The key question isn't whether memory demand is collapsing, but whether this selloff reflects fundamentals or simply a valuation reset triggered by tighter liquidity and leverage unwinding.
HBM demand from AI accelerators remains exceptionally strong, and supply is still constrained. On that front, SK Hynix arguably has the stronger near-term positioning given its lead in HBM and close relationship with NVIDIA. However, that advantage is also why Hynix traded at a much richer valuation and became more vulnerable to profit taking.
Samsung is the more diversified bet. If it successfully closes the HBM technology gap and wins larger AI memory share, today's correction could look like an attractive entry point for long-term investors.
My view: this feels more like a policy-driven shakeout than the end of the AI memory cycle. The bigger risk isn't demand disappearing, but AI capex slowing faster than expected in 2027–2028.
If forced to choose: Hynix for higher growth, Samsung for better risk-adjusted returns.
Curious what others think — is this a buying opportunity or the first crack in the AI memory supercycle?
Comments