth to $150.9 billion and his position on the Bloomberg Billionaires Index to thirteenth, from ninth.

An even larger one-day loss fell to Musk: approximately $79.7 billion, bringing his wealth to $785 billion. Tesla’s 10.77% closing decline and a 9.05% fall in SpaceX accounted for the reduction. His exposure has a particular focus in Texas, where the Democratic governor-elect has threatened to halt new data centre construction. Approval conditions in the state are closely linked to the AI infrastructure plans of both SpaceX and Tesla.

A more structural problem confronts Ellison. Oracle’s stock is down 39% so far this year and more than 53% from its peak in November 2025. Ellison has lost $159 billion over the past year, leaving a net worth of $164 billion, almost half its previous level. A further $13.2 billion was lost on 5 November alone. As a core part of Oracle’s business involves selling computing capacity via cloud services, the rate of data centre approvals directly dictates the speed at which it can add capacity.

Across these three business empires, the shared exposure is substantial dependence on building data centres at scale to satisfy demand for AI computing. The chips come from Nvidia; SpaceX constructs computing capacity for its own use; Oracle sells capacity to customers. If approvals for that construction move from a presumed unobstructed route to a process involving a wait at every junction, all three categories of assets face an increase in their valuation denominator.

Democratic control of the House and Senate brings the power to investigate, oversee funding allocations and examine appointments. An outright legislative ban on data centres is not simply within the party’s reach: most ordinary legislation must still secure 60 Senate votes to overcome a filibuster and can face a presidential veto. Nevertheless, sustained policy uncertainty, longer approval procedures and higher compliance costs can arise from committee hearings and investigations. Consider an asset priced on the expectation that computing capacity will double over the next three to five years. Approvals need not slow to the point of a “ban”; a “delay” on its own changes the result of discounted cash flow calculations.

The deeper assumption laid bare by the selling was that, before 5 November, the AI trade treated computing capacity as capable of indefinite expansion, limited only by chips and capital expenditure. That view changed with the midterm elections. Constraints now also encompass electricity, land, environmental permits and the willingness of local politicians to support expansion. Spending more capital cannot resolve these issues at once: the relevant timescales run to years rather than quarters.

Leverage and Crowded Holdings Deepen the Decline

Extreme advances in the AI hardware supply chain this year have coincided with extreme concentration of institutional holdings. Should expectations turn, retail sales, automated stop-loss selling and the unwinding of leveraged positions can arrive in quick succession and magnify losses. Michael Monaghan, a portfolio manager at Founder ETFs, observed that falling share prices forced hedge funds with high leverage and tighter risk limits to sell out quickly, making intraday trading more volatile.

Particularly notable was the combination of moves at CoreWeave. Its five-year credit default swap spread reached a record high as its shares declined approximately 10% that day. The widening CDS spread indicates that a reassessment of AI computing operators’ risks is under way in credit markets. Equity selling conveys a similar signal: rising financing and regulatory costs are becoming harder for investors to overlook on the strength of the scarce-computing-capacity narrative.

Nvidia Earnings Bring Growth Expectations Into Focus

Later this month, Nvidia’s earnings report will offer a key point of assessment. Monaghan forecasts a further 70% increase in revenue next year. With data centre construction, in his view, “likely to continue”, he expects chip demand to remain strong and considers the steep share-price decline an opportunity to buy. David Wagner of Aptus Capital Advisors set out a more cautious view. Although Nvidia’s chip performance, software ecosystem and developer lock-in still give it a wide competitive moat, its valuation relies heavily on AI infrastructure expenditure. Markets, he noted, believe spending on that scale will ultimately slow.

This is a repricing of how quickly AI expands and the route it takes, rather than a rejection of the long-term trend. Green lights along an entire route allow a different journey time from red lights at some junctions. The traffic does not stop altogether, but takes longer to arrive. That extra time itself implies a lower valuation for assets whose worth is derived chiefly from future cash flows.

What distinguished the sell-off on 5 November was that markets, for the first time, built the political costs of AI infrastructure into their pricing models. The consequences included Huang dropping out of the top ten, nearly $80 billion being wiped from Musk’s wealth and Ellison’s fortune almost halving over a year. None was the cause. Rather, the market had recognised that the bottlenecks facing computing expansion include the ballot alongside chips and electricity.

Markets must now assess the speed at which approval delays change the computing capacity actually entering service at data centres. Nvidia’s earnings will also be examined for a new balance that investors can accept between robust orders and a longer path to future growth. Unless and until approval problems actually delay the first data centre, the decline reflects the pricing of a probability, not an event that has already occurred.