[News] SK hynix Debuts 16-Layer 48GB HBM4 at CES 2026, Alongside SOCAMM2 and LPDDR6

As CES approaches, SK hynix is set to highlight its flagship next-generation memory product. According to the company’s press release, it will debut a 16-layer HBM4 with 48GB. The new offering represents the next generation of its 12-layer HBM4 (36GB), which previously demonstrated industry-leading speeds of 11.7 Gbps, and is currently under development in line with customer schedules.

Beyond the 16-layer HBM4, the company is also set to showcase its 12-layer HBM3E with 36GB, which is expected to be a key driver of the memory market this year. Notably, SK hynix will co-exhibit AI server GPU modules incorporating HBM3E alongside a customer, underscoring the technology’s role within AI systems.

SOCAMM2 and LPDDR6 in Focus

In addition to HBM, the company will also present SOCAMM2, a low-power memory module designed for AI servers, along with LPDDR6, which is optimized for on-device AI and delivers markedly higher data throughput and improved power efficiency compared with earlier generations.

321-Layer 2Tb QLC NAND Targets AI Data Center Demand

Turning to NAND flash, the company will showcase its 321-layer 2Tb QLC product, designed for ultra-high-capacity eSSDs, as demand rises alongside the rapid expansion of AI data centers.

At the event, the company will also set up an “AI System Demo Zone,” where visitors can experience how its future-oriented AI system memory solutions interconnect to form a broader AI ecosystem. The zone will feature custom HBM optimized for specific AI chips or systems, PIM-based AiMX, CuD enabling computing in memory, CMM-Ax integrating computing capabilities into CXL memory, and data-aware CSD solutions.

Rubin Momentum Highlights NVIDIA–SK hynix HBM Discussions

Notably, according to iNews24, industry sources say SK hynix CEO Kwak Noh-jung and AI Infrastructure President (CMO) Kim Joo-sun held a private meeting with NVIDIA officials immediately after CEO Jensen Huang concluded his special keynote at CES 2026. The industry is closely watching whether HBM demand could rise further as Rubin-based systems scale. As iNews24 indicates, discussions may have focused on next-generation product roadmaps and future supply cooperation.

During the keynote, NVIDIA CEO Jensen Huang also provided an update on the company’s next-generation AI accelerator, Rubin, saying the chip is now in full production and that customers are expected to begin using it within the year, according to iNews24.

Read more

• [News] SK hynix 2026 Outlook: HBM3E Dominates, HBM4 Dual Strategy Amid 3 Market Headwinds
• [News] SK hynix Reportedly Plans First U.S. 2.5D Packaging Line, Eyes Turnkey HBM to Challenge TSMC

(Photo credit: SK hynix)