SK Hynix has begun supplying samples of its next-generation HBM4E memory chips to major customers, aiming to strengthen its leadership in the rapidly expanding AI memory market. The South Korean chipmaker stated that it successfully delivered the 12-layer HBM4E samples to customers as planned, highlighting the company’s expertise in high-bandwidth memory (HBM) development and production. Mass production is expected to begin after close collaboration with customers and partners.
Designed for artificial intelligence applications, the new chip delivers data transfer speeds of up to 16 Gbps per pin and offers more than 20% better power efficiency than previous generations. These improvements are expected to enhance AI training and inference workloads by enabling faster and more efficient data processing.
SK Hynix said the HBM4E also reduces data transfer latency through an upgraded interface and optimised design, while maintaining stable performance in high-bandwidth computing environments. Using its advanced MR-MUF packaging technology, the company has achieved a 48GB capacity in a 12-layer stack and improved heat resistance by 17% compared with HBM4.
HBM chips play a critical role in AI processors by providing the high-speed memory needed to handle massive datasets. They are widely used in AI accelerators, including those produced by Nvidia.
Also Read- Samsung Takes BYD, Google, AMD Orders Amid TSMC Shortage
As the leading supplier of HBM chips to Nvidia, SK Hynix is looking to maintain its competitive edge as rivals Samsung and Micron ramp up efforts in the AI memory segment. The company said its experience with HBM3, HBM3E, and HBM4 products provides a strong foundation for supporting next-generation AI infrastructure and meeting growing demand for advanced memory solutions.
