Micron lost 13% and $138 billion in a single session, and the reason has less to do with chips than with who actually pays for them.
The selloff hit fast. Micron Technology fell 13%, Intel dropped 9%, and AMD slid 7%, dragging the VanEck Semiconductor ETF down 5% in one trading day. This wasn't a single earnings miss. It was a sector-wide reckoning with a question that's been building for months: are hyperscalers actually getting a return on the hundreds of billions they've poured into AI infrastructure, or are chipmakers riding a spending wave that's about to crest?
The trigger, according to reporting from Seeking Alpha, was news that Meta is building out a cloud business to resell its excess AI computing capacity to outside customers. Read that plainly and it's a strange admission: a company that has spent tens of billions on GPUs now has more compute than it knows what to do with. Investors read it the same way. If Meta has spare capacity to sell, the assumption that every hyperscaler needs every chip it can buy starts to look shaky.
That's a different story from the one already circulating around Oracle. Oracle's stock slid earlier in June after the company disclosed roughly $95 billion in fiscal 2026 capex plans, funded partly by $40 billion in new debt that pushed its long-term borrowing above $100 billion, according to Yahoo Finance. Oracle's problem is company-specific: one balance sheet stretched thin to chase AI contracts. What hit chip stocks this week is broader. It's not one company's debt load. It's the market pricing in doubt across the entire hyperscaler cohort at once.
Two other forces compounded the drop. SK Hynix signaled it plans to slow expansion of high-bandwidth memory production and shift some capacity toward cheaper commodity DRAM, according to Yahoo Finance. HBM is the memory that feeds Nvidia's most advanced AI accelerators, so any hint of a supply pullback there raises questions about whether the AI hardware chain can keep scaling at the pace investors had priced in.
Then there's the Federal Reserve. Under new chair Kevin Warsh, nine of eighteen policymakers now back rate hikes in 2026, up from zero in March. Higher rates hit capital-intensive chipmakers twice: borrowing gets more expensive, and the future earnings that justify today's valuations get discounted harder. Growth stocks don't like that math, and semiconductor stocks are about as growth-priced as it gets.
The scale of the spending underneath all of this is what makes the doubt matter. Combined capital expenditure from Meta, Microsoft, Google, and Amazon is tracking toward roughly $725 billion in 2026, a 77% jump from last year, according to analysis reported by Investing.com. Yet the incremental revenue these companies can directly attribute to AI so far sits in the tens of billions, not the hundreds. That gap is the whole argument in one number.
Not every signal points down. On July 2, a $20 billion investment in TSMC's Arizona subsidiary cleared regulatory approval, and Nvidia confirmed its Blackwell architecture has reached volume production at TSMC's Phoenix plant. Power semiconductor makers also raised prices industry-wide around the same time, a sign that demand from AI and automotive customers hasn't disappeared. So this isn't a straight line down. It's a market trying to figure out which parts of the AI buildout are real demand and which parts were momentum.
For founders and investors building anything adjacent to AI, this is worth watching closely rather than shrugging off as a Wall Street mood swing. Chip stocks move first because they sit closest to actual capital commitments: an order for GPUs or HBM modules is a real dollar figure, not a slide deck. When that order flow gets questioned, the valuations built on top of it, including the ones attached to AI startups that haven't shipped a comparable order themselves, get questioned next. Nobody has answered whether Meta's spare compute is a blip or the first crack in the capex story. That's the number to watch, not the daily close.
Also read: Bitcoin holders are buying the dip even as traders brace for another leg down • Germany's Small Town Banks Are Becoming Europe's Biggest Crypto On-Ramp • SK Hynix is racing to list on Nasdaq just as the AI chip trade it built its fortune on wobbles
