SEOUL, July 28 - South Korea's semiconductor sector experienced a sharp selloff on Tuesday, with major chipmakers seeing double-digit percentage losses as investor appetite for AI-linked names retreated.
Shares of Samsung Electronics slid up to 9.5% intraday, while SK Hynix dropped as much as 10.9%, reflecting heavy selling across the space. U.S.-listed SK Hynix had already declined overnight, closing at $143.02, a level below its initial public offering price of $149. The broader market also suffered: the benchmark KOSPI index was down about 7.3% as of 0032 GMT.
Market participants pointed to a cluster of developments that together undermined confidence in the durability of the AI-driven rally in chip stocks. SK Hynix, which supplies high-bandwidth memory (HBM) crucial to AI workloads and a key supplier to Nvidia, has been among the largest beneficiaries of the surge in AI spending. That exposure left its shares particularly vulnerable to any shift in sentiment about future AI demand.
One catalyst for the renewed caution was a report saying Nvidia could extend a roughly $250 billion financial backstop to a data-centre project tied to OpenAI. The suggestion that Nvidia might be providing sizable financing for a customer-related infrastructure project knocked almost 5% off Nvidia's share price, as investors questioned the implications of the company financing its customer base.
Concerns were compounded by reports of technological progress in China. Accounts that China has made strides toward producing homegrown deep ultraviolet (DUV) lithography tools have unsettled investors, since such a development would represent a step toward reducing reliance on Western semiconductor equipment despite U.S. restrictions on Chinese access to imported chip technology.
At the same time, market dynamics in memory chips are evolving. The strong market debut of Chinese memory-chip maker CXMT heightened fears of intensifying competition in the global memory industry. Reports that Apple had lobbied the U.S. administration about permitting the use of Chinese-made chips in some products were also cited as an additional disquieting factor for investors already focused on China’s growing capabilities.
Another element cited by analysts was the rising popularity of lower-cost Chinese open-source AI models, including Kimi K3. The growing adoption of these models raised questions about whether future AI workloads will demand the same intensity of compute and high-bandwidth memory as previously anticipated, potentially reducing demand for advanced AI chips and HBM.
Taken together, these developments prompted broad-based selling in semiconductor names linked to AI hardware and memory production. Observers noted the moves reflected an aggregation of financing concerns for AI infrastructure, technological and supply-chain shifts in China, and heightened competition within the memory sector.
Market context: The episode underscores how intertwined investor sentiment around AI spending, corporate financing arrangements, and geopolitical-technology competition can materially affect semiconductor valuations in a short period.