The reaction confused the White House. “With a great Jobs Report, like just announced, stocks should go up, not down,” President Donald Trump posted on Truth Social Friday morning. “That’s the way it was for 200 years. Growth does not mean inflation!”

Demand is running unusually hot—overall spending in the economy grew 5.9% last quarter, well above the roughly 4% pace that is consistent with the Fed’s inflation target—while the economy’s capacity to meet that demand is being short-circuited from multiple supply-chain crises. The war in Iran has shut off most oil shipments through the Strait of Hormuz, sending gas prices skyrocketing. Tariffs have raised the cost of imported goods. And the labor force is slowly shrinking, partly due to the Trump administration’s immigration crackdowns. 

The combination of too much demand chasing too little supply is the textbook definition for inflation—and is why consumer prices have now run above the Fed’s 2% target for more than five years. It is also why growth itself has changed meaning for markets; when supply is abundant, a strong jobs report signals more output and more profits, more room to grow. When supply is constrained, it signals more spending power pressing against that same ceiling of inflation. So if the labor market is hot, it’s evidence that the Fed has no room yet to cut rates, and might in fact have to raise them. 

AI is at the center of the insatiable demand

Driving all this is a historic appetite for capital. The AI buildout has hyperscalers on pace to spend close to a trillion dollars a year on data centers and chips. Even Alphabet, among the richest companies on earth, raised $85 billion from the stock market. The issue is everyone and everything wants money now, but they are stuck with borrowing costs that keep climbing. Friday’s report pushed it higher still: The 10-year Treasury yield jumped to 4.54% while the 30-year crossed 5%, and futures markets priced better-than-60% odds of a rate hike by year-end.

Why rising yields fall hardest on technology stocks is a timing question. “You can think of equities as long-duration assets,” Venu Krishna, head of U.S. equity strategy at Barclays, told Fortune.

What he means: A stock’s price isn’t the market’s estimate of future profits, it’s the market’s estimate of all the profits the company will ever earn, translated into today’s dollars. And that translation has to run through interest rates. A dollar of profit arriving in 2031 is worth less than a dollar arriving now, because money in hand can be parked in Treasuries and earn some yield in the meantime. The higher that yield, the worse that penalty is. When the 10-year pays 3%, waiting is cheap, and the future is worth not much more than face value. But when the yield is 4.5% and rising, every year of waiting costs more, and distant profits shrink fast.

So AI companies, many of them valued on revenue they are expected to earn three to five years from now, sit at the far end of that spectrum. “Your discount factor is going up, and it shows up via pressure on valuation multiples,” Krishna said. That is why hot memory chip companies like Marvell and Micron fell 9% on Friday while the Dow slipped just 0.3%, and investors rotated into the classics—Coca-Cola and Colgate-Palmolive—companies whose profits arrive now, not in five years.

Barclays’ research, Krishna said, has identified that the dynamic becomes especially acute as the yield approaches 5%. At 4.54%, “we are in the warning zone, but just in the warning zone,” Krishna said. “Five percent is more of a clear level. But as we start approaching it, the market starts pricing that risk.”

For nearly a year after last spring’s tariff selloff, retail investors decided to sit out the market recovery and came to regret it. So with this recent boom, they’ve come rushing back, trying to buy up equities while the systematic funds are also at full exposure. Barclays calls the result an asymmetric risk-reward: When nearly everyone who might buy has already bought, any good news that comes no longer drives new demand, while bad news has a market full of potential sellers.

Secondly, there’s froth. Barclays says that markets are more euphoric than ever, tracking it through an index that scans the options market across roughly 700 stocks for the signs of speculative chasing. In data going back through the dotcom era, about 7% of stocks show that pattern at any given time. The share recently climbed to roughly 10%, Krishna said, and the last time it ran meaningfully higher, reaching about 14% early this year, the March selloff followed.

Friday’s selloff, in Krishna’s view, doesn’t spell the end of the AI trade. Rather, it marks a sign of discipline in the markets. The hyperscalers, trading near 26 times forward earnings and cheaper than they were in January despite stronger profits, would be “a huge buying opportunity” if they fell further, he said. The pressure concentrates instead on companies whose valuations rest mostly on the future—“not because those business models are falling apart,” Krishna said, but “because the discount factor is going up.”

In fact, the repricing might come at a welcome moment for the IPO pipeline. Friday is set to start off a hot IPO summer, with SpaceX debuting at a $1.77 trillion valuation, the largest IPO in history—priced, like almost everything in the AI industry, on years of growth yet to come. Anthropic, having filed its confidential prospectus already, is not far behind, and OpenAI is in the rearview mirror; analysts started to worry that there might be too much supply of stocks and not enough buyers. Friday’s selloff eases those worries a bit. 

When asked whether it was surprising that a rally that withstood a war and an oil shock yet finally buckled on good news, Krishna was matter-of-fact: “It doesn’t surprise us.”