A piece of computing power news from Meta has pierced the one-sided narrative of AI hardware momentum trading.

BlockBeats news, July 2: Meta’s stock surged in Wednesday’s U.S. trading session, but it unexpectedly put AI hardware trading under a sudden pressure test. The market had originally expected the start of July to be relatively quiet. U.S. stocks performed strongly in Q2, with the S&P 500 just posting one of its best quarters since the pandemic rebound in 2020. However, before the U.S. market opened, news about Meta potentially releasing or selling “excess computing power” suddenly changed the market narrative.

The news is positive for Meta itself. The market interpreted it as the company shifting from continually ramping up capital expenditures to placing more emphasis on financial discipline and free cash flow. Meta’s stock therefore jumped; the article said its single-day gain was about 10%, one of the best single-day performances this year.

But for the AI hardware chain, it’s another story. In the past few months, one of the most crowded trades in the market has been betting that cloud providers will continue expanding computing power, storage, and data center capital expenditures. If Meta starts releasing excess computing power, investors will naturally ask: Is AI computing power demand really as endlessly tight as previously expected? Will cloud providers’ capital expenditures only be revised upward, with no revisions downward?

UBS trader Christina Dwyer said the Meta event pushed the market narrative toward “stronger financial discipline,” while easing outside concerns about capital expenditures continuing to rise. This benefits platform tech companies whose valuations are already close to the lows, but it weakens the “long-term computing power shortage” logic that previously supported neocloud, semiconductors, storage, and AI supply chain stocks.

The market reacted quickly. The BofA Neocloud Basket fell noticeably, while storage stocks and momentum stocks were hit. Stocks that had previously surged, such as SanDisk and Micron, were sold off. Micron in particular is viewed as a key observation target: since April, it has held above its 20-day moving average, and if it breaks below that level, it could open up room for a pullback toward the 50-day moving average, implying roughly 20% potential downside risk.

This adjustment also quickly evolved into a momentum trade unwind. BTIG’s Jonathan Krinsky pointed out that the Bloomberg Mag7 Index, relative to the Philadelphia Semiconductor Index SOX, logged the largest single-day rebound since 2015. In other words, money flowed out of chip, storage, and high-beta AI hardware stocks and back into large-cap platform tech stocks. Goldman Sachs’ high-beta momentum basket fell by about 9% in a single day, while the long-short high-beta momentum combination fell by about 10%, approaching the worst performance since the 2020 vaccine-news shock.

The Meta event reminds investors that cloud providers—the ones actually responsible for capital expenditures—also calculate returns. Once the certainty of upward capital expenditure revisions declines, the segments in the AI hardware chain that have seen the biggest gains and have the most crowded valuations will face pressure first.

However, money has not completely left the AI theme. Software stocks strengthened relative to semiconductors, and Bitcoin also rebounded as money pulled back from AI/storage momentum trades. Another beneficiary is capacitors and other AI bottleneck assets, indicating that the market is still searching for scarce points in computing infrastructure—but no longer indiscriminately chasing storage and chip momentum stocks.

Another risk the market needs to pay attention to right now is that liquidity is currently poor. Goldman Sachs’ trading desk said that in June, top-of-book liquidity for S&P E-mini futures fell 33% month-over-month, while U.S. stock trading volume hit the highest level since 2026. This means the market appears active on the surface in terms of trading activity, but its actual ability to absorb orders is weakening; once large sell orders appear, prices are more prone to sharp volatility.

The real meaning of the Meta event may not be about Meta itself, but that it hit the most sensitive point in the AI rally: whether capital expenditures will continue to grow without limit. AI demand hasn’t disappeared, but the market has started to differentiate between two types of companies—one being platform companies that can steadily monetize compute investments, and the other being hardware and storage suppliers that have already fully priced in expectations of a computing power shortage.