Price-to-earnings ratios (PER) for Samsung Electronics and SK hynix have fallen to historical lows, but an analysis said that should not be taken as a simple undervaluation signal. While earnings are rising quickly on the back of the artificial intelligence (AI) semiconductor boom, large-scale capital expenditures (CAPEX) needed to sustain AI investment and the burden of future supply increases are acting as discounts to corporate value.

A visitor examines the unveiled SK hynix HBM4 in person at SEDEX 2025 at COEX in Gangnam-gu, Seoul, in October last year. /Courtesy of News1

Hwang San-hae, an LS Securities analyst, said in a report on the 9th that the memory sector has entered a "value trap" zone. PER alone makes it look cheap, but the market has already priced in future supply expansion and slowing profitability, making it hard to justify aggressive overweighting based solely on valuation.

Hwang said, "Recently, Samsung Electronics and SK hynix have seen share price declines alongside upward earnings revisions, putting 12-month forward PER at 4.8 times and 5.3 times, respectively, near historical lows," but added, "Although valuation merit stands out at these levels, considering the unique valuation discount structure of AI cycle leaders and valuation errors that occur during periods of rapid earnings re-rating, it falls short as a rationale for further overweighting."

Hwang viewed the structure of the AI industry itself as creating valuation discount factors. To sustain the AI ecosystem, massive investment must continue even as technology prices are gradually lowered.

Hwang said, "Corporations positioned as AI cycle leaders resist bubble theories and maintain cycle durability," adding, "But in that process, they face the tasks of continuous CAPEX expenditure and lower technology unit prices."

He added, "Because cutting investment leads directly to falling out of the cycle, expanding CAPEX, even at the cost of bleeding, is closer to an obligation than a choice," and said, "Corporations benefiting from AI are facing the cost burden of sustaining the AI cycle, which is translating into valuation-level discounts."

The memory sector is no exception, the analysis said. If the current supply shortage continues, capacity additions will be inevitable, which could raise long-term concerns about margin erosion.

Hwang said, "For memory as well, before the current bottleneck slows the entire cycle, supply expansion (a return to cyclical) is unavoidable," adding, "The recent surge in the memory CAPEX growth rate reflects this."

The market also worries that if profits at AI intermediate-goods corporations become excessively large, it could actually weaken the sustainability of AI investment.

Hwang said, "Operating profit at Samsung Electronics and SK hynix has surged to 57% of hyperscalers' CAPEX," diagnosing that "if profits at AI intermediate-goods corporations expand excessively, pressure intensifies on hyperscalers' return on investment (ROI) from AI spending, undermining the justification for additional investment."

He also noted that rising memory prices cannot be seen as an unconditional positive. AI devices require more memory, but as memory bottlenecks persist, the price burden on devices grows, which can slow the spread of AI.

Hwang explained, "Another risk created by memory bottlenecks is constraints on device upgrades," adding, "While the memory threshold needed for practical AI use continues to rise, memory bottlenecks are putting pressure on the device market."

He added, "Low PER during phases when earnings forecasts surge at a time of structural growth or at the peak of a cycle may not be an undervaluation signal," and said, "The cases of Nvidia, Amazon, Alphabet, and Meta attest to this, and we judge that the memory sector is currently in the same value trap zone."

※ This article has been translated by AI. Share your feedback here.