The tape over the past week looked like a thesis breaking. The Nasdaq fell more than 4 percent on June 4, its worst session since April 2025. Nvidia shed 6 percent, Broadcom nearly 8, and AMD and Intel each gave up better than 10 in a single day. Then it bounced, then it sold off again, then it stabilized as oil rolled over. Micron jumped 9.9 percent one day, fell 13.3 percent two sessions later, opened up 4.2 percent the next and closed down 7.6. That is not a market digesting new information about artificial intelligence. That is a market that has lost its grip on price.
Nothing in the demand data cracked. Nvidia’s most recent quarter put up $82 billion in revenue, up 85 percent year over year, on what management called the fastest product ramp in the company’s history. Super Micro disclosed roughly $39 billion in AI server orders from more than 20 customers in a matter of weeks. Hyperscaler capital expenditure is tracking above $690 billion for 2026. Alphabet raised a record $85 billion in equity to keep building. The orders are real, the spending is committed, and the chief executive at the center of the trade called the drop a buying opportunity. If the AI demand story were ending, it would show up in the order books and the capex guidance first. It has not.
What moved was the discount rate and the mood, which are not the same thing as the business. The proximate trigger was a hot May payrolls report that pushed the unemployment rate to 3.4 percent and revived the higher-for-longer rate narrative. Long-duration growth equities are the most sensitive assets in the market to that shift, because their value sits years out in cash flows that a higher rate discounts harder. Layer on a Middle East risk-off bid, an oil spike, and a single soft data point in Broadcom’s AI revenue outlook, and you have every ingredient for a violent repricing of the multiple without a single revision to the underlying demand. The selloff was a rate-and-sentiment event wearing an AI costume.
The whipsaw is the tell. When a name has tripled in a year, as Micron has, there is no valuation cushion left to absorb bad news, so price swings on positioning rather than fundamentals. Double-digit daily moves in both directions are the signature of a market where the marginal buyer has become exquisitely sensitive to entry price and indifferent to it on the way up only as long as momentum holds. That is fear overtaking greed in real time. The greed built the multiple; the fear is now testing how much of it was real.
The supply side of the same overheating has been hiding in plain sight. The flood of AI paper into the market — Alphabet’s $85 billion, Super Micro’s $7 billion, the $2.3 billion CoreWeave’s founders have sold and the $5.5 billion its anchor investor Magnetar has unloaded — is not a demand signal. It is the issuers and the early holders rationally monetizing into strength. More equity arriving into a tape where the buyer has turned picky on price is precisely the mechanism by which an overheated market cools. The dilution and the insider selling did not cause the correction. They are the same phenomenon viewed from the supply side.
The distinction that matters for positioning is between a derating and a demand break, because they have opposite endings. A demand break impairs the cash flows permanently and the multiple compression is justified. A valuation reset compresses the multiple while the cash flows keep compounding underneath it, which means the business grows back into the price from a lower entry. Everything in the current data points to the second. The earnings are accelerating, the orders are stacking, the capex is funded, and the only thing that fell was what investors were willing to pay for a dollar of it today.
Fear took the multiple. It did not touch the demand. Those two numbers diverged this week, and they will reconverge — the only open question is the price at which.