Foreign Holdings Hit Historic Lows: Is South Korea's Semiconductor Valuation Gap a Trap or an Opportunity?

Deep News
Sep 29

South Korea's memory chip market is facing a rare pricing divergence: the market is pricing in a cyclical peak with extremely pessimistic valuations, while analysts' earnings forecasts continue to be revised upward. Behind this contradiction lies a deep tug-of-war between the massive withdrawal of foreign capital and the expansion of HBM demand.

According to Goldman Sachs data, foreign investors' holdings in South Korea's KOSPI semiconductor sector have fallen to 49.1%, which is 2.1 standard deviations below the historical average, and this position loss is almost entirely concentrated in the memory chip segment. Meanwhile, the consensus earnings-per-share forecast for South Korea's tech sector in 2027 has been revised up by as much as 122% over the past six months, and the divergence between earnings forecasts and market pricing has reached a level that is difficult to ignore.

The core variable in this divergence is the demand trajectory for HBM (High Bandwidth Memory). TrendForce raised its 2027 HBM price outlook in September and expects HBM to absorb 30% of DRAM wafer input in 2027, up from 22% in 2026. If this trend holds, traditional DRAM supply may remain persistently tight, and the duration of the memory cycle could far exceed current market expectations.

Foreign Positions at Extreme Lows, Concentrated in the Memory Segment

The foreign holdings structure in South Korea's semiconductor sector shows a highly uneven pattern. According to Goldman Sachs data, foreign holdings in the overall KOSPI semiconductor sector are 2.1 standard deviations below the historical average, but if Samsung Electronics and SK Hynix are excluded, foreign holdings in the remaining constituents are close to normal levels.

This means that the foreign capital retreat is not a broad avoidance of South Korea's tech sector, but rather a precise concentration in the most fiercely debated sub-segment: memory chips. The market's concern about a memory cycle peak has been fully reflected through the extreme skew in positioning.

Earnings Forecasts Keep Rising, Diverging from Market Pricing

However, the signals on the earnings side are diametrically opposed to market pricing. The consensus EPS forecast for South Korea's tech sector in 2027 was revised up 8% over the past three months, with a cumulative upward revision of 122% over the past six months. Although this data covers the broader tech sector rather than pure memory chip estimates, its directional signal is difficult to ignore.

The market is pricing in cyclical peak risk, while the analyst community continues to raise earnings forecasts. This divergence itself constitutes the most central tension in the current investment logic for South Korean semiconductors.

4.7x P/E Ratio: Deeply Undervalued or a Peak Trap?

From a valuation perspective, South Korea's semiconductor sector currently trades at just 4.7 times forward earnings, compared to a historical average of 10.6 times, representing a staggering discount.

But this low valuation in itself does not constitute a margin of safety. The memory chip industry has an inherent rule: when earnings approach a peak, valuations often appear cheapest. If the current level indeed represents peak earnings, the earnings figure in the denominator could decline rapidly, and at that point the 4.7x P/E ratio would no longer be undervaluation but rather a classic peak-earnings trap.

The key question is whether we are truly at peak earnings right now.

HBM Expansion May Extend the Traditional DRAM Boom Cycle

The continued expansion of HBM demand provides an important counterargument to the above question. TrendForce expects HBM to account for 30% of total DRAM wafer input in 2027, a significant increase from 22% in 2026.

More wafer resources flowing to HBM production means that available capacity for traditional DRAM will be squeezed, and the tight supply pattern may persist. The market worries about a memory cycle peak, but the expansion of HBM itself may precisely be one of the reasons the traditional memory cycle can be extended.

It is worth noting that industry forecasts support the continuation of pricing power, but they cannot lock in the margin trajectory for Samsung Electronics or SK Hynix. HBM growth does not necessarily mean the memory cycle is about to reverse, and this logical chain deserves re-examination.

Re-rating or Trap: Key Variables Lie in Contract Pricing and Supply-Demand Data

The current investment logic is not complicated. If 2027 HBM contract prices remain strong and traditional DRAM supply stays persistently tight, the market will have to re-price for a longer earnings cycle. With foreign holdings at extreme lows, once such a re-rating begins, it could unfold in a rather dramatic manner.

Conversely, if supply catches up with demand, the 4.7x P/E ratio will reveal its true nature — that is the classic form of a peak-earnings trap.

The key indicators that will determine direction are the trajectory of 2027 HBM contract pricing, as well as changes in traditional DRAM spot prices and major manufacturers' profit margins. Low valuation itself is not a catalyst — data is.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10