U.S. stock market's three major indices collectively closed lower. Why did the storage sector plunge? AI computing power shifts from scarcity expectation to reversal.

On July 2, Beijing time, the three major U.S. stock indexes all fell on the day. At the close, the Dow Jones Industrial Average fell 13.96 points, or 0.03%, to 52,305.24; the Nasdaq Composite Index fell 173.69 points, or 0.66%, to 26,040.03; and the S&P 500 Index fell 16.13 points, or 0.22%, to 7,483.23.

From an index-level perspective, the declines were not severe. However, the divergence within sectors was extremely stark. Big tech stocks were mixed: Meta rose more than 8%, Microsoft rose more than 3%, and Google, Apple, Tesla, and Amazon all rose more than 1%; while Nvidia fell more than 1% and SpaceX fell more than 7%.

The true “eye of the storm” was in the semiconductor and storage sector. The Philadelphia Semiconductor Index plunged more than 6%, and the storage sector index was down as much as about 9% at one point during the session. At the individual stock level, Corning fell more than 13%, Micron Technology and SanDisk fell more than 10%, Intel fell more than 9%, ASML fell more than 7%, AMD and TSMC fell nearly 7%, Western Digital fell more than 6%, and Supermicro and Seagate Technology fell more than 5%.

One noteworthy detail is that while Meta surged more than 8%, the storage and semiconductor sectors collectively collapsed. This “icy and fiery at once” pattern itself points to a core narrative— the internal logic of the AI industry chain is undergoing structural restructuring.

Meta’s Cross-Over Cloud Business: How One Spark Can Ignite the Entire Storage Sector

The market generally attributes the sharp plunge in the storage sector to a piece of news from social media giant Meta. Reports say Meta is formulating plans to launch a cloud infrastructure business, selling AI computing power and model access permissions to external customers. Meta CEO Mark Zuckerberg revealed at the annual shareholders’ meeting that entering the cloud computing market is “absolutely within consideration,” with external companies asking almost every week whether they can purchase Meta’s computing power at a premium.

The reason this news triggered an intense reaction in the storage sector is not a straightforward bearish signal for storage, but rather an expectation reversal.

Over the past two years, the market’s investment logic for AI hardware has been built on the assumption that “computing power will always be scarce.” The persistently high capital expenditures of technology giants (Meta, Microsoft, Google, Amazon, etc.) have been viewed as a rigid demand guarantee for upstream hardware such as AI chips, storage chips, and optical communication modules. Storage chips, as core components of AI servers—HBM (high-bandwidth memory), DRAM, and NAND Flash—directly benefited from this capital expenditure cycle.

However, Meta’s news about selling “excess AI computing power” to the outside world breaks this narrative. The implicit inference is that if Meta’s computing power is already so abundant that it needs to be leased out externally, then the industry’s computing power supply may be shifting from “undersupply” to “oversupply.” Analysts note that the core logic shift triggered by this news is that market expectations have moved from “buying at any cost” to “oversupply and optimization of existing inventory.”

In other words, the market interpreted Meta’s move as a signal that AI infrastructure investment is “shifting from offense to defense.” Since existing computing power is already in surplus and needs to be leased out, the growth rate of technology giants’ capital expenditures may slow down, or even reach an inflection point. For the storage chip industry, which is highly dependent on AI investment to drive demand, this means a sudden downward revision of demand expectations.

This logic propagated at an extremely fast pace. After the overnight plunge in the U.S. storage sector, panic sentiment spread quickly across borders. On July 2, when South Korean stocks opened, both Samsung Electronics and SK Hynix shares fell by more than 5%. The South Korea KOSPI index plunged by more than 6%, and KOSPI 200 futures dropped by 5%, directly triggering a circuit breaker mechanism. A blanket sell-off sentiment formed across the global storage industry chain.

Goldman Sachs’ “Three Warnings”: Structural Risks Had Been Lurking for a Long Time

Meta’s news was the direct spark, but the storage sector’s plunge was by no means without warning. Days earlier, Goldman Sachs trader Ippei Yamaura had released a report that clearly warned of three major downside risks facing the storage chip sector.

First, momentum in HBM prices is slowing down. HBM is the core associated storage for AI server GPUs. In the past two years, its price surged due to persistent tight supply. But as major manufacturers continue to expand capacity, Goldman Sachs expects that in fiscal years 2027 to 2028, HBM supply will increase significantly, and the high gross-margin advantage brought by prior tight supply may gradually shrink, leaving price support under challenge.

Second, the market competition landscape is worsening. Chinese manufacturers represented by Changxin Memory Technology are accelerating their entry into the DRAM market, intensifying a competitive landscape that was originally dominated by Samsung, SK Hynix, and Micron. The addition of new entrants implies rising price pressure and a redistribution of market share.

Third, AI server investment is slowing down overall. This is a risk point directly related to Meta’s news. Goldman Sachs has explicitly listed “an overall sudden slowdown in AI server investment” as one of the core downside risks for the storage sector. If technology giants cut capital expenditures because of excess computing power, the demand foundation for storage chips will be hit directly.

