Don't bottom-fish Samsung Electronics, SK hynix just yet, analysts warn

Shares of Samsung Electronics and SK hynix have fallen sharply in recent months, pushing their valuations to historically low levels — but analysts say that alone is not enough reason to aggressively add to positions.

Forward earnings estimates are being revised up rapidly amid a semiconductor supercycle, meaning the currently low price-to-earnings ratios may not simply signal undervaluation.

In a recent report, LS Securities said the 12-month forward PER of Samsung Electronics and SK hynix had fallen to 4.8 times and 5.3 times, respectively, entering historically low territory. The simultaneous decline in share prices and upward revision of earnings forecasts has compressed valuation metrics quickly.

However, LS Securities cautioned against buying on the numbers alone.

"The 12-month forward PER of Samsung Electronics and SK hynix has reached historical lows of 4.8 times and 5.3 times, respectively, as share price declines and earnings upgrades have occurred together," said Hwang San-hae, a researcher at LS Securities. "While the valuation appeal is notable, it falls short as a basis for adding further exposure when you factor in the discount structure typical of AI-cycle leaders and the valuation distortions that arise during periods of rapid earnings re-rating."

The report also flagged that the AI industry's structure carries burdens that differ from past semiconductor cycles. Big tech companies driving AI investment — including Alphabet and Meta — face a dual challenge: sustaining massive capital expenditure to avoid falling behind competitors while also keeping their service prices competitive.

That surge in investment is lifting memory chip demand, but it is also intensifying pressure to expand supply. Hwang said memory chipmakers will have little choice but to increase production capacity to ease bottlenecks, which could ultimately push the industry back toward a structure sensitive to investment and supply variables.

If the profits of AI intermediate suppliers grow too large, it could also undermine the sustainability of investment across the broader AI ecosystem.

"Mounting pressure on hyperscalers' return on investment from AI spending could erode the justification for further capital deployment," Hwang said. "The combined operating profit of Samsung Electronics and SK hynix has surged to 57 percent of hyperscaler capital expenditure — a level that exceeds even the Nvidia bottleneck of the past."

A memory supply shortage could also slow the pace of AI adoption, the report noted. While the memory specifications required to run AI services smoothly continue to rise, persistent bottlenecks could delay the replacement cycle for devices such as smartphones and PCs.

"The memory threshold needed for practical AI use is consistently moving higher," Hwang said. "But memory bottlenecks are squeezing the device market at the same time, and that is a factor fundamentally delaying the spread of AI."

He added that the current moment is "a period in which the market is verifying the sustainability of this earnings level, and pricing in supply expansion and future margin erosion in advance," drawing parallels with Alphabet and Meta between 2020 and 2022, Amazon between 2017 and 2018, and Nvidia between 2023 and 2024. "The memory sector is now in the same valuation trap," he said.

Analysts broadly agree that investors should look beyond the low PER alone and monitor whether the AI investment cycle will hold, how big tech capital expenditure trends, and how quickly memory supply expands. Earnings remain solid, but the market has already begun to price in the possibility of rising supply and slowing profitability ahead — making a simple undervaluation argument an insufficient basis for investment, they said.