Chip Stocks Bleed as SK Hynix Rattles AI Spending Thesis
SK Hynix's disappointing results triggered a sector-wide selloff: Micron lost 9.9%, NVDA fell 3.5%, AMD dropped 5.5% in a single brutal Wednesday session.
Key Points
- SK Hynix's disappointing Q2 results triggered a single-session wipeout: Micron lost 9.9%, Applied Materials fell 8.4%, AMD shed 5.5%, and Nvidia dropped 3.5% on Wednesday.
- The market's read-through is that AI-driven memory demand may be peaking or decelerating at the margin, challenging the bull case that has driven chip stocks to elevated valuations throughout 2025 and 2026.
- Traders should watch Nvidia's and Micron's reaction to today's MSFT print and tonight's Amazon AWS number — a strong AWS result could stabilize the sector, but another guidance miss from a major hyperscaler would accelerate the rotation out of semis.
Micron Technology lost 9.9% in a single session on Wednesday, July 29 — the largest one-day drop in the stock in months — after South Korea's SK Hynix delivered quarterly results that rattled the core assumption underpinning the entire semiconductor bull market: that AI-driven demand for advanced memory and compute infrastructure is insatiable. It is not. Or at least, the pace of that demand is more variable than the multiples assigned to chip stocks had priced in, and Wednesday's broad selloff was the market adjusting the odds in real time.
What SK Hynix Actually Said
SK Hynix's results pointed to softness at the margin in high-bandwidth memory demand, the premium memory product that has been the single biggest beneficiary of the AI infrastructure buildout. HBM is the memory architecture used in Nvidia's H100 and successor GPU chips, and SK Hynix has been one of the primary suppliers — meaning any signal of demand deceleration from that company carries direct implications for the entire AI hardware supply chain. Samsung's stock fell 13.4% in Seoul, SK Hynix itself dropped more than 14.7%, and the contagion spread immediately to U.S.-listed names before the New York open.
The mechanism is straightforward. If hyperscalers — the Microsofts, Amazons, Googles, and Metas of the world — are pulling back on or simply not accelerating their orders for the most advanced memory, it suggests either that their data center buildouts are temporarily sated, that their AI models are becoming more memory-efficient, or that capital allocation priorities are shifting. None of those scenarios are bullish for memory chip pricing, and Micron, as the primary U.S. competitor to SK Hynix and Samsung in the HBM market, absorbed the most direct damage at 9.9%. Applied Materials, which supplies the fabrication equipment used to manufacture these chips, fell 8.4% — a secondary but equally telling data point, since equipment orders are a leading indicator of future fab capacity investment.
The Broader Damage and the Meta Connection
Wednesday's selloff was not contained to semiconductors. The S&P 500 fell 1.5%, the Nasdaq 100 dropped 2.1%, and the Dow Jones Industrial Average shed 1,153 points in its worst single-day loss in several months. The chip sector was the epicenter, but the damage radiated outward. Softbank declined 4.43% in Tokyo — notable because SoftBank has made aggressive bets on AI infrastructure through its Vision Fund and its planned Arm-related investments. Advantest, which makes semiconductor testing equipment, fell more than 10% in Tokyo. The geographic breadth of the selloff, spanning Seoul, Tokyo, and New York simultaneously, signals that this was not a one-company earnings miss — it was a reassessment of a global investment thesis.
Meta Platforms compounded the negative sentiment after the close on Wednesday, falling more than 6% in after-hours trading after its Q3 revenue forecast missed Wall Street estimates. Meta's situation is distinct from the memory chip story, but the market is reading both through the same lens: AI capital expenditures are enormous — Meta guided to $125B to $145B in 2026 capex — and the revenue justification for that spending is not arriving on the timeline the market had assumed. CNBC's live coverage Wednesday captured the dual pressure on markets: the Fed's three-dissent vote to hold rates, combined with the AI monetization anxiety, created a risk-off environment with few places to hide inside the Nasdaq 100.
The Federal Reserve backdrop adds a second layer of pressure specifically on high-multiple technology stocks. With three FOMC members voting for a rate hike at the most recent meeting — the most hawkish dissent in years — the probability distribution around future rate cuts has shifted. The fed funds effective rate sits at 3.63% as of July 28, and SOFR is at 3.65%, but the 10-year Treasury yield at 4.61% reflects a term premium that is reasserting itself. High-multiple growth stocks, and semiconductors in particular, are acutely sensitive to discount rate movements because so much of their perceived value is embedded in earnings several years out. A world where the Fed is more hawkish than expected is mechanically a world where those distant earnings are worth less today.
Where the Sector Goes From Here
The critical variable for semiconductor stocks over the next 48 hours is the Amazon Web Services revenue number, due after today's close. AWS is one of the largest buyers of Nvidia chips and advanced memory in the world, and its Q2 results will either validate or undercut the SK Hynix demand-softening signal. Analyst estimates tracked by Yahoo Finance show the Street had been raising AWS growth forecasts heading into this print, with consensus expecting strong AI-driven cloud demand. If Amazon delivers an AWS acceleration and provides bullish forward guidance, the SK Hynix read-through gets significantly complicated — it may be a company-specific or Korea-specific inventory issue rather than a systemic demand signal.
Microsoft's blowout print — EPS of $4.74 versus $4.24 expected, with Azure cloud revenue up 27% — argues that the demand for AI compute at the hyperscaler level remains robust. The tension for semiconductor traders is that MSFT's strength is in software, services, and cloud delivery, not in raw chip procurement volume. A strong Azure print does not necessarily mean Microsoft is ordering more HBM memory next quarter — it means the workloads running on the infrastructure already in place are generating strong revenue. Those are related but distinct signals. Nvidia fell 3.5% Wednesday despite no company-specific negative news, which tells you the market is applying a blanket discount to AI hardware exposure until the demand picture clarifies.
For traders with active positions in NVDA, MU, AMD, or AMAT, the key level to watch in Nvidia is the $115–$118 support zone that has acted as a technical floor during prior pullbacks. A close below $115 on heavy volume would suggest the selloff has further to run and that institutional holders are reducing exposure rather than buying the dip. Micron's next major fundamental catalyst beyond today's tape is its own earnings report, currently expected in late September — meaning MU holders have roughly two months of headline risk from competitor reports and macro data before they get their own management commentary. The Fed's next scheduled meeting and any intervening inflation data between now and then will determine whether the rate-hike dissent from Wednesday's FOMC meeting hardens into a genuine policy pivot that resets the entire valuation framework for the sector.
The Weekly Investor
Daily market analysis for active traders. Free.
