Hook
A broad selloff in semiconductor stocks spread from Asia to the United States on July 28, 2026, as investors weighed intensifying competition from Chinese chipmakers and growing unease over how the largest AI infrastructure projects are being financed. Micron, SanDisk, Western Digital, Intel, Marvell, AMD, Super Micro, and the U.S.-listed shares of SK Hynix all fell sharply, while the Philadelphia Semiconductor Index dropped to a fresh two-month low. The selling followed an overnight cascade in Korea and Japan and reflected a deeper shift in market sentiment that has been building for weeks.
The overnight Asian trigger
South Korea’s Kospi tumbled roughly 11 percent on July 28, triggering its eighth circuit breaker of the year. The index was briefly halted after falling 8 percent, then extended losses to close down 10.8 percent once trading resumed. Index heavyweights Samsung Electronics and SK Hynix ended the day 13 percent and 15 percent lower, respectively. In Japan, the Nikkei 225 fell 4 percent, with memory-chip maker Kioxia, one of the strongest performers in the first half of the year, sinking 18 percent.
The trigger in Asia was the soaring debut of CXMT, a Chinese memory chipmaker, which surged 466 percent on its Shanghai Stock Exchange listing on July 27. The IPO raised $8.6 billion and pushed CXMT’s market value to 3.3 trillion yuan, or about $487.73 billion, nearly half the valuation of U.S. rival Micron. The debut was the largest IPO in Asia in 2026 and signaled that the gap between Chinese and Western memory chipmakers is closing faster than Western investors had assumed.
Credit-default swaps reach record highs
Adding to the pressure, a key barometer of risk in the debt of AI-boom companies has climbed sharply. Credit-default swaps on Oracle, SpaceX, Alphabet, Amazon, Meta, Broadcom, and Nvidia, instruments investors use to bet against corporate debt, hit record highs in recent days, according to LSEG data. The moves mirror a broader selloff in bonds issued by hyperscalers, which are spending hundreds of billions of dollars on data centers and AI models.
“Credit markets don’t deal well with uncertainty, and the sheer unpredictability of the pace and cost of AI financing is triggering a serious crisis of confidence right now,” said John Aylward, chief investment officer at credit manager Sona Asset Management. The pressure has been sharpest at Oracle. The database company’s five-year CDS traded at 215 basis points on Monday, up from 144 basis points at the start of the year, meaning investors must now pay $215,000 a year to insure $10 million of its debt against default.
Oracle’s downgrade and the new math of AI capex
Oracle said last month it would spend $70 billion over the coming year on data-center expansion, a plan that prompted S&P Global Ratings to cut its credit rating to triple-B-minus, just one notch above junk. S&P cited an uncertain path to profitability amid the company’s heavy AI investment. The downgrade crystallized the market’s concern that the AI capex cycle is outrunning the cash flow profile of even the largest hyperscalers.
For chip stocks, the implication is direct. Demand for memory and logic chips has been driven by the same hyperscaler capex that credit markets are now questioning. If investors begin to discount the durability of that spending, the multi-year revenue tailwind that has supported chip stock valuations becomes harder to defend.
The China factor and the new competitive reality
Behind the credit market pressure sits a more fundamental competitive shift. Chinese memory chipmakers are now producing high-bandwidth memory and DRAM at specifications that, until recently, only Samsung, SK Hynix, and Micron could match. The CXMT listing gives Chinese competitors a domestic capital base to expand capacity. The Wall Street Journal reported that Nvidia is in talks to provide around $250 billion toward a data-center project tied to OpenAI, a figure that, while huge, has also raised questions about the assumptions baked into current AI infrastructure spending.
The memory chip market’s exposure to AI risk
Hynix and Micron have built their 2026 capacity plans around the assumption that AI model training and inference workloads would absorb every high-bandwidth memory chip they could produce. The CXMT listing, combined with credit market pressure on hyperscaler capex, has introduced a wider uncertainty band around that assumption. Hynix is the most exposed of the three memory chipmakers, given that its high-bandwidth memory has been the dominant choice for the latest generation of AI accelerators from Nvidia.
The market is now repricing that exposure. Hynix’s U.S.-listed shares fell 6.8 percent on July 28, and the stock remains one of the most volatile in the semiconductor complex. The next round of guidance from the memory chipmakers will reveal whether they are adjusting capacity plans or doubling down. The size of the Hynix capex response over the next two quarters will be the clearest signal of how the industry is reading the new competitive landscape.
What the next earnings cycle will reveal
The next test comes this week. Meta reports earnings on July 29, and Microsoft follows shortly after. Both companies have guided to higher AI capex than the Street had modeled, and both have emphasized that they expect the resulting compute base to support monetizable products. A clear path to revenue from the AI infrastructure so far under construction would go a long way to settling the credit market jitters that have spilled into chip stocks. A weaker signal, particularly on monetization, would extend the repricing that began in Asia on the night of July 27.
Conclusion
The July 28 selloff is not a single-event reaction. It is the convergence of three threads: a record Chinese memory chip IPO, a credit market reckoning with the cost of AI infrastructure, and a sudden repricing of the AI capex cycle. Each thread alone might have been absorbed. The combination has reset the risk premium for the entire semiconductor complex, and the next set of earnings reports, beginning with Meta on July 29, will test whether the selloff is a rotation or the start of a longer correction.