TLDR: SEOUL—SK Hynix has surged about 1,694% in five years, and Nomura raised its price target May 17 to 4,000,000 SKW for about 78% upside. The AI memory boom lifts DRAM and HBM demand, keeping supply tight into 2028.
Key Takeaways:
- AI data centers drive DRAM and stacked HBM demand, boosting memory makers after supply takes longer to catch up.
- Nomura analyst CW Chung lifted SK Hynix price target from 2,340,000 to 4,000,000 SKW May 17, citing AI demand outpacing supply. SK Hynix leads HBM revenue with 57% share.
- If memory supply stays constrained until mid 2028, SK Hynix could trade like a higher growth semiconductor name despite its cyclical reputation.
Investors who missed Nvidia got a different kind of front row seat: the chips that feed the GPUs. SK Hynix now looks like the part of the AI supply chain that takes the longest to build, so hope has a way of sticking.
Investors who missed Nvidia got a different kind of front row seat: the chips that feed the GPUs. SK Hynix now looks like the part of the AI supply chain that takes the longest to build, so hope has a way of sticking.
Q&A
Why does HBM leadership matter more than DRAM volume when AI clusters keep scaling?
HBM is closer to the GPU compute bottleneck because it stacks bandwidth where the model runs. That makes HBM capacity a tighter, higher value constraint than general DRAM throughput.
What happens to memory stocks if AI demand stays strong but HBM manufacturing ramps faster than expected?
The valuation floor could drop quickly. The bullish thesis depends on constrained supply, so faster ramping would pressure pricing power and compress forward earnings multiples.
How could an ADR listing change the stock market behavior for SK Hynix?
It can widen the buyer pool by reducing friction for US investors. A larger shareholder base may also improve liquidity and reduce discount to peers, tightening spreads around the current value.
If memory companies are cyclical, what is the historical clue that this cycle could last longer?
Long lead times in semiconductor manufacturing and packaging often delay supply response. If AI accelerates faster than those timelines, the cycle can extend beyond typical demand spikes.
What would falsify Nomura's argument that memory should trade like a premium semiconductor stock?
Evidence of sustained oversupply, sharp HBM pricing declines, or a drop in contract demand. Any of those would pull the narrative away from an AI supercycle and back toward a traditional memory cycle.
Top in Stocks
Morningstar questions SpaceX price, xAI alarms investors
Investors scramble for SpaceX IPO edge before debut
Goldman Sachs bets big on NVIDIA as June pressure rises
Nikkei 225 clears 68,000 as AI chip buying surges
No comments yet. Be the first to share your thoughts!