For two years, the answer to almost every market question was "AI." Now investors are asking a harder one: when does all that spending actually pay for itself?

The doubt has a price tag. The "Magnificent Seven" — the mega-cap group of Microsoft, Nvidia, Apple, Amazon, Alphabet, Meta and Tesla — shed roughly $2.3 trillion in combined market value in June, their largest monthly loss on record, Yahoo Finance reported. And the selling didn't stop when the month did: this week it spread to Asian chipmakers, with memory-chip makers SK Hynix and Samsung dragging down South Korea's Kospi index, according to Seeking Alpha.

Who fell, and how far

The June damage was broad but uneven. Microsoft led the decline, down about 19% on the month; Nvidia fell 10.7%, Amazon 8.8%, Meta 6.1%, Apple 5.5% and Alphabet 5.0%, per Yahoo Finance. (These are one-month moves, not one-day crashes — but for companies this large, a 19% slide erases hundreds of billions of dollars in value.) Notably, the broader Nasdaq-100 held up better, a sign that investors were rotating out of the biggest AI names rather than fleeing tech altogether.

The fear: a spending "loop"

At the center of the anxiety is capital expenditure — "capex," the money companies pour into data centers, chips and other AI infrastructure. Across the big cloud providers, that spending is up about 84% from a year earlier and now equals roughly 98% of their operating cash flow, Yahoo Finance reported — meaning nearly every dollar the businesses generate is being plowed back into AI build-out. (Capex is spending on long-lived physical assets; operating cash flow is the cash a business throws off from its core operations.)

What unnerves investors is how circular some of the revenue looks. Cloud giants are funding the very AI labs that then buy computing power back from them: Alphabet has committed up to $43 billion to Anthropic, Amazon as much as $50 billion to OpenAI, and Microsoft's arrangement with OpenAI involves a reported $250 billion of Azure services, per Yahoo Finance. Critics worry this counts the same dollars twice — an AI economy that looks like booming demand may partly be the giants paying each other. The counter-argument is that real external demand for AI is coming; the bet is enormous either way.

Meta lights the fuse

The immediate trigger this week was Meta. Its shares fell as much as 4% on Thursday — trading down about 3.7% late morning — after Wolfe Research analysts projected Meta's capex could climb to $200 billion in 2027, up from a prior estimate of $160 billion, and might require the company to raise fresh capital, Yahoo Finance reported. The analysts framed it as a long-term opportunity — Meta could sell cloud computing into an estimated $500 billion market — but the sheer size of the upfront bill spooked the market. (For its part, Meta has guided its own full-year capex to $125–145 billion; the $200 billion figure is Wolfe's estimate for a later year, not company guidance.)

Why it matters

For investors, the selloff is a repricing of risk after two years of near-unbroken AI optimism: the market is no longer taking "we'll spend whatever it takes" on faith. For the chip industry, the pain in SK Hynix and Samsung shows how quickly doubts about U.S. hyperscaler spending travel down the supply chain — if the giants trim capex, memory-makers feel it first. And for the broader economy, AI capex has become a genuine driver of growth, so a sustained pullback would be felt well beyond tech. Boursel gives no investment advice and takes no view on where these stocks go next; the takeaway is that the market has begun, for the first time in a while, to demand that the AI build-out show its math.