$1 Trillion Erased Overnight: US Tech Stocks Hit by Black Thursday as AI’s Political Costs Are Repriced and Jensen Huang Loses His Top Ten Richest Ranking

5.52% was the Nasdaq’s loss on 5 November. All three major US stock indices suffered steep declines, with the Dow down 3.44% and the S&P 500 lower by 4.36%. Technology took the hardest hit. The Philadelphia Semiconductor Index finished 12.25% lower; SanDisk slumped more than 22%; AMD, Micron and Western Digital each dropped more than 17%; SK Hynix lost over 13%; Broadcom retreated more than 10%; and Nvidia shed 9.26%.

The midterm election results supplied the immediate catalyst: a “blue wave” restored Democratic control of the House and Senate. The party also made gains in major hubs for data centre construction, including Texas and Georgia. For markets, the concern was tighter Democratic conditions on the physical expansion of AI infrastructure. These could cover environmental approvals, power connections and land-use permits, or even temporary moratoriums. UBS puts the proportion of planned additional US data centre capacity potentially affected at approximately 30%.

Political risk has therefore entered the valuation equation. AI assets had previously been priced for unlimited demand and an eventual supply catch-up, with politics carrying effectively no weight. That zero-weight assumption was no longer tenable after 5 November.

AI Business Models Face Regulatory Approval Risk

A single session erased $447 billion from Nvidia’s market capitalisation as the shares finished 9.26% lower. The stock’s retreat from its record closing high on 6 October reached 24.1%, officially putting it in a technical bear market. Jensen Huang moved from ninth to thirteenth on the Bloomberg Billionaires Index after a $15.1 billion drop in net worth left him with $150.9 billion.

For Musk, the one-day wealth loss was greater still at approximately $79.7 billion, leaving $785 billion. Tesla finished 10.77% lower, while SpaceX lost 9.05%. His particular vulnerability lies in Texas. The Democratic governor-elect has threatened to freeze new data centre construction following the election victory. Both SpaceX and Tesla rely heavily on the state’s approval environment for their AI infrastructure plans.

A structural issue weighs on Ellison. This year’s 39% fall in Oracle shares has left them more than 53% below their November 2025 peak. Ellison’s net worth now stands at $164 billion following a $159 billion reduction over the past year, almost halving his fortune. The loss on 5 November alone added another $13.2 billion. Oracle earns part of its core revenue by supplying computing capacity through cloud services. How fast it can increase that capacity is therefore directly determined by the pace of data centre approvals.

Large-scale data centre development is the common dependency across these three business empires, underpinning their ability to meet AI computing demand. Each occupies a different position: chip supply at Nvidia, self-built computing capacity at SpaceX and capacity sales at Oracle. All three asset categories face a higher valuation denominator if construction approvals change from a presumed open route to a wait at every junction.

Investigations, scrutiny of appointments and oversight of funding allocations fall within Democrats’ powers now that they control both chambers. An outright data centre ban cannot simply be legislated. For most ordinary legislation, overcoming a Senate filibuster still takes 60 votes, and a presidential veto remains possible. Committee hearings and investigations can nonetheless keep policy uncertain, lengthen approval procedures and lift compliance costs. Suppose an asset’s valuation relies on computing capacity doubling over the next three to five years. Slower approvals need not amount to a “ban”: a “delay” is enough to change its discounted cash flow result.

Beneath the selling lay a broken assumption. Before 5 November, the AI trade rested on indefinitely expanding computing capacity, with chip supply and capital expenditure treated as the only constraints. The midterm vote broadened that list to electricity, land, environmental permits and local political willingness. Extra capital spending cannot immediately overcome those barriers. They take years to resolve, not quarters.

Crowded AI Positions and Leverage Accelerate Selling

Extreme price gains across the AI hardware supply chain this year have been matched by extreme concentration in institutional portfolios. A change in expectations can compress automated stop-loss selling, leveraged position unwinds and retail sales into a short period, making losses larger. Founder ETFs portfolio manager Michael Monaghan noted that highly leveraged hedge funds subject to tighter risk limits had to liquidate quickly when prices began to fall. Their exits intensified intraday volatility.

At CoreWeave, the credit signal was particularly significant. Alongside an approximately 10% share-price decline that day, its five-year credit default swap spread hit a record high. Wider CDS spreads show that credit markets have begun to reprice AI computing operators’ risks. That assessment is consistent with the equity retreat. Investors are less prepared to overlook mounting financing and regulatory costs simply because computing capacity is scarce.

Nvidia Earnings and Approval Timelines Set the Next Tests

Later this month, attention will centre on Nvidia’s earnings report. Monaghan projects another 70% rise in Nvidia’s revenue next year. He believes data centre construction is “likely to continue”, supporting strong chip demand, and views the steep share-price fall as an opportunity to buy. Aptus Capital Advisors’ David Wagner takes a more cautious approach. Nvidia still has a wide competitive moat built on chip performance, its software ecosystem and developer lock-in. Yet its valuation is strongly dependent on AI infrastructure spending, and markets believe expenditure of that magnitude will eventually decelerate.

The adjustment concerns AI’s expansion timetable and route, rather than the validity of its long-term trend. A route that previously offered green lights throughout now has red lights at some junctions. Vehicles still reach their destination, but the journey takes longer. For assets drawing most of their value from future cash flows, that longer wait alone implies a lower valuation.

Political costs entered AI infrastructure pricing models for the first time in the 5 November sell-off. The resulting wealth losses were effects, not explanations: Huang fell outside the top ten, Musk lost nearly $80 billion, and Ellison’s fortune almost halved over a year. What drove the repricing was a recognition that ballots, as well as chips and electricity, can constrain computing expansion.

The practical test is now the time it takes for slower approvals to alter the computing capacity actually brought into service at data centres. Nvidia’s earnings pose a parallel test: can strong orders and a longer path to future growth produce a new balance acceptable to markets? Until approval issues actually hold up the first data centre, this decline prices a probability, not an event that has already happened.