On the morning of August 7, Beijing time, the US stock market experienced a rare "split" market: while all three major indices collectively closed lower, the semiconductor sector stubbornly closed red under the drag of storage chip stocks, with the Dow Jones falling more than 400 points, a drop of 0.85%, the S&P 500 falling 0.18%, and the Nasdaq falling only 0.06%. Compared with the mild pullback of the indices, the divergence at the individual stock level is dramatic—on one side, storage chip giants encountered "death after exposure" selling after delivering bright earnings, while on the other side, ARM surged more than 4%, driving AI computing hardware stocks to fluctuate higher. This multi-battle is writing a new footnote for the pricing logic of US tech stocks in the second half of the year.
Storage Chips "Good News Is Out": The Story Behind Western Digital's 13% Plunge
The most eye-catching in this round of market movements is the collective plunge of storage chip stocks. Western Digital plummeted 13.03% in a single day, SanDisk fell 6.81%, SK Hynix ADR fell nearly 5%, and Micron Technology fell more than 1%. Notably, both storage giants SanDisk and Western Digital delivered strong quarterly results recently, but the bright data not only failed to impress the market but triggered a chain of profit-taking selling.
From an industry fundamental perspective, the storage price increase cycle remains stable. Citigroup's latest research shows that in the second quarter, DRAM average selling price (ASP) increased by about 40% quarter-over-quarter, while NAND ASP increased by as much as 60% quarter-over-quarter, and DRAM manufacturers generally expect supply shortages to intensify in 2027 and continue into 2028 and beyond. Samsung even plans to allocate 60-70% of its capacity to long-term agreement customers to lock in profits from this super cycle. However, precisely because the market has formed a high consensus on storage price increases, and stock prices have fully priced in this cycle in the past few months, the moment when earnings are realized becomes the starting point for funds to cash out.
Goldman Sachs gave a precise diagnosis: the core contradiction facing the storage industry currently is not fundamental deterioration, but market expectations have overly led reality. In other words, it's not that the companies are not good enough, but the fact that they are "good" is no longer fresh. This "good news is out" script is common in cyclical stock movements, reminding investors that when industry consensus becomes too crowded, even the brightest earnings reports can become a catalyst for decline.
Semiconductor Sector Deep V Divergence: ARM's Surge Behind Oversold Recovery Logic
In stark contrast to the decline of storage chip stocks, the Philadelphia Semiconductor Index, which once fell more than 2% at the open, stubbornly turned red, finally closing up 0.33%. Among its components, ARM surged 4.41%, Qualcomm rose 1.82%, AMD rose 1.50%, ASML rose 1.56%, TSMC ADR rose 0.98%, and Broadcom rose 0.55%, with only Intel slightly down 1.24%.
This divergence pattern is the continuation of the recent "oversold rebound" in AI stocks. Data shows that the Philadelphia Semiconductor Index has rebounded about 15% from its low point on July 29, while the Nasdaq 100 and S&P 500 are also approaching record highs in sync. This rebound began with massive short covering: the deepest decliners before late July have seen the largest rebound magnitudes, with the median rebound of the worst-performing fifth of S&P 500 stocks reaching 8%, and lagging chip stocks' gains being about three times that of strong stocks. ARM, AMD, Applied Materials, Marvell, and other chip stocks all recorded double-digit rebounds, but most stocks are still 20% to 40% away from their previous highs.
Market analysts point out that after the violent adjustment at the end of July, the technical picture has completed a阶段性重置, but the sustainability of the rebound still needs observation. On key support levels, if the S&P 500 loses the 50-day moving average near 7490 points, it will increase the risk of a false breakout; the Nasdaq 100 needs to hold the 29,000 to 30,000 point area; and only after the semiconductor index stabilizes above 12,500 points can it hope to challenge higher levels.
AI Application Software Stocks Plunge: Market's Repricing of "High-Valuation Narrative"
Notably, this selling pressure is not limited to storage chips. AI application software stocks also suffered a bloodbath, with Datadog, AppLovin plunging more than 19%, EPAM Systems falling more than 15%, and Axon Enterprise falling more than 14%. Among them, marketing technology platform AppLovin was sold off due to lower-than-expected quarterly revenue, while cloud security company Datadog plummeted due to expected third-quarter revenue growth slowdown.
This logic is similar to storage chips—after two years of狂飙 in AI trading, the market has higher requirements for the realization rhythm of "AI faith." Any signs of growth slowdown can trigger a valuation kill. This also explains the divergent performance within the tech seven giants: Microsoft rose 2.54% to a new high since November last year, Apple and Meta rose slightly, while Google fell nearly 1%, Tesla fell 0.63%, Amazon fell 0.14%, and NVIDIA fell 0.10%. Funds are not leaving tech but are reordering within tech—who can fastest convert AI investment into profits will receive a premium.
