The frothy rally in artificial intelligence stocks is cooling fast. Chip makers that have soared on AI euphoria are now stumbling, and the culprit is a harsh dose of reality: companies are discovering that building out AI infrastructure costs vastly more than they budgeted.
The Nasdaq 100 dropped 3.3% on Tuesday after a selloff in South Korean semiconductor stocks rattled investors. Micron Technology, which had skyrocketed over the past year, plunged 13.2%. The broader S&P 500 also slipped 1.4%. Much of the selling came after Google parent Alphabet tanked on news of losing key AI talent to rivals.
The pattern is clear: stocks that benefited most from the AI boom are now unwinding some of those explosive gains as investors get nervous about how long the feeding frenzy can last. Chip companies critical to AI processing had posted triple-digit returns over twelve months. That momentum is evaporating.
The Cost Reality Check
Here is what is happening behind the scenes. Companies racing to deploy AI across their organizations are hemorrhaging money faster than they expected. Only 26% of U.S. executives surveyed by KPMG in May said they have full visibility into their AI operating costs. The rest are flying blind.
Uber burned through its entire 2026 budget for AI coding tools in just four months. Now it limits what employees can spend. Other firms report costs spiraling beyond projections as they shift from testing pilots to deploying AI at scale across their entire operations.
Rahsaan Shears, AI transformation leader at KPMG, told Axios: "Costs can get quickly out of hand. We're hearing from clients that they're going through their budgeted amounts faster than they anticipated."
What started as companies casually piloting AI to see if it worked has turned into massive, sustained spending to roll out the technology company-wide. That spending is real money with real constraints.
There is also a pricing question emerging. The cheapest AI models are getting cheaper. But companies that want access to frontier models from OpenAI and Anthropic still face premium costs. Not every business needs the supercar version. Some just want the reliable workhorse. As Deutsche Bank analyst Jim Reid noted, many firms will soon start asking whether paying the frontier premium is worth it.
Meanwhile, demand for AI computing power still vastly outstrips what is available. Bloomberg Intelligence estimates demand exceeds supply by a factor of five to ten times. That dynamic keeps the door open for companies selling the chips and infrastructure that power AI.
Mandeep Singh, global head of technology research at Bloomberg Intelligence, sees the current selloff as temporary noise. Lower compute costs might actually help hyperscalers like Microsoft and Google that are spending hundreds of billions on AI infrastructure. Cheaper processing power means their investments go further. "I'm not in the camp who believes that the drivers have changed meaningfully," Singh says.
Still, the market is sending a signal. The easy gains in AI stocks may be behind us. Companies are waking up to what this buildout actually costs, and investors are pricing that reality in.
Author James Rodriguez: "The AI gold rush just hit a wall of actual spreadsheets, and that's where the real story begins."
Comments