AI Chip Trade Narrows to Nvidia as AMD, Marvell, Intel and Micron Sell Off

High-multiple AI semiconductor names sold off sharply while the trade concentrated into Nvidia and Broadcom. AMD fell about 8%, Marvell about 7%, Intel about 6% and Micron more than 5%, with the iShares Semiconductor ETF down roughly 5%. The move reads as a valuation rotation rather than a demand signal: Intel sold off despite Data Center and AI revenue up 59% in Q2, and the names falling hardest are the ones carrying the highest multiples and largest year-to-date gains.

The AI semiconductor trade narrowed sharply toward NVDA, with the rest of the complex selling off hard. AMD fell about 8%, Marvell about 7%, INTC about 6% and MU more than 5%. The iShares Semiconductor ETF was down roughly 5%, a smaller move than the individual names, which is itself the tell: this was concentrated in the high-multiple end of the group rather than broad semiconductor weakness.

The evidence points to a valuation rotation rather than a demand problem. Intel reported Data Center and AI revenue up 59% in Q2 and sold off anyway. What separates the decliners from the holdouts is what investors are paying for the earnings, not the earnings themselves. Nvidia trades near a 30x trailing multiple and Broadcom around 63x, while Marvell sits near 60x, AMD near 150x and Arm near 289x. Capital moved toward the names where the AI thesis is least stretched relative to what is already in the price.

The year-to-date positioning explains the asymmetry of the move. AMD entered the session up roughly 112% for the year, Marvell up about 106%, Intel up about 133% and Arm up about 122%, against Nvidia up only about 6% and Broadcom about 10%. The laggards on the year were the safe havens on the day. That is a crowded-trade unwind: the stocks with the most embedded expectation and the largest accumulated gains gave back the most, while the sector ETF remains up about 61% year to date.

Micron carries a separate overhang, with reports of softening memory-chip demand and pricing pressure adding a fundamental leg to its decline that the logic names do not share. What to watch is whether this holds as a durable re-rating or resolves as a single-session rotation. The distinguishing evidence will be whether the high-multiple names recover on their own earnings prints, which would argue rotation, or whether the compression persists through good results, which would argue the market is repricing how many winners the AI buildout actually supports.

Related Stocks

Powered by SentiSense - Intelligent Market Analysis