A trillion-dollar semiconductor selloff shows investors are demanding clearer proof that AI demand can justify stretched valuations.
The artificial-intelligence trade has been powerful enough to lift indexes, reshape capital spending plans and turn semiconductor earnings into macro events. This week, it showed its other side.
U.S.-traded chipmakers lost more than $1 trillion in market value on Friday, according to Reuters, as investors sold AI-linked names including NVIDIA, Micron, AMD, Marvell and Broadcom. The PHLX semiconductor index suffered one of its sharpest one-day declines since the 2025 tariff shock, while the broader market was already under pressure from a stronger-than-expected jobs report.
The selloff was not a rejection of AI as a technology. It was a repricing of how much certainty investors are willing to assign to future AI revenue. Broadcom’s earlier report had raised doubts about whether custom AI chip demand was matching the most aggressive expectations. When that concern collided with higher-rate fears, the sector’s valuation premium became harder to defend.
Semiconductors sit at the center of the AI economy because every frontier model, inference product, cloud contract and data-center expansion ultimately flows through chips, networking and power. That position has made the sector a proxy for the whole AI buildout. It also means any crack in demand assumptions can ripple quickly across indexes, ETFs and portfolio risk models.
The market structure amplifies the move. Many investors own the AI theme through broad technology funds, semiconductor baskets and passive index exposure. When the same mega-cap and chip names dominate multiple vehicles, selling pressure can become self-reinforcing. A move that begins with one company’s guidance or one macro data point can become a factor-wide unwind.
Still, the deeper question is about timing rather than destiny. AI infrastructure spending remains enormous, and major cloud and enterprise customers continue to invest. But equity prices had already priced in a long runway of growth, high margins and limited cyclicality. Semiconductor investors are now asking whether the sector can behave like infrastructure while still being valued like hypergrowth software.
That question will define the next phase of the trade. If earnings keep proving that AI demand is broad, durable and profitable, pullbacks may be treated as resets. If demand becomes lumpier or financing costs stay high, the market may keep separating real AI infrastructure winners from companies that merely benefited from the glow around the theme.
Image source: i.ibb.co