At CES 2024, the world’s most significant consumer electronics and IT expo in Las Vegas, NVIDIA CEO Jensen Huang made unexpected remarks about Samsung Electronics’ high bandwidth memory (HBM) design.
During a global press briefing, Huang suggested that Samsung must revisit its HBM design, implying potential flaws.
HBM, a high-performance memory chip critical for AI accelerators, is built vertically by stacking multiple DRAM chips.
While SK Hynix and Micron, both based in the U.S., are already supplying HBM to NVIDIA, while Samsung Electronics is still in the testing phase.
Huang attributed Samsung’s delays in HBM testing to Korea’s tendency to prioritize speed.
Huang commented, “They have to engineer a new one (HBM), a new design. But they could do it, and they are working very fast.”
This marks the first public acknowledgment of design issues as the primary reason for Samsung’s extended HBM testing phase.
Despite the criticisms, Huang expressed confidence in Samsung’s potential. “I have confidence that Samsung will succeed with HBM memory. I have confidence like tomorrow is Wednesday,” he said.
“Remember Samsung created HBM originally? The very first HBM memory that Nvidia ever used was from Samsung. They will recover, it’s a great company,” Huang added.
Huang further emphasized the importance of Samsung and SK Hynix to NVIDIA, stating that they are two of NVIDIA’s largest suppliers. He praised them as excellent memory manufacturers and hoped they would continue to thrive.
Furthermore, Huang confirmed plans to meet SK Chairman Chey Tae Won during CES and expressed enthusiasm about their upcoming discussion.
These remarks underscore Korean firms’ opportunities and challenges in the competitive global semiconductor industry.
Most Commented