AI-related stocks sold off on June 23, 2026, as investors reconsidered a question that's shadowed the sector's two-year rally: whether the staggering sums being spent on AI infrastructure will ever pay off. The decline extended a pattern that started earlier in June, with tech and semiconductor stocks sliding again on June 23–24 as selling pressure spread from U.S. markets into Asia.

The scale of the spending behind this anxiety is concrete and enormous. The hyperscalers — the few cloud giants operating the world's largest AI infrastructure — were projected to spend over $800 billion in capital expenditure in 2026, according to Reuters. To put that in perspective: that's roughly what every non-technology company in the S&P 500 combined was spending on capex. Capital expenditure, or capex, is the money companies invest in building or upgrading physical and digital assets — factories, data centers, software systems. That comparison matters because it shows how concentrated this buildout has become. A handful of firms are making one of the largest coordinated capital allocation decisions in corporate history. And the revenue base that will eventually need to justify it doesn't yet exist at scale.

This June 23 selloff wasn't the first sign of trouble. On June 8, tech stocks declined after Federal Reserve policy expectations shifted, rattling the AI trade. Asian markets experienced turbulent trading in the week that followed, driven largely by semiconductor selloffs in the U.S., per Reuters. The pattern is consistent across these episodes: sentiment turns on any signal — whether from the broader economy or from the Fed — that suggests the long-term growth assumptions baked into AI stock valuations might need to be trimmed.

The Federal Reserve Is Watching — and Skeptical

The Federal Reserve has been tracking the AI buildout closely, and its tone is cautious rather than bullish. At his June 17 press conference, Fed Chair Warsh said the central bank would monitor the pace, reach, and economic impact of general-purpose technologies including AI, according to the FOMC press conference transcript. A general-purpose technology is one with broad potential to reshape how the entire economy works — like electricity or the internet. That careful language — pace, reach, measured impact — signals the Fed isn't ready to assume AI will deliver a productivity bonanza automatic enough to guide interest rate decisions.

The Fed's own research found that just over 20% of U.S. firms expected to use AI in the first half of 2026, per a Federal Reserve note published in April. That's intent to use, not proven gains in how much work gets done. The distance between companies adopting AI and companies actually making more money or producing more output with it is where the bullish case for AI stocks looks weakest.

On employment, Fed Governor Cook noted in a May 27 speech that unemployment was tracking in line with estimates of the natural rate — the joblessness level consistent with a stable economy. That tells you labor supply and demand are roughly balanced. This matters in two ways for the AI story: a balanced labor market creates no urgent pressure to adopt AI quickly to fill a worker shortage, and it also means the immediate deflationary boost that AI bulls often promise — prices falling because workers become more productive — isn't guaranteed.

Governor Barr went further in February, saying plainly that AI could seriously disrupt labor markets and hurt some workers in the short term, in a speech on February 17. The fact that policymakers are already flagging labor market risks before AI adoption has even crossed 25% among firms tells you how seriously they're thinking through the downside scenarios.

What the Market Is Actually Recalibrating

The core tension is simple. Hyperscaler capex is a present-day cash drain of historic size. The productivity and revenue gains that would justify it remain firmly in the future — at best a probability, not a certainty. When interest rates are low and growth is hard to find, markets place high value on those future gains. When rates stay elevated, or when policy signals suggest they'll stay that way, the discount rate applied to future earnings goes up, and stock valuations fall. This is a mechanical shift, not a judgment call.

The June 23 selloff reflected exactly this mechanism: investors recalibrating the odds that $800 billion in annual infrastructure spend will actually translate into earnings on a timeline that supports today's stock prices. Whether it does hinges on variables — how quickly companies actually adopt AI, what regulators decide to do, which competing AI systems prove most efficient — that no one can know for certain. What is known is the capex itself. It's committed. It's enormous. And the market is increasingly reluctant to treat its payoff as guaranteed.