Benzinga and Yahoo Finance LLC may earn commission or revenue on some items through the links below.

Big Tech has spent years pouring record sums into artificial intelligence. Now veteran strategist Ed Yardeni says hitting the brakes could be bullish.

“I’ve had the view that a slowing down in the AI build-out is actually a bullish story,” the Yardeni Research president told CNBC Thursday.

He was explaining a striking split in Wednesday’s post-midterm rally, when the Dow surged 1,569 points, its best day in more than 18 months. Microsoft Corp. (NASDAQ:MSFT) gained 6.7% and Amazon.com Inc. (NASDAQ:AMZN) rose 6.1%, while Nvidia Corp. (NASDAQ:NVDA) fell 3.2% and CoreWeave Inc. (NASDAQ:CRWV) sank 9.9%.

Why Spending Less Could Mean More Cash

Yardeni pointed to concerns that hyperscalers and smaller cloud providers have been building data centers and buying chips too quickly, “creating a lot of havoc,” without any guarantee of adequate returns.

Amazon illustrates the problem. The company generated $168.3 billion in operating cash flow over the past year, yet reported negative $42.4 billion in free cash flow as AI investment accelerated.

It expects approximately $220 billion in capital expenditures this year.

If hyperscalers slow their expansion, “their free cash flow will be less negative, maybe even turn positive,” Yardeni said.

Fewer new data centers would mean less cash going out the door, potentially improving returns on existing investments.

Political Speed Bumps

Texas could become a major obstacle to the AI build-out. Gina Hinojosa won the governor’s race Tuesday after promising to halt new data-center permits until lawmakers pass regulations.

Texas accounts for nearly 30% of planned U.S. data-center additions in 2028, SemiAnalysis estimates.

Politics isn’t the only obstacle. Yardeni also pointed to tighter financing, labor shortages and semiconductor constraints.

“I think these are all putting speed bumps in the AI build-out. And that’s a good thing,” he said.

He also welcomed the role of “bond vigilantes,” investors who demand higher returns for lending to risky borrowers, arguing that some “neo-clouds” shouldn’t be getting cheap credit.

But there’s a catch: Amazon, Microsoft, Alphabet and Meta still plan more than $700 billion in capital investment this year. Slower construction could also constrain future revenue if AI demand keeps growing.

On Polymarket, traders see an 8% chance of an AI bubble bursting by December 31, rising to 23% by June 30, 2027. Yardeni’s argument points to a gentler path: a slower build-out that keeps overspending from turning into a full-blown downturn.