The artificial intelligence (AI) rout that accelerated towards the end of last week on the back of the Democrats winning the US mid-terms and concerns they could halt data centre build-outs has been some time coming, analysts note.
Democrats won control of both chambers of US Congress in the 3 November mid-terms, with 56% of voters calling AI and data centres a crucial or major factor in their congressional vote, according to an NBC exit poll.
UBS analysts estimate that around 30% of expected US data centre capacity additions in 2028 could be at risk from potential construction pauses, Reuters reports.
The tech-heavy Nasdaq closed last week 3.99% down over the past five days, with the Semiconductor Index – otherwise known as SOX – dropping 12.75% as of close of trade in the US on Friday, although it had recovered by 4.21% on the last trading day of last week.
Trouble brewing
Worries about an AI investment boom predate last week’s mid-term elections, says Brian Levitt, chief global market strategist and head of strategy and insights at global asset manager Invesco. “Investors have increasingly questioned the enormous amount of capital being committed to AI infrastructure, particularly as interest rates moved higher during much of the year,” he writes.
Strate Street Investment Management also notes that, “while the election was not the sole cause of the AI correction, it added political and regulatory uncertainty that accelerated a repricing already underway”.
Global semiconductors were already in a 20% drawdown before the mid-term elections, while broad US equities had held up comparatively well, Strate Street notes. It adds that such drawdowns are not unprecedented, with semiconductor equities having experienced considerably deeper ones in 2022 and 2025.
“Higher borrowing costs naturally raise the hurdle for new projects. At the same time, semiconductor and memory companies have been investing aggressively to satisfy expectations for extraordinary compute demand,” Levitt notes.
“The scale and financing of AI capital spending, elevated earnings expectations and questions around monetisation had raised the hurdle for further gains. Against that backdrop, investors had become less forgiving of challenges to the AI growth story,” says Strate Street.
Bills, bills, bills
Dr Robert Castellano, a high-tech industry analyst with more than 40 years’ experience, notes that companies once praised for massive AI infrastructure spending are now being criticised over depreciation, negative free cash flow, borrowing needs and uncertain returns.
“The issue is not whether AI is a legitimate technological transformation. It is whether Wall Street confused capital expenditure announcements with profitable, financed and ultimately utilised computing capacity,” writes Castellano.
The expansion is concentrated among Microsoft, Amazon, Alphabet and Meta, with Oracle also playing an increasingly important role, he states.
“The market is moving from valuing announced AI capacity toward questioning the economics of operating it.”
Higher financing costs can undermine data centre economics despite contracted demand, Castellano notes, while operators’ financial troubles do not necessarily signal weaker demand for semiconductor suppliers.
Power play
Levitt says investors fear Democratic victories could slow data centre construction, particularly after candidates sceptical of more development won in states including Texas, Ohio and Georgia. “That’s a legitimate new uncertainty. It could slow the pace of the build,” he writes.
The election brought concerns around power demand, infrastructure and the impact of data centres on local communities into sharper focus, says Strate Street.
Bianca Botes, MD of Citadel Global, says in a weekly note published last Friday that the first market reaction to the Democrats’ win in the US mid-terms split AI stocks in two on Wednesday, with AI infrastructure companies falling, while the big tech companies paying for it rose.
“Investors expect a Democratic Congress, together with candidates opposed to data centre expansion, like the candidates who won in Texas, to slow new construction. That was read as negative for AI infrastructure suppliers,” writes Botes.
Yet, Levitt believes that a reset isn’t the end of the AI story. “None of this changes our longer-term views that demand for compute will continue to grow and that AI represents a powerful structural investment theme.”
Levitt says the speed at which capacity gets built may change. “A more measured pace could ultimately make the AI build healthier and more sustainable.”
Price check
Invesco does not view the recent decline in AI stocks as a bubble bursting. “Lower prices and more reasonable valuations are improving the longer-term opportunity set,” says Levitt.
“The long-term demand for AI and compute remains intact. The price investors are being asked to pay for that opportunity is simply being reset.”
Strate Street Investment Management notes the “market reaction overstates how much the election changes the longer-term trajectory of AI development”. It adds that the “election may add to near-term volatility, but it’s unlikely to change the long-term AI investment case”.
Share
