The artificial intelligence trade that minted trillions in wealth over the past 18 months is showing its first serious cracks. A brutal selloff has erased more than $1 trillion in market value from semiconductor stocks in July alone, as investors question whether massive AI infrastructure spending can ever generate adequate returns.

The Selloff in Numbers

The carnage has been swift and indiscriminate. Nvidia, once the undisputed king of AI computing, has shed roughly $500 billion in market capitalization in the past two weeks. Micron Technology plunged 13% in a single session earlier this month. Intel dropped 9%. AMD, Broadcom, and Qualcomm have all suffered double-digit percentage declines.

The Philadelphia Semiconductor Index, which had surged nearly 65% in the first half of 2026, has given back a substantial portion of those gains. What began as a measured rotation out of technology stocks has accelerated into something resembling panic selling.

The Custom Silicon Threat

A key catalyst for the selloff has been the emergence of custom AI chips from the very companies that were supposed to be Nvidia's biggest customers. OpenAI is developing its own silicon for inference workloads. Amazon's Trainium and Inferentia chips are gaining traction in its data centers. Meta's MTIA chip is being deployed at scale.

Cerebras Systems, which makes wafer-scale chips for AI training, has captured contracts from several major research institutions. The message is clear: Big Tech no longer wants to be dependent on Nvidia's supply chain.

"The moat around GPU makers was always going to be temporary," noted one semiconductor analyst. "Custom silicon optimized for specific workloads was inevitable once AI reached this scale."

The Valuation Reckoning

The fundamental issue is valuation. At their peaks, AI chipmakers were trading at price-to-earnings multiples that assumed years of hypergrowth extending indefinitely. Nvidia at one point commanded a forward P/E ratio above 50, pricing in perfection.

The problem is that perfection isn't possible when your customers are actively working to replace you. Even TSMC, which just reported record quarterly results with 77% profit growth, has seen its stock struggle to find footing amid the broader selloff.

Where the Money Is Going

The rotation has been textbook. Money fleeing semiconductors has poured into industrials, financials, and defensive sectors. The Dow Jones Industrial Average recently hit record highs even as the Nasdaq stumbled, a divergence that underscores the breadth of the reallocation.

Small-cap stocks have also benefited, with the Russell 2000 outperforming large-cap indices for the first time in months. Investors appear to be betting that the broader economy can thrive even if the AI trade cools.

What Comes Next

The next major test arrives in late July when Big Tech reports second-quarter earnings. Microsoft kicks off the season on July 29, followed by Alphabet, Meta, Amazon, and Apple in quick succession. Combined 2026 capital expenditure guidance from these five companies now sits between $635 billion and $725 billion.

If AI revenue fails to show commensurate growth, the selloff could deepen. If the numbers surprise to the upside, chipmakers could stage a recovery. Either way, the easy money phase of the AI trade appears to be over.

For investors, the lesson is familiar: even the most compelling technology stories eventually face valuation gravity. The companies building AI infrastructure are fundamentally sound. The question is whether their stocks, after historic runs, still offer attractive risk-reward at current levels.