The AI Trade Split in Two as Meta Rallied and Semis Slid

Bottom line

The market did not have a broad selloff. It had an AI stack rotation.

The headline move on July 1 was not a macro crash. Reuters reported that Meta Platforms was up more than 10% intraday after Bloomberg said it is building a cloud business to sell excess AI capacity, while semiconductors were weaker and the semiconductor index was off 5.9%.

That is a very specific message from the market: monetization is being rewarded, but the parts of the AI trade that depend on ever-higher infrastructure expectations are being repriced more aggressively.

This was a split tape, not a broad panic. The market was rewarding cash-flow optionality and penalizing parts of the stack that already looked fully owned.

What moved

The tape separated the AI beneficiaries from the AI suppliers.

Reuters said the S&P 500 was little changed, the Nasdaq was down, and the Dow was slightly higher. The same report said semiconductors were under pressure while Meta was the standout winner. BlackRock's SOXX fact sheet later showed a 6.48% daily NAV drop on July 1, confirming that the selloff was real inside the chip basket.

So the best read is not that investors suddenly hated AI. They just became much more selective about which part of the AI stack deserved a premium multiple.

Who won and who got repriced
Instrument	Move	Interpretation
Meta Platforms	up >10% intraday	The market liked a new monetization path for excess compute.
Semiconductor index	down 5.9% intraday	Supply-side names were hit by oversupply and expectation reset fears.
SOXX ETF	-6.48% daily NAV	The sector as a whole de-risked, not just one stock.
S&P 500	+0.05%	No broad market panic; the move stayed concentrated.
Nasdaq	-0.22%	Tech weak, but not a full index break.

Why semis sold

The problem is not demand. The problem is what the demand is already priced to be.

If Meta can turn spare compute into a product, the story for the AI stack changes. The market no longer only asks whether chips are needed; it asks whether the buyer of those chips can fully monetize the infrastructure before depreciation, financing, and customer concentration become problems.

That is why semis can sell off even when the AI theme is still intact. The immediate pressure lands on names whose valuation already assumes several more rounds of capex growth and perfect utilization. Once the market starts questioning that, the multiple contracts first and the fundamentals later.

Neoclouds are most exposed because they rely on renting capacity rather than owning a broader monetization engine.
Chip makers still benefit from AI demand, but the market becomes less forgiving when guidance no longer beats an already elevated bar.
Power, cooling, and grid infrastructure remain important because density and utilization still matter even in a slower tape.
Software can get a relative lift when the market shifts from capex excitement to monetization discipline.

Long-term read

The AI trade is not ending. It is splitting into monetizers and suppliers.

That split matters for the next leg of returns. Companies that can convert infrastructure into recurring revenue should keep earning a premium. Companies that depend on uninterrupted capex growth will need cleaner proof that the capital cycle still has room to run.

For the supply chain, the takeaway is still bullish on the physical buildout, but more selective on the equities. Chips, networking, cooling, and data-center gear remain central. The difference is that the market now cares more about utilization and customer quality than about raw spend alone.

The rotation was concentrated, not random

Bar lengths show the absolute size of the daily or intraday move. Notes carry the direction so the chart can show how sharply the AI trade split apart.

단위: Percent move