The Magnificent Seven, the group of America’s largest technology companies, ended June with a combined loss of about $2.3 trillion in market value, their worst month since 2022. The selloff was not a panic. It was a repricing, a collective decision by investors that the companies spending the most on artificial intelligence had not yet shown the returns the spending was supposed to produce.
The seven companies, Apple, Microsoft, Nvidia, Alphabet, Amazon, Meta Platforms and Tesla, drove most of the market’s gains over the past two years, and they have driven its recent losses with equal efficiency. In June, the declines came from different directions but pointed to the same question: who actually earns the money that AI is supposed to generate?
Microsoft’s slide showed the cost side of the equation. The company’s cloud margins came under pressure as AI infrastructure spending ramped ahead of revenue, and investors marked down the stock even as growth remained strong. Apple’s decline traced to the memory shortage, with the company raising prices on its hardware as component costs tripled. Broadcom, not a member of the group but the industry’s biggest seller of custom AI chips, fell about 13 percent after its earnings disappointed, a drop that rippled through the entire semiconductor complex.
The pattern is new. For two years, AI capital expenditure was treated as a virtue, and companies that spent the most were rewarded with the highest valuations. The June selloff is the moment investors started asking when the spending becomes profit, and the answers, from earnings calls and disclosures across the industry, have been slower to arrive than the spending itself.
The scale of the June losses invites comparison with 2022, when the same group lost trillions as interest rates rose and growth stocks were repriced. The 2022 decline was about the cost of capital; the 2026 decline is about the return on capital, a different question with a less certain answer.
The dispersion within the group matters. Nvidia, which sits at the center of the AI build-out, fell less than its peers, because its earnings have kept pace with its valuation. Apple and Tesla, whose AI stories are more about future products than current revenue, fell more. The differences suggest investors are not abandoning AI so much as becoming discriminating about which companies profit from it.
The selloff has also shifted the market’s attention to the companies selling the inputs. Memory makers, power-chip suppliers and data-center landlords have all been repriced as investors look for the businesses that collect revenue from AI regardless of which model or platform wins. This rotation, from the companies spending on AI to the companies selling to them, is one of the clearest moves of the year.
Valuations have come down with the prices. The group now trades at a lower multiple of forward earnings than it did at the start of the year, and some of the most beaten-down stocks screen as cheap by their own historical standards. Cheapness alone has not stopped the selling, because the uncertainty is about earnings two and three years out, not about the current quarter.
What would restore confidence is evidence: cloud revenue accelerating, AI products converting users into paying customers, margins stabilizing as infrastructure comes online. The companies themselves have pointed to such evidence arriving in the second half of the year, and the next earnings season will be the first real test.
The second half of the year will test whether the June repricing holds. The largest cloud companies report earnings within weeks, and each report will be read for the same signals: how fast AI revenue is growing, how much of the capital spending is converting to profit and whether margins are stabilizing. The companies have guided to improving economics as new data centers come online, and the market’s willingness to believe them will determine whether the group recovers.
The comparisons to earlier booms are unavoidable. In 2000, investors paid for connectivity that took years to become profitable, and the correction punished even the companies that eventually succeeded. The current selloff is far smaller than the dot-com collapse, and the AI companies involved have real earnings, but the pattern, money spent ahead of revenue, is the same.
The rotation has beneficiaries. Memory makers, power-chip suppliers, data-center landlords and the companies selling the picks and shovels of AI have been re-rated as investors look for businesses that collect revenue regardless of which model wins. The divergence between the companies spending on AI and the companies selling to them is now one of the market’s defining trades.
The question at the center of the June selloff, who earns the profit from AI, has not been answered. It has simply been asked, in public, at scale, by every investor who sold a large technology stock this month. Until the earnings data answer it, the $2.3 trillion loss will stand as the market’s way of saying the bills have arrived before the payoff.
