Three sessions, three markets, one U-turn. Friday, the Nasdaq Composite fell 4.18 percent, its worst single day since April 2025. Monday morning, Seoul’s KOSPI index opened down more than 8 percent and triggered a circuit breaker within minutes. Monday night, U.S. stock futures turned higher, with the Nasdaq contract up about 1.5 percent and Nvidia rising more than 2 percent in premarket trading. The whipsaw left traders describing the week as a test of how much pain the AI trade can absorb.
The rebound started with a headline. Iran announced it had ended its military operations against Israel, according to reports carried by CNBC, and the easing of geopolitical risk was enough to flip the mood in a market that had spent three days selling everything. Futures turned green before the announcement was an hour old, and the chip names that led the decline were the first to recover.
The bounce does not erase the damage. The Friday session erased more than a trillion dollars of market value from the Philadelphia Semiconductor Index in a single day, and the Monday rout in Seoul showed how the selling migrated across borders. Samsung Electronics fell 5 percent and SK Hynix dropped 2 percent before the Korean market stabilized, and the circuit breaker that halted KOSPI trading was the loudest alarm of the day.
Jeremy Siegel, the Wharton professor whose name has been attached to market forecasting for four decades, said on Monday that pullbacks of this kind “rarely mark tops.” His argument is that a crash in a handful of AI names is not the same as a market-wide top, and that the underlying earnings growth in technology remains intact. The caveat in his own analysis is the part the market is arguing about: the AI complex is trading at valuations that leave no room for the disappointment that a Broadcom-style guidance miss can produce.
The catalyst for the whole episode was Broadcom. The company’s quarterly report beat expectations but declined to raise its long-term guidance, and the market took the restraint as a signal that the AI capital expenditure cycle is maturing. Combined with a stronger-than-expected nonfarm payrolls report that reinforced bets on a Federal Reserve rate increase, the guidance decision was enough to start an unwind that no single stock could stop.
The unwind had a specific shape. Positions that had been built on borrowed money and crowded into the same AI names came out in a hurry, and the selling fed on itself until the circuit breakers and the Iran headline broke the sequence. Traders described the pattern as a standard deleveraging event: the trades that had worked for two years stopped working at the same time, and everyone tried to exit through the same door.
The valuation pressure that built the fragility has not gone away. The S&P 500’s top ten stocks trade at multiples that assume years of uninterrupted growth, and the concentration means the index moves with the fortunes of a handful of companies. Siegel’s view that the pullback is healthy rests on the assumption that the underlying earnings arrive; if they do not, the selloff resumes from a lower base.
What comes next depends on the data. The payrolls report that helped trigger the selloff points to an economy that does not need rate cuts, and the swaps market is now pricing a rate increase this year. Higher rates are the enemy of long-duration assets like AI stocks, and every strong data point between now and the Fed’s next meeting keeps the pressure on.
For the chip makers at the center of the storm, the fundamentals have not changed in a week. Demand for AI accelerators and the networking that surrounds them continues to outrun supply, and the companies reporting in the coming weeks are expected to show it in their order books. What changed is the price the market is willing to pay for that demand, and the rebasing that happened Friday and Monday was the market’s way of saying the old price was too generous.
The rebound in futures was also a vote on the fundamentals. The chip companies at the center of the storm report earnings in the coming weeks, and the order books they show will either confirm the demand story or open the second leg of the correction. Until then, traders said, the bounce is a bet rather than a conclusion.
The Iran headline gave the market an excuse to buy; the question is whether it will hold. Futures were up Monday night, but the session that matters is the one that opens with the chip stocks that led the decline. Traders said the rebound is real but shallow until the next earnings report confirms the demand story, and until then, the week’s lesson stands: the AI trade can fall faster than it rose, and the recovery will not be a straight line.