Fundamentals Have Not Been Absent: Cloud Capital Expenditure and AMD Earnings Dual Endorsement
Despite short-term volatility, the medium-to-long-term logic supporting the AI computing chain has not动摇. Based on second-quarter data, Citigroup has raised its 2026 and 2027 capital expenditure growth expectations for the US "Big Five" cloud service providers to 90% and 46% year-over-year, respectively. Among them, Alphabet raised its 2026 capital expenditure guidance to $195-205 billion, more than double last year; Amazon also revised its guidance upward from $200 billion to $220 billion. This real money investment will be directly converted into orders for GPUs, server CPUs, storage, optical interconnects, and advanced packaging.
AMD's latest earnings report is also highly persuasive. In Q2 2026, AMD achieved $11.5 billion in revenue, a 50% year-over-year increase and 13% quarter-over-quarter increase, with profitability at a historic high. Its data center business revenue doubled year-over-year to $6.7 billion, accounting for 58% of total revenue. Even though this report led to a post-market plunge of more than 9% because it didn't meet some institutions' "super explosion" expectations, Lisa Su still gave stunning guidance: server CPU revenue in the second half of 2026 will grow more than 80% year-over-year, and the overall data center segment revenue in 2027 could more than double year-over-year. She estimates that the high-performance computing and AI computing market will have a compound annual growth rate of about 40% in the coming years, with the overall market size approaching $2 trillion by 2030.
Institutional Portfolio Rotation: From Chip Design to Semiconductor Equipment
Within sectors, institutional preferences are also quietly changing. Citigroup clearly stated in its report that compared to the entire chip industry, it is more optimistic about semiconductor equipment stocks because "the upward momentum of earnings forecast upgrades driven by capital expenditure revisions is stronger." Data shows that the average upward revision of consensus expected EPS for semiconductor companies in 2026 and 2027 is 9% and 6%, respectively, while for equipment manufacturers it stabilizes at 10% and 11%. Citigroup accordingly recommends buying AMD, Texas Instruments, and Applied Materials, and has placed Applied Materials on its 90-day positive catalyst watch list—the company will report earnings on August 13, and Citigroup expects its earnings guidance to exceed market consensus.
Industry data also confirms the equipment sector's prosperity: KLA, Lam Research, and TE Connectivity have all revised their 2026 global fab equipment spending expectations upward to over $150 billion; TE Connectivity even expects equipment spending to reach over $190 billion in 2027 and could enter the $200-250 billion range from 2027. KLA admits to "unprecedented visibility" for the second half of 2027 and has even begun early discussions about demand beyond 2029; Lam Research describes the current demand environment as an "extraordinary layout" for 2027 growth. As storage chip stock valuations are fully discounted, the equipment sector may be becoming a new direction for funds seeking "expectation differences."
Macro Variables: Geopolitical Risks and Oil Price Disturbances
At the macro level, geopolitical risks remain a variable hanging over the market. On August 6 local time, Salimi, a member of the Iranian Parliament's Presiding Board, stated that Iran's National Security Council is reviewing a draft bill aimed at "ensuring the security and sustainable development of the Strait of Hormuz and the Persian Gulf," which includes provisions banning US, Israeli, and other hostile countries' ships from passing through the Strait of Hormuz. Affected by this, international oil prices continue to strengthen, with New York crude jumping more than 1% to $78.08/barrel, with WTI and Brent both closing up more than 4% the previous day. Recalling early August, positive signals in US-Iran talks once caused oil prices to plummet more than 5% in a single day and boosted US stocks to record highs; now the negotiation outlook has new twists, and energy price volatility will continue to affect global risk appetite.
Meanwhile, US Department of Labor data shows that as of the week ending August 1, initial jobless claims were 199,000, slightly lower than the expected 202,000, with the four-week moving average falling to 199,000, indicating the labor market remains resilient. This data to some extent alleviates market concerns about a US hard landing and provides more flexibility for the Fed's subsequent policy.
Conclusion: AI Narrative Enters "Verification Period", Divergence Will Be the Norm
Overall, the plunge of storage chips and the deep V rebound of the semiconductor sector together point to a core conclusion: the AI investment narrative is transitioning from "faith-driven" to "earnings verification." In this phase, broad market gains will become increasingly difficult, replaced by intense multi-battle between sectors and individual stocks—overpriced narrative stocks may face valuation kills at any time, while segments with solid fundamentals and visible capital expenditure are expected to continue receiving premiums.
For investors, the key now is not to judge whether the AI wave is ending, but to identify where the "expectation difference" lies in the industry chain. The storage price increase logic still holds, but stock prices have already discounted it in advance; the visibility of semiconductor equipment, CPU, and HBM supply chain is actually higher. As Citigroup says, after oversold comes opportunity, but opportunities only belong to those who can remain clear-headed in crowded trades. Against the backdrop of continuous expansion in AI computing demand, every violent fluctuation may be an opportunity to re-examine portfolio structure.
