Business 4 min read By Arthur Ellington
Investors on Edge as AI Safety Warnings Threaten Data Centre Spending Boom
Wall Street is reassessing the AI infrastructure trade after frontier lab leaders called for slower, safer model development, splitting chipmakers from platform giants and drawing a sharp political response from the White House.
Investors are growing nervous about the scale of spending on artificial intelligence infrastructure after a series of warnings from industry leaders that frontier model development may need to slow down. The shift in tone has unsettled a market that has priced in years of uninterrupted growth in data centre construction, advanced chip demand and the wider AI supply chain.
The immediate trigger was a coordinated call from some of the most prominent figures in the sector. Anthropic chief executive Dario Amodei urged the industry to «pace the frontier», while OpenAI chief executive Sam Altman and SpaceX chief executive Elon Musk have backed stronger safety measures as frontier models demonstrate increasingly powerful cyber capabilities. OpenAI has called for mandatory, capability-based national AI safety requirements, and Amodei has proposed independent evaluators, industry coordination and deeper government oversight as models grow more powerful.
The market reaction was swift and uneven. Nvidia and other chipmakers fell more than 3 per cent as investors contemplated the possibility that frontier labs could deliberately slow model development, and with it the massive infrastructure buildout designed to support it. Yet Meta and Alphabet rose, exposing a strange divide inside the AI trade. Cybersecurity firms such as CrowdStrike jumped 15 per cent or more, as traders bet that tighter safety rules would drive demand for protection rather than reduce it.
Analysts argue that regulation may ultimately entrench the position of the largest AI companies. Gil Luria, head of technology research at D.A. Davidson, said rules written with significant input from OpenAI and Anthropic could turn safety requirements into a moat around the companies already at the frontier, because only they can afford expensive safety teams, lawyers and government compliance staff.
The political dimension has added to the uncertainty. President Donald Trump has rejected the idea of slowing AI development, warning Americans not to «kill the Golden Goose» — his term for AI data centres — and arguing that China would be the primary beneficiary of any slowdown. In a post on Truth Social, he said the only guardrails AI needs are a «strong and smart» president, and claimed his administration already holds tremendous criminal and regulatory power over the companies involved.
That intervention sits awkwardly with public opinion. A poll from the Annenberg Public Policy Center at the University of Pennsylvania found that 61 per cent of Americans oppose a new data centre in their community, including 54 per cent of Republicans. The White House has spent much of the past year deepening its involvement in the AI economy. A June executive order directed national-security agencies to develop benchmarks for advanced models' cyber capabilities and to determine when a system qualifies as a «covered frontier model», alongside a voluntary framework for government access before broader release.
The administration has also floated taking stakes in AI companies themselves, extending an interventionist playbook already used in semiconductors and steel. When it allowed Nvidia to resume sales of advanced chips to China, the federal government arranged to take a share of the revenue. For investors, the question now is whether safety concerns, political pressure and community resistance will slow the spending cycle that has underpinned the AI rally — or merely redirect it towards the largest players and the security firms that guard them.
