The AI Trade Is Rotating: -24% in Semis, Money Slamming Into These 4 Sectors

Broadcom triggered a rotation, not an AI collapse

Miss this rotation, and the cost is not just pride. It can be months of relative returns. The clean break came when Broadcom, down 14%, dragged the chip complex. That did not end the AI trade. It showed where crowded chip exposure had become vulnerable and pushed capital to look for other ways to own the same AI buildout.

What changed in the tape

The market message is straightforward: the Philadelphia Semiconductor Index has dropped nearly 24% since late June. Over the same stretch, healthcare and financials gained about 9% and 5%, while the equal-weight index rose 3.3% and the cap-weighted index fell 3.5%. That points to broadening leadership rather than a simple market top. In plain terms, money is moving out of the most crowded AI chip names and into broader parts of the market that can still benefit from infrastructure spending.

Why bulls still have the stronger evidence

Bulls still have the cleaner backdrop. J.P. Morgan sees little evidence of a fundamental slowdown and does not expect hyperscalers to scale back AI compute investment into 2027. Bears are focusing on something different: the market is demanding proof that massive AI capex can turn into profits faster than investors have assumed. That keeps the debate on timing and returns, not on whether AI demand exists.

> If hyperscalers actually cut AI compute spending, or if rotation fails and capital surges back into the same crowded leaders, this thesis weakens. For now, the evidence still looks more like reallocation than collapse.

The real trade is a move down the infrastructure stack

Money did not leave AI. It started moving toward the less crowded parts of the buildout.

Demand is still intact

CNBC has reported that AI infrastructure spending is driving record growth for industrial, utility and power equipment companies. That supports a simple point: the broader AI buildout is still funding-heavy. Even within semis, the latest weakness looks more like a repricing of expectations than proof that demand disappeared. J.P. Morgan still sees little evidence of a fundamental slowdown and does not expect hyperscalers to cut AI compute spending into 2027.

Why bottlenecks matter more after a 24% semi pullback

That helps explain the rotation. Semiconductors remain the glory area of the trade, but power, grid equipment, utilities, and related industrial names sit closer to the constraint. Infrastructure cycles often reward the parts of the system that are hardest to scale quickly. That is why capital is drifting toward sectors tied to data-center construction, power delivery, and broader industrial capacity instead of sticking only in the most crowded chip stocks.

Oracle showed the market's new worry: payback, not demand

Oracle did not just fall. It led the latest AI selloff after earnings and spending guidance revived fears about excessive capital expenditures. That is the real fault line now. The debate is less about whether AI spending is happening and more about whether the return on that spending will arrive quickly enough to justify today's valuations.

Weakness in other large tech names added pressure, with Apple losing 6% while Alphabet and Meta also closed lower. But that broad tech weakness supported the move rather than driving it. The core shift is simpler: investors are becoming more selective about where they put incremental AI exposure.

What would confirm the rotation, and what would break it

Confirmation: breadth holds even if the headline tape softens

The cleanest confirmation is a flat or soft headline backdrop while the equal-weight index keeps outperforming as capital moves into under-loved sectors and more cyclical names. That would suggest rotation with staying power, not just panic management.

Invalidation: chip leadership returns while breadth fades

The clean break in the thesis is not another semi pullback. It is chip leadership coming back while the broader market loses breadth. If capital rotates back into crowded AI leaders instead of holding the newer winners, the original market hierarchy is still in charge.

How to think about the trade from here

Stay flexible. Rotation wins when the market keeps questioning how quickly AI spending converts into economic payback, including concerns highlighted around Oracle-led AI selloff and delays in data center construction.

If inference demand, hyperscaler budgets, and ecosystem profitability keep improving faster than the market can rerate the bottleneck names, this could still look like a temporary head fake. If breadth and sector broadening persist, the rotation has more substance.