Nasdaq's 5.5% plunge prices in a political risk the AI boom ignored: the electricity bill behind data centers

The AI trade had one of its worst days in years on Thursday. The Nasdaq Composite fell 5.52% to 24,315 points, according to market data compiled by CNN Chile from Investing.com. That puts the index in correction territory, more than 10% below its October record. The S&P 500 lost 4.30% and the Dow Jones 3.44%, or 1,834.84 points. The Nasdaq 100, which tracks the largest tech companies, dropped 6.05%. More than $2 trillion in market value disappeared in a single session. That wiped out Wednesday's post-election rally.

The reported trigger was political. Democrats won control of Congress in the midterms and also won state races. Investors fear new rules, higher costs or pauses in data-center construction. Data centers are the server farms that train and run AI models. Their effect on rising electricity prices had become a campaign issue. Semiconductors took the worst of it. The Philadelphia Semiconductor Index fell 12.25%, roughly twice the Nasdaq 100's loss. SanDisk dropped 22.26%, Micron 17.46% and AMD 17.26%. Nvidia lost 9.26% to $181.60. Qualcomm added to the pressure with a weak outlook for smartphone chips in the fourth quarter. AI infrastructure names also fell sharply: CoreWeave 9.83%, Palantir 9.73% and Oracle 6.93%.

The pattern matters as much as the size of the drop. Big platforms with diversified revenue held up much better. Apple fell 1.53%, Amazon 2.19%, Alphabet 2.20% and Microsoft 3.58%. Investors sold the companies whose story depends on building ever more compute: memory, GPUs and leased capacity. Musk's companies were a separate case. Tesla fell 10.77% and SpaceX 9.05% after Republican governor Greg Abbott lost in Texas. Analysts quoted in the report say that loss complicates the regulatory path for a possible merger of the two. Money moved to safe havens. The 10-year Treasury yield eased to 4.83% and gold rose 0.63% to $4,098.67 an ounce.

Our thesis: this was not a verdict on AI's capability. It was the market discovering that AI's real bottleneck is political and physical. The industry has priced in endless capex, meaning ever more spending on chips, power and buildings. The binding constraint is no longer only chip supply. It is also whether voters accept the cost on their electricity bills. When that becomes an election issue, the most leveraged links in the supply chain reprice first. One day's move proves little. Markets overshoot, and Friday's jobs report could shift sentiment again. No specific rule has been proposed yet. Investors are trading fear of regulation, not regulation itself.

Our reading: this matches a point we keep making. Regulation is needed, but the risk is regulating panic rather than evidence. A data-center build-out that pushes up household power bills is a real and legitimate concern, especially in the short term. The answer is to pay for new generation and grid capacity, not to freeze compute. The long-term prize of AI depends on abundant compute: drug discovery, longer healthy lives, cheaper services. That compute needs abundant, clean and fairly priced energy. If Washington pairs any new data-center rules with an energy build-out, this sell-off will look like a healthy correction to a speculative run. If it turns into simple restriction, the cost will show up later in slower progress, not just in share prices.