The artificial-intelligence boom is entering a more difficult phase. After months of soaring valuations and massive spending on chips, data centres and power, investors are beginning to question whether demand, profits and public consent can keep pace with the buildout.
The warning arrived in markets on Thursday, when the Nasdaq Composite fell 5.5%, its steepest one-day decline since the panic around President Donald Trump’s “liberation day” tariffs in April 2025. The Nasdaq 100, dominated by large technology companies, has fallen about 14% from its October peak, according to Axios. The selloff reportedly erased roughly $2.5 trillion in market value across global equities, while Asian technology markets suffered further losses in early trading.
From enthusiasm to scrutiny
For much of the past two years, AI investment was treated as a self-reinforcing cycle. Demand for computing power encouraged companies to expand data-centre capacity; that expansion boosted chipmakers, cloud providers and utilities; rising share prices then made it easier to finance still more construction.
That logic has not disappeared, but its assumptions are being tested. AI infrastructure requires enormous upfront capital, long permitting processes and reliable electricity. Companies must ultimately demonstrate that software revenues can justify the cost of the servers and energy needed to deliver them. Reuters has described a broader market already strained by wars, trade tensions, high government debt and rising borrowing costs, leaving investors less willing to overlook uncertain returns.
“The key question is whether investment in technology and infrastructure can sustain growth while higher operating costs constrain spending elsewhere,” the International Strategic Analysis organisation wrote in its November outlook.
The concern is not simply that AI will fail. It is that the industry may be economically successful while still producing returns too slowly, unevenly or narrowly to support today’s valuations. Anthropic’s decision to publish financial information ahead of a highly anticipated initial public offering has made that tension more visible: public markets will demand evidence of durable revenue rather than promises of eventual transformation.
The backlash is local before it is national
Resistance to AI infrastructure is also becoming a political force. Data centres consume large quantities of electricity and water, compete for land and can raise questions about grid reliability, noise and tax subsidies. Axios reported that local moratoriums and political opposition have accumulated across the United States, threatening to slow an infrastructure buildout that some estimates place at $10 trillion.
The opposition is not uniform. Supporters argue that data centres bring construction jobs, tax revenue and strategic capacity at a moment when the United States and its allies want to reduce dependence on foreign technology supply chains. The industry also says that delaying facilities could leave the country behind China and other competitors.
Critics counter that communities should not be expected to absorb environmental and infrastructure costs for private projects whose gains may flow mainly to technology investors. Their preferred remedies include tougher disclosure rules, limits on public subsidies, new electricity pricing and requirements that operators finance grid upgrades.
Those disputes became harder for investors to ignore after the US midterm elections. Axios reported that AI-cautious Democratic candidates won gubernatorial contests in technology hubs and that voters backed ballot measures restricting data-centre construction. The results do not amount to a national rejection of AI, but they signal that the politics of permitting, power and land may become as important as the politics of innovation.
What comes next
The immediate test is whether the selloff remains a correction or becomes a deeper repricing. A resilient US economy could help: employers added 165,000 jobs in October and unemployment fell to 4.1%, according to the Labor Department data reported by NPR. Yet strong employment can also keep interest rates higher for longer, increasing the cost of financing data centres and pressuring high-growth technology valuations.
Three outcomes are plausible. First, the largest firms may consolidate their advantage, absorbing weaker projects and securing power and chips through long-term contracts. Second, utilities and governments may impose conditions that make growth slower but more publicly acceptable. Third, capital may shift away from speculative infrastructure toward applications that show measurable productivity gains.
That final question matters beyond Silicon Valley. Labour productivity rose at a 3.6% annual rate in the July-to-September period, Axios reported, suggesting that technology investment is already contributing to economic output. But productivity gains do not automatically settle distributional disputes. Workers, households and local governments will judge AI by wages, prices, electricity bills and public services—not by the market value of its leading companies.
The AI boom therefore faces a credibility test on two fronts: financial discipline and democratic legitimacy. If the industry can show durable earnings while sharing infrastructure benefits and limiting local costs, the current correction may become a healthy reset. If not, the backlash will not merely interrupt construction. It could redefine who controls the pace, location and purpose of the next technological revolution.