Samsung Developing 8-Layer HBM4E Custom-Built for Nvidia, Targeting Transfer Speeds Up to 18Gbps — BigGo Finance
Samsung (005930.KS) is developing an 8-layer HBM4E memory chip for Nvidia (NVDA), targeting 18Gbps transfer speeds, a 20% improvement. Nvidia's NVHBM platform aims to enhance AI GPU performance, with Samsung holding an integration advantage. SK Hynix (000660.KS) is outsourcing base die production to TSMC, focusing on design capabilities. Success could reshape the AI memory market.
How this was made
The 30-second read
Why it matters
The news signals a step forward in AI memory performance, potentially influencing investor sentiment toward memory manufacturers and AI chip makers.
Market read
First‑time disclosure of a custom high‑bandwidth memory design for Nvidia, highlighting competitive dynamics in the AI memory market.
What to watch
Potential supply‑chain constraints for 4‑nm base dies and competition from TSMC‑based HBM solutions could limit Samsung's advantage.
Background
The article details Samsung's custom HBM4E development for Nvidia's upcoming AI GPU platform, contrasting it with SK Hynix's alternative approach and broader industry trends.
Ticker impact
Samsung announced development of an 8‑layer HBM4E memory chip for Nvidia, a new product specification not previously disclosed.
Short‑term upside for Samsung shares on news of a high‑profile Nvidia contract.
First report of a custom HBM4E design for Nvidia; market may price in increased demand for Samsung's advanced process capacity.
Nvidia is specifying 17‑18 Gbps transfer speeds for the new Samsung HBM4E, indicating a material design change for its upcoming AI GPU platform.
Neutral to mildly positive for Nvidia as the news confirms progress on its Rubin Ultra platform.
The announcement is about a supplier's development; Nvidia's own product timeline remains uncertain.
SK Hynix is mentioned as pursuing a different HBM4E strategy, highlighting competitive dynamics in the AI memory market.
Limited immediate impact; longer‑term competitive implications.
The article only references SK Hynix's plans without new concrete milestones.
Market effects
Strengthens the outlook for the high‑end AI memory sector and may spur further investment in advanced DRAM technologies.
Positive for South Korean semiconductor exporters, especially Samsung and potentially SK Hynix.
Relevant to global AI hardware supply chains and investors tracking AI‑driven demand for memory.
Counterpoint
If Samsung's yield or cost targets are not met, the project could delay, dampening any near‑term share uplift.
Key entities
- CompanySamsung Electronics
South Korean semiconductor leader developing 8‑layer HBM4E for Nvidia.
- CompanyNvidia
AI GPU designer specifying new memory requirements.
- CompanySK Hynix
Competitor pursuing a different HBM4E strategy.