Goldman Sachs’ warnings were validated by the market a few days later—Meta’s news exactly triggered the market’s concentrated repricing of these three risks. The storage sector’s year-to-date gains were extremely high, and trading crowding was at historic levels. Any marginal negative signal could spark intense profit-taking. Some analysts point out that the core of this overseas sell-off may stem from loosening trading crowding in the AI hardware track, with the market worrying that prior AI hype became overly overheated, leading funds to collectively start de-risking and reducing positions; AI beneficiaries became the primary targets for selling.

The “Triple Pressure” from Macroeconomic Data: Economic Slowdown and Policy Uncertainty Combine

Besides the industry-level logic reversal, the macro environment also provided a backdrop for this drop.

On economic data, the U.S. June ADP National Employment Report released on July 2 Beijing time showed that private-sector employment added 98,000 jobs, below economists’ consensus forecast of 118,000 and also below May’s 122,000. In the same period, the June ISM Manufacturing PMI came in at 53.3, below the market expectation of 54.0 and the prior value of 54.0 in May. Both figures missed expectations, intensifying market attention on a potential U.S. economic slowdown.

On monetary policy, the same day Federal Reserve Chair Kevin Warsh spoke at the European Central Bank’s annual central banking forum in Sintra, Portugal, stating that the Federal Reserve does not provide forward-looking guidance and will make policy decisions based on the latest economic data. Warsh also said that in recent weeks, both inflation expectations and the risks of inflation rising have declined. According to the CME FedWatch Tool, traders’ probability forecast for a September rate hike fell from 80% on Tuesday to 65%.

Although Warsh’s remarks were relatively dovish, the phrase “does not provide forward guidance” in itself increased market uncertainty. Against the backdrop of weaker employment data and slower manufacturing expansion, concerns about economic growth momentum are rising. For technology and semiconductor sectors with high valuations, any macro-level change could be amplified.

On the U.S. dollar index, the index measuring the dollar against six major currencies rose 0.2% on the day, closing at 101.39. The strengthening dollar has a certain dampening effect on commodities priced in dollars and on the global technology industry chain.

The combined effect of three factors—industry logic reversal, the release of structural risks, and weaker macro data—together forms the complete logical framework behind the storage sector’s July 2 crash.

Conclusion

The sharp decline in the U.S. storage sector on July 2 was the result of multiple factors converging.

The direct fuse was news about Meta’s cross-border cloud business— it shattered the market’s belief that “computing power will always be scarce,” triggering an expectation reversal across the AI hardware industry chain. The deeper structural risks came from three downside factors previously flagged by Goldman Sachs—slowing HBM price momentum, worsening competitive landscape, and slowing AI investment—risks that were concentratedly repriced in the wake of the news. At the macro level, both ADP employment and ISM Manufacturing PMI came in below expectations, and combined with uncertainty around Federal Reserve policy, they provided a macro backdrop for a pullback in high-valuation sectors.

Have the long-term supply-and-demand fundamentals of storage chips truly undergone fundamental change? The answer may be no. Micron CEO had previously said that storage is the underlying bottleneck that AI has been underestimating, and structural constraints on the supply side mean that storage shortages will remain a long-term condition, with tight supply at least continuing beyond 2026. But from a short-term trading perspective, when a sector’s gains within the year have been too high and trading crowding is at extreme levels, any marginal negative signal may trigger a severe price correction.

For investors, understanding the logical chain behind this sell-off is more valuable than trying to predict short-term ups and downs. It reveals a fact that is unfolding: the AI industry chain is moving from “indiscriminate benefit” to “structural differentiation.” In this process, identifying the true value creators versus those driven purely by sentiment will be key to distinguishing winners from losers.

FAQ

Q1: How much did the three major U.S. stock indexes fall on July 2?

The Dow Jones Industrial Average fell 0.03% to 52,305.24 points; the Nasdaq Composite Index fell 0.66% to 26,040.03 points; and the S&P 500 Index fell 0.22% to 7,483.23 points.

Q2: What is the direct reason for the sharp drop in the storage sector?

The direct fuse was news that Meta plans to launch a cloud infrastructure business and sell AI computing power to external customers. The market interpreted this as a sign that AI capital expenditures by tech giants may have peaked, sparking concerns about the outlook for demand for storage chips.

Q3: How much did the Philadelphia Semiconductor Index drop on that day?

The Philadelphia Semiconductor Index fell by more than 6%. Some reports show the specific decline was 6.27%.

Q4: Which storage-related stocks saw the largest declines?

Corning fell by more than 13%, Micron Technology and SanDisk fell by more than 10%, Intel fell by more than 9%, ASML fell by more than 7%, AMD and TSMC fell by nearly 7%, Western Digital fell by more than 6%, and Supermicro and Seagate Technology fell by more than 5%.

Q5: What warning did Goldman Sachs previously issue about the storage chip sector?

Goldman Sachs summarized the risks to the storage chip sector in three points: slowing momentum in HBM prices, intensified market competition brought by Chinese manufacturers such as Changxin Memory Technology, and an overall slowdown in AI server investment.