AI’s trillion-dollar question: What happens if the boom slows?

The AI trade is not cracking yet. But one of its central assumptions that model capabilities, infrastructure spending and demand will keep rising without interruption is facing a fresh test.

Anthropic chief executive Dario Amodei on September 12 urged AI companies to slow the development of advanced models so that safety measures could catch up.

“We must slow the pace at which we improve the capabilities of AI models,” Amodei wrote in an essay titled We Must Pace the Frontier.


Also Read: Dario Amodei, Sam Altman, Elon Musk: AI’s biggest rivals suddenly agree the race needs brakes

OpenAI chief executive Sam Altman and Elon Musk, who runs xAI, backed the call. US President Donald Trump dismissed the warnings about AI risks as exaggerated.

The debate comes with large valuations at stake. OpenAI was valued at $852 billion in March and Anthropic at $965 billion in May. xAI was valued at $250 billion when SpaceX acquired it in February, creating a combined entity worth $1.25 trillion.

Anthropic is also moving ahead with a potential initial public offering, while OpenAI has pushed its listing plans beyond this year. That puts greater scrutiny on how investors value companies whose projected growth depends on increasingly powerful models and sustained demand for computing capacity.

A market priced for no speed limit

AI-linked stocks fell sharply on September 14 as investors reacted to the industry’s warnings, although rising oil prices and Treasury yields also weighed on markets.

The S&P 500 closed 0.48% lower, the Nasdaq Composite lost 0.56% and the S&P 500 information technology index fell 1.68%. The Philadelphia Semiconductor Index tumbled 5.9%.

The selloff also showed how concentrated the AI trade has become. Since the S&P 500 bottomed in late March, investors have put $52 billion into information technology exchange-traded funds, compared with just $4 billion into the rest of the market, according to Bloomberg, citing data compiled by Baird Strategas.

Also Read: Global AI stocks fall as industry chiefs call for slowing development

Almost $13 billion flowed into the $481 billion Invesco QQQ Trust in August, marking its biggest monthly inflow on record. That concentration makes markets vulnerable to even a small change in assumptions about AI growth.

“AI semiconductor and infrastructure stocks have long priced in an uninterrupted capex boom, leaving virtually zero margin of error for an industry-imposed speed limit,” James Humphries, managing partner at Mindset Wealth Management, told Reuters.

The unlikely winners of a slower AI race

The selloff also revealed a widening divide within technology. Companies supplying chips, data centres and power systems for the AI buildout came under pressure. Meanwhile, software and IT services companies previously seen as vulnerable to AI disruption got a breather.

Europe's STOXX 600 Technology Index fell 2.1% on September 14, with French chipmaker Soitec dropping 12.5%. But shares of software companies including Sage, RELX and Capgemini rose between 5% and 7.5%.

“These stocks were victims of the SAASpocalypse, on fears AI would wipe out their businesses,” Chris Beauchamp, chief market analyst at IG, told Reuters.

Investors seem to be betting that slower advances in frontier AI could give software companies more time to adapt before the technology erodes their revenues or makes parts of their products obsolete.

India saw a similar reaction on September 15. The Nifty IT index surged as much as 5.2%, with HCLTech, Infosys and Tata Consultancy Services among the leading gainers.

Also Read: AI restraint debate lifts outlook for India tech services stocks

India’s $315 billion IT services industry has been hit particularly hard by concerns that AI could reduce billable hours and force companies to overhaul their pricing models. The Nifty IT index remains about 21% lower this year, more than twice the decline in the broader Nifty 50.

The rally does not necessarily mean investors are walking away from AI. It suggests they are reassessing which parts of the technology industry benefit if the AI race takes longer than expected.

Capex is the fault line

The biggest risk lies in the capital spending that has powered the AI boom. Microsoft, Alphabet, Amazon, Meta and Oracle are expected to spend about $795 billion in capital expenditure this year and nearly $1.08 trillion in 2027, according to BofA Global Research.

That spending has flowed into semiconductors, networking equipment, data centres and power infrastructure. On September 14, fibre-optic cable maker Corning fell 13.7% and power systems company Vertiv dropped 7.8%.

A slowdown in frontier model development would not automatically bring those investments to a halt. Companies could continue spending to meet existing demand, deploy current models and build safety systems. But investors fear it could weaken the assumption that every new generation of models will require substantially more computing capacity.

There is little evidence so far of an actual spending pullback.

“I need to see something concrete that, in fact, there is a slowdown versus just talk,” Chuck Carlson, chief executive of Horizon Investment Services, told Reuters.

Stronger safety standards could even make long-term investments easier to justify, according to Erik Kratz, chief investment officer and co-head of wealth at Arena Private Wealth.

“The buildout doesn’t stop because the CEOs asked for guardrails. If anything, a credible safety framework makes the long-duration capex easier to underwrite,” Kratz told Reuters.

Too strategic to slow

Increase in regulation could slow model development and put pressure on companies valued for rapid growth. But greater government involvement could also produce more spending if AI is treated as a strategic and national-security priority.

Goldman Sachs expects cumulative global AI spending through the end of the decade to reach as much as 5% of global gross domestic product.

Also Read: Governments worldwide are way behind on AI, says Bill Gates

The US and China have strong incentives to continue advancing the technology even as concerns about its risks intensify. Any unilateral slowdown could allow a rival country or company to catch up, making a broad international agreement difficult to achieve.

Continued spending, however, does not guarantee that current valuations will hold. The question is whether investors have already priced in too much growth and too little risk.

European Central Bank President Christine Lagarde has said a correction in AI valuations is “entirely possible”, pointing to high asset prices and the circular financing arrangements emerging across the industry.

Can the AI boom slow without breaking?

AI has survived market scares before, including the emergence of China’s DeepSeek in early 2025 that raised doubts about how much computing power and capital would be required to develop competitive models. However, that selloff was short-lived.

While DeepSeek challenged the cost of building powerful models, Amodei’s intervention questions whether the pace of AI development itself can or should continue unchecked.

The recent market turmoil brings what is at stake into the spotlight. Infrastructure suppliers were hit because their valuations depend on the speed and scale of the AI buildout, while software and IT services companies rose because slower development could delay the disruption they fear.

If safer AI means more regulation and slower development, some of the sector's richest valuations could come under pressure. But if AI becomes too strategically important for governments to let up, the spending boom may simply take a different form.