AI Memory & Chip Stocks Sell Off Across the Board as South Korea's Two Storage Giants Plunge Over 12% in a Single Day
Miles Bennett
Samsung Electronics and SK Hynix each fell roughly 12% on Tuesday, dragging the Kospi 10% off its all-time high and triggering a circuit breaker; the sell-off spread quickly to U.S. chip and cloud names, pressure-testing the AI compute growth narrative against stretched valuations, rising rates, and crowded positioning.
What lit the fuse?
SK Hynix led the decline after Korean outlet Chosun Biz reported the company is slowing its AI memory chip (HBM) capacity expansion, pivoting toward cheaper commodity DRAM.
This means → the market's worst-case signal arrived: the single most critical supplier in the AI chip stack is easing off the gas, implying demand may not be as strong as priced in.
SK Hynix is Nvidia's primary supplier of high-bandwidth memory (HBM — ultra-fast memory purpose-built for AI training). Its strategic shift directly undermines the consensus that AI compute demand only goes up.
Who fell and by how much?
Korea: Samsung and SK Hynix each dropped roughly 12%; the Kospi triggered a circuit breaker intraday.
U.S. memory and equipment: Micron (MU) fell about 8%, Sandisk 9%; Lam Research, Applied Materials, KLA, and ASML each dropped roughly 7%; TSMC fell about 5%.
U.S. chip designers: Nvidia slid about 3%, AMD and Qualcomm each about 6%, Broadcom nearly 4%, Marvell and GlobalFoundries each close to 8%.
In plain terms = from memory makers to equipment suppliers to chip designers, the entire AI hardware supply chain sold off 5%–12%, with no link spared.
Why did the cloud giants get dragged in?
Alphabet fell roughly 2%, Amazon and Meta each about 1%, Oracle nearly 4%; Microsoft bucked the trend with a slight ~1% gain.
Bloomberg anchor Lisa Abramowicz noted: investors worry that tech firms' ability to charge for frontier-model access has hit a ceiling, squeezing potential margins.
This means → hyperscaler stock prices — Google, Amazon, Microsoft and peers that buy AI chips at scale — are now tightly coupled to AI inference pricing. When the chip end wobbles, the downstream feels it immediately.
How did the macro picture pile on?
New Fed Chair Kevin Warsh dropped forward guidance and changed the inflation reference metric upon taking office; the market promptly repriced the 2026 rate path, pushing Treasury yields higher.
Fundstrat's Tom Lee said: "Yields are becoming a headwind" — the market is digesting expectations for more hikes this year.
In plain terms = chip stocks were already expensive; rising rates make the "discount cost" of high-valuation equities steeper — squeezed from both ends, capital exits fastest.
Just how stretched are valuations?
Per Barron's, the PHLX Semiconductor Index has surged 92% since the start of Q3; meanwhile Microsoft and Meta have each fallen roughly 30% over the past year, and Oracle is down nearly 50%.
This reflects a leadership rotation from consumer-facing tech toward AI digital infrastructure — Marvell was just added to the S&P 500 this week, replacing Pool Corp and Campbell Soup.
Carson Group portfolio management director Blake Anderson confirmed: market leadership is "increasingly tied to the digital infrastructure powering AI, not the consumer-facing names that led before."
What to watch next?
UBS European equity strategist Gerry Fowler warned: investors are growing aware of crowded-trade risk, questioning "how much upside is left relative to the risk."
The Nasdaq faces roughly 700 points (~2.3%) of downside pressure on Tuesday; the S&P 500 may fall to its lowest since June 10.
This means → with Fed policy uncertainty, fragile U.S.–Iran diplomacy, and a cloudy economic outlook stacking up, whether this AI-chip valuation reset stabilizes before quarter-end is the key test of market conviction.
Content is for reference only, not financial advice.