Kospi Rises for Second Day, but Retail Investors Net Buy $151 Million of Inverse ETFs

South Korea’s Kospi rose for a second straight day, but retail investors were heavily buying inverse exchange-traded funds that profit when the benchmark index falls.

According to data from the Korea Exchange and Koscom CHECK, from September 7 through 10:55 a.m. on September 8, the top net buy among individual investors on the benchmark KOSPI market was KODEX Inverse, with net purchases of 107.4 billion won ($79.3 million). KODEX 200 Futures Inverse 2X, which tracks twice the inverse of the index’s decline, ranked second after retail investors bought a net 97.2 billion won ($71.8 million) of the fund.

The Kospi has been rising for two straight sessions as money flowed into semiconductor shares. Even so, individual investors appear to be putting more weight on the possibility of a short-term correction. Net purchases of the two inverse ETFs alone exceeded 200 billion won ($147.7 million).

Brokerages, by contrast, say chip stocks could have further upside as artificial intelligence investment expands.

In a report on September 8, KB Securities Head of Research Kim Dong-won wrote that the AI technology rivalry between the U.S. and China would drive a new expansion in memory-chip demand. From next year, demand for HBM4, the sixth generation of high-bandwidth memory, along with server DDR5 and enterprise solid-state drives, is set to rise rapidly, making Samsung Electronics and SK Hynix the main beneficiaries.

He also projected that launches of new AI models in both the U.S. and China, combined with competition in data-center investment, would lift memory demand at the same time.

"Chinese AI accelerators are less efficient than U.S. graphics processing units, so handling the same AI computing workload requires more accelerators and more memory," Kim wrote. "U.S. AI models are raising the upper bound of performance and expanding the ceiling for demand, while China’s efficiency-focused AI models are sharply lowering token costs and are expected to speed the mass adoption of AI." He named Samsung Electronics and SK Hynix as his top picks.