Volatility in SK Hynix shares has surged this year as it has become a target of global investors betting on a sustained boost to profits from a shortage of high-bandwidth memory chips used in AI data centres, with many investors using leveraged exchange-traded funds that have amplified returns and losses.
In Hong Kong, a single-stock ETF tracking SK Hynix 7709.HK offered by fund manager CSOP, which uses leverage to target twice the daily returns of its shares, lost more than a third of its value on Monday, its biggest one-day decline since listing in October.
After the rout in the Seoul market on Monday, SK Hynix's U.S. ADRs, which represent one-tenth of a share and were last at $154.70 on Monday, were trading at a premium of about 25.6% to the South Korean share price.
"It's typical for ADRs to trade at a premium because they give U.S. investors direct access to the stock for the first time," said Nic Puckrin, cross-asset analyst and founder of Coin Bureau.
"Though some investors have been taking advantage of the arbitrage opportunity, these trades tend to get crowded and the price eventually evens out, so the premium likely won't hold forever."
Ryu Young-ho, a senior analyst at NH Investment & Securities, said investors were profit-taking after the conclusion of the U.S. listing, while sentiment also suffered from caution with regard to SK Hynix's second-quarter earnings.
He said investors had expected shipments of SK Hynix's HBM4 chips to increase from the second quarter, but that the increase does not appear to have materialised at scale.
Ryu also said investors had moderated earnings expectations because SK Hynix, with its greater exposure to the HBM market than crosstown rival Samsung, was set to benefit less from a recent rise in prices for conventional DRAM chips.
SK Hynix led the market for high-bandwidth memory chips with a 58% revenue share in the first quarter, whereas Samsung and U.S. competitor Micron Technology MU.O each held 21%, Counterpoint Research data showed.
HBM chips are primarily used in artificial intelligence systems for customers such as Nvidia NVDA.O and Alphabet's GOOGL.O Google.