AI-Linked Stocks Fall as Chipmakers Lead Selloff After Record Rally

Artificial intelligence continues to develop very quickly, while the cost of some computing power—that is, the price of renting the resources needed to run models—is falling. Launching flagship models from OpenAI and Anthropic, however, remains significantly more expensive. At the same time, demand for AI technologies continues to grow.

Over the past year, shares of most chipmakers critical to AI infrastructure have risen by hundreds of percent. Investors have now begun partly taking profits and looking more cautiously at further growth.

On Tuesday, the technology-heavy Nasdaq 100 fell 3.3%, while the broader S&P 500 lost 1.4%. The selloff intensified after a decline in the South Korean market, which made investors nervous about the overvaluation of chipmakers’ shares.

Shares of Micron Technology, which had risen sharply amid the AI boom, fell 13.2%.

Pressure on the technology sector intensified for different reasons. On Monday, the decline was led by Alphabet, Google’s parent company, after reports that it was losing key AI specialists to competitors.

At the same time, AI infrastructure is entering a new phase. Companies increasingly understand that adopting such technologies costs real money, but many still do not fully understand what exactly those expenses are going toward.

According to a KPMG report published on Wednesday, only 26% of 204 surveyed executives at U.S. companies said they have full visibility into the operating costs associated with AI.

“Costs can get out of control very quickly,” Rahsaan Shears, KPMG’s head of business transformation with AI, told Axios. According to him, the company’s clients report that they are spending allocated budgets faster than expected.

For example, according to media reports, Uber exhausted its 2026 budget for AI coding tools in just four months and is now limiting such spending by employees.

Until recently, companies were mostly testing AI and encouraging employees simply to use the new tools. Now, Shears says, they are trying to scale AI across the entire organization.

At the same time, the cost of computing power is indeed falling—at least outside the most expensive advanced models from Anthropic and OpenAI.

Some market participants believe that shares of companies linked to AI are falling in line with the declining cost of computing. Others point out that not every company needs the most expensive models. As Deutsche Bank’s Jim Reid wrote in a recent note, many businesses need “a reliable workhorse, not a supercar.”

More and more companies will ask whether paying a premium for the most advanced AI models is justified.

However, demand for computing power for AI still exceeds available supply by at least five times, and possibly by ten times, says Mandeep Singh, global head of technology research at Bloomberg Intelligence.

According to him, lower computing costs may affect expectations for individual memory producers such as Micron, but overall it does not change the valuation of the largest cloud and technology companies. Moreover, cheaper computing could benefit them, since they are the ones spending enormous sums on it.

“I’m not in the camp that thinks the key drivers have changed materially,” Singh said.