The evidence so far points more towards a policy-driven valuation reset than a collapse in the memory super-cycle.
The sell-off was triggered by two macro factors arriving together:
The Bank of Korea unexpectedly raised its policy rate by 25 basis points to 2.75%, its first increase in three and a half years, and signalled that further tightening is possible. Higher interest rates reduce the present value of future earnings, which particularly affects stocks that had risen sharply on AI optimism.
Korean regulators have also been tightening scrutiny of single-stock leveraged ETFs, which can amplify both rallies and sell-offs through daily rebalancing.
That does not automatically mean the underlying demand has deteriorated. The key drivers of the current cycle remain largely unchanged:
HBM demand from AI accelerators remains strong.
Server DRAM pricing has stayed firm after substantial contract price increases earlier this year.
AI infrastructure spending is still supporting premium memory products, even if investors have become more cautious about valuations.
The risk to watch is whether this becomes more than a macro correction. If cloud providers begin delaying GPU deployments, hyperscalers cut AI capex, or HBM order growth slows materially, then the memory cycle itself would deserve re-evaluation. At present, there is little concrete evidence that this has happened.
SK Hynix vs Samsung
SK Hynix
Clear leader in HBM.
Most direct beneficiary of the AI boom.
Highest earnings leverage if HBM demand continues.
Also the more volatile stock because expectations are extremely high.
Samsung
More diversified across memory, foundry, smartphones, consumer electronics and displays.
Offers greater resilience if memory pricing weakens.
Still has upside if it narrows the HBM technology gap.
If I expected the AI memory upcycle to continue for another 12 to 24 months, I would favour SK Hynix because it has the strongest exposure to the highest-margin segment.
If I expected macro uncertainty, higher interest rates and multiple compression to dominate over the next few quarters, I would lean towards Samsung, whose diversified businesses provide a cushion.
Overall, I would currently rate the situation as approximately:
70% probability: policy and positioning driven correction while the memory upcycle remains intact.
30% probability: the market is correctly anticipating that AI-related memory demand is approaching a cyclical peak.
The next major test will be upcoming earnings and management guidance. If HBM shipments, pricing and order backlogs remain strong despite this sell-off, this episode is likely to be remembered as a valuation reset rather than the end of the memory super-cycle.
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.
