SK hynix said it has completed the world’s first development of high-bandwidth memory 4 (HBM4) and is now ready to begin mass production of the advanced chip.
Next-generation chip for AI
HBM4 is the latest generation of stacked memory designed for artificial intelligence accelerators and high-performance computing. The chip is engineered to dramatically increase data throughput while maintaining power efficiency, enabling faster training and inference for AI models.
Cho Joo-hwan, who leads SK hynix’s HBM development, said the company is focused on meeting customer requirements in performance, energy savings, and reliability as demand surges in the AI computing market.
Partnership with Nvidia
Industry observers expect SK hynix’s HBM4 to be integrated into Nvidia’s next-generation graphics processing units (GPUs) as early as the end of this year. Nvidia, the market leader in AI accelerators, has been one of the main adopters of earlier SK hynix HBM versions, from HBM2E through HBM3E.
SK hynix was also the first company to mass-produce HBM chips a decade ago. With HBM4 entering volume production, the company aims to maintain its lead in a market critical to powering AI infrastructure worldwide.
Comments