The technology sector suffered a sharp sell-off on Thursday, November 5, as the Nasdaq Composite closed down 5.5 percent. The Philadelphia Semiconductor Index closed down more than 12 percent, its steepest drop since March 2020, with more than a dozen large technology stocks suffering double-digit declines. The immediate triggers included weaker-than-expected phone chip sales at Qualcomm and political uncertainty. Democrats won majorities in both the House and the Senate in Tuesday’s midterm elections, along with governors’ races in Texas and Georgia where limits on data centers were a campaign issue, raising the prospect of tighter rules for artificial intelligence infrastructure. But the sell-off did not begin on Thursday. The Nasdaq had reached a record closing high on October 6, and October marked its worst month in more than a year and a half.
Analysts are divided on whether the bottom is near. In a forecast published before Thursday’s rout, Morgan Stanley’s Mike Wilson and his team suggested semiconductor stocks could fall another ten to fifteen percent from then current levels, with the Philadelphia Semiconductor Index finding support around 9,500 points and the correction potentially ending in mid-November. That forecast used an earlier index close of 10,779; on Thursday the index closed at 9,011, already below Wilson’s support level. In a July note, Academy Securities’ Peter Tchir warned that mechanical selling could deepen a decline and that the US economy could slow without increased compute spending. JPMorgan’s Daniel Pinto warned in November 2025 about elevated AI valuations and the potential for a correction. Not every signal points down: the S&P 500 has gained in the twelve months after October’s close in every midterm year since 1950, according to TradingView data cited by Benzinga. Most of the forecasts cited here, however, point to further near-term losses.
The deeper problem is not political but financial. The AI boom has been powered by an unprecedented surge in capital expenditure from a handful of hyperscalers (companies that run data centers at vast scale) including Amazon, Alphabet, Meta, Microsoft, and Oracle. Vanguard, citing Bloomberg consensus estimates as of July 31, projected their combined capital expenditure would approach $800 billion in 2026, up from $416 billion in 2025 and $97 billion in 2020. These figures cover total capital expenditure, much of it supporting the AI buildout, rather than AI spending alone.
Investors are now asking whether that spending will ever generate returns commensurate with its scale. The early evidence is not encouraging. Amazon reported negative trailing twelve month free cash flow of $42.4 billion for the period ended September 30, despite generating $168.3 billion in operating cash flow. Its investment spending, rather than an absence of operating cash generation, drove the outflow. Alphabet, which in July raised its 2026 capital spending guidance to $195 billion to $205 billion, posted roughly $9 billion in negative free cash flow in the third quarter.
McKinsey’s 2026 State of AI survey found that only about six percent of respondents qualified as AI high performers. These high performers attributed at least five percent of their earnings before interest and taxes (EBIT) to AI and reported significant value from its use. Thirty-seven percent reported at least some enterprise-level EBIT impact, so the findings point to limited widespread financial transformation rather than an absence of benefits. The gap between the hundreds of billions being poured into infrastructure and the returns investors expect is the defining tension of this market. “AI stocks are already in a private bear market that virtually nobody is talking about,” Adam Sarhan, chief executive of 50 Park Investments, told Reuters.
This has the anatomy of a classic bubble: an investment boom whose valuations depend on a narrative of transformative future returns rather than demonstrated returns on the new spending. Negative free cash flow does not prove those investments will fail, but it raises the stakes if demand disappoints. The midterm elections did not create that risk, but they added another source of uncertainty to an investment thesis that had been priced for perfection.
The political side of the sell-off shows how much of the AI trade depends on permits, power, and local consent. Democrats Gina Hinojosa in Texas and Keisha Lance Bottoms in Georgia won governors’ races in which data center development was a campaign issue, Reuters reported. Hinojosa had promised a data center moratorium, the Washington Examiner reported in October. Limits were not only a Democratic position, however. Her Republican opponent, Governor Greg Abbott, had himself announced a moratorium in August on approving new data centers until state regulators audit projects seeking connection to the Texas grid, the Texas Tribune reported. The policy tracker MultiState rated both Texas candidates as backing a moratorium, and Georgia Republican candidate Rick Jackson as backing new regulations. Goldman Sachs estimated that Democratic victories at state and local level could lead to temporary halts potentially affecting some 45 percent of planned US data center capacity expansion, with Texas alone accounting for about a quarter of planned capacity, though the bank expects such halts to be short-lived. That is an estimate of exposure, not a forecast that all that capacity will be cancelled.
JPMorgan’s Michael Cembalest estimated that 37.5 gigawatts of planned 2028 capacity could face pauses and another 15 gigawatts could face conditions such as grid cost payments, water limits, and labor mandates. He also indicated that expected 2027 deliveries should largely be unaffected, with much of that capacity already under construction.
For heavily indebted neoclouds, companies renting access to computing infrastructure, such delays are potentially existential, because they can postpone revenue while financing obligations continue. The breadth of the sell-off suggests investors are weighing not only more regulation but the possibility that the physical buildout of AI infrastructure, which was supposed to be the engine of the entire trade, may be slower and more expensive than they had assumed. The pressure has extended beyond chipmakers to engineering and construction companies, cooling suppliers, and power equipment manufacturers exposed to the buildout.
On one hand, Oxford Economics has warned that Britain would face a substantial risk of recession if the AI boom falters. The consultancy estimated that AI spending contributed around 0.2 percentage points of UK GDP growth in 2025 and was on track to do the same in 2026.
On the other hand, some economists and investors argue that a slowdown in AI capital expenditure could be beneficial. Evercore’s Krishna Guha has argued that slower spending could reduce overheating risks and extend the investment cycle by helping keep demand and supply in balance. The Financial Times’ Lex column has argued that delays could also reduce the risk of committing enormous sums to equipment that becomes obsolete or depreciates faster than its financiers expected.
Applied to this week, that logic suggests the political pushback against data centers, which the midterms reinforced, may do the economy a favor: a brake now on a boom at risk of overheating could lower the odds of a sharper bust later. The paradox of the moment is that the political forces that are bad for AI stocks in the short run may be good for the long-run health of the AI industry itself.
The big hyperscalers (Microsoft, Amazon, and Alphabet) have established businesses and substantial financial resources that give them room to adjust their investment plans.
Microsoft rose 6.7 percent, Amazon 6.1 percent, and Alphabet 3.3 percent in Wednesday’s post-election rally, while Oracle fell 2 percent. One reading is that investors expect slower infrastructure spending to improve the three companies’ finances. The companies most exposed to a buildout slowdown are those whose growth assumptions rely heavily on continuing investment: chipmakers, neoclouds, and the engineering and power suppliers counting on relentless demand growth.
For the AI industry itself, a period of consolidation and slower spending may be healthy. It would put greater pressure on companies to generate returns rather than simply scale capacity, and could intensify the focus on agents and enterprise tools that demonstrate productivity and profitability gains. The bearish argument is that narrative-driven valuations are running into a fundamental reckoning. The first test is Morgan Stanley’s mid-November window for a bottom. The companies best placed to survive will be those that can demonstrate that AI is not just a story but a business.