Memory

South Korea memory stocks wobble on AI demand doubts

Late‑July selloff shows investor nerves over AI capex, prices and China competition

Late‑July selloff shows investor nerves over AI capex, prices and China competition

Shares of South Korea’s two memory giants slid sharply in late‑July trading as investors questioned whether AI‑driven demand will sustain current pricing and capital‑spending plans, knocking the KOSPI into one of its worst sessions of the year.

The move capped a month of volatility that began with U.S. semiconductor declines in early July and accelerated as traders reassessed how fast cloud and AI customers will keep expanding data‑center capacity.

The stock pain arrived despite blowout second‑quarter results: Samsung reported a record operating profit for April–June, and SK hynix posted historic quarterly earnings as memory prices and demand surged.

Investors said they were pivoting from short‑term earnings to the sustainability of the AI investment cycle, triggering profit‑taking and a broader rotation out of highfliers. Market commentators pointed to over‑leverage in the trade and a recalibration of AI capex forecasts.

Capital spending plans are a big part of the debate. SK hynix has raised capex guidance as it rushes to expand HBM and DRAM capacity, while Samsung and SK have outlined multiyear investments to scale advanced fabs and packaging. Those commitments help explain why investors worry about spending trajectories as prices evolve.

Geopolitics and competition have intensified the mood. Chinese developments — including a frenzied IPO for a domestic memory maker that underlined Beijing’s push to build local supply chains — added to fears that competition could erode margins and pricing power.

The market mechanics magnified the moves. A heavy concentration of large memory names in the South Korean index, growth‑style positioning and popular single‑stock ETFs helped turn a sector pullback into a marketwide rout that prompted regulators to weigh measures to limit retail leverage.

At the product level, demand for high‑bandwidth memory used in AI accelerators has been a key earnings driver, but analysts warn that the timing and persistence of HBM and server DRAM purchases depend on cloud customers’ capex cycles and chip pricing trends. That makes revenue forecasts unusually sensitive.

Analysts described the late‑July selloff as a mix of profit‑taking and renewed skepticism about the durability of the AI boom. Some called the reaction overdone, while others said it reflected genuine uncertainty about how quickly hyperscalers will keep expanding capacity.

The near‑term watchlist is straightforward: follow capex statements from major cloud and AI players, upcoming earnings from chipmakers, and any signals on memory pricing from contract negotiations. Those datapoints will reset investor views on whether current margins are transitory or sustainable.

For investors and policymakers the episode is a reminder of vulnerability: a small shift in AI capex forecasts or a geopolitical surprise can quickly translate into large market moves, especially when a handful of firms dominate an index. Regulators in Seoul have already signaled they are monitoring leverage and volatility.

Fundamentals remain strong — demand for AI‑optimized memory has lifted revenues and margins — but stock prices now more tightly reflect confidence in multi‑year demand growth and protection against rising competition. That leaves memory makers and their investors exposed to rapid swings in sentiment.