SEOUL, SOUTH KOREA – SK hynix, a global leader in semiconductor manufacturing, announced today the successful development of the world’s first High Bandwidth Memory 4 (HBM4) chip and has established a mass production system, marking a significant milestone in the AI memory market. This achievement is set to solidify the company's leading position, following its successful supply of HBM3E chips to Nvidia last year.
HBM technology is critical for next-generation artificial intelligence (AI) systems, as it vertically stacks DRAM chips to enable faster data transfer speeds compared to conventional DRAM. This innovation is crucial for powering data-intensive applications and AI servers by accelerating the data handling capabilities of graphics processing units (GPUs).
According to SK hynix, the new HBM4 chip doubles bandwidth by adopting 2,048 input/output (I/O) terminals, a significant leap from the previous generation's 1,024. The chip also boasts an operating speed of over 10 gigabits per second (Gbps), surpassing the Joint Electron Device Engineering Council (JEDEC) industry standard of 8 Gbps. This increased performance translates into a potential 69 percent improvement in AI service performance while simultaneously improving power efficiency by over 40 percent, which could lead to substantial energy savings in data centers.
To prepare for mass production, SK hynix has utilized its advanced Mass Reflow Molded Underfill (MR-MUF) stacking method in conjunction with its fifth-generation 1b, 10-nanometer process technology. The MR-MUF process involves injecting liquid protective materials between stacked chips to enhance heat dissipation and ensure the stability required for large-scale production. The company stated that this technological approach minimizes production risks and ensures the reliability and stability of the final product.
Industry analysts forecast that SK hynix's HBM4 chips could initially be priced 60 to 70 percent higher than their predecessor, a price that is expected to gradually decrease as competitors like Samsung Electronics and Micron Technology enter the market. Samsung is also actively preparing for HBM4 mass production, planning to use a more advanced 1c, sixth-generation 10nm DRAM technology for its version of the chip.
"Completion of HBM4 development will be a new milestone for the industry," said Cho Joo-hwan, Head of HBM Development at SK hynix. "By supplying the product that meets customer needs in performance, power efficiency, and reliability in a timely manner, the company will fulfill time to market and maintain a competitive position."
SK hynix's HBM4 is expected to be integrated into Nvidia’s next-generation GPU platform, Rubin, which is slated for release in the second half of 2026. This partnership will further solidify SK hynix's role as a key supplier in the rapidly expanding AI sector. The company's President and Head of AI Infra, Kim Ju-seon, emphasized that HBM4 represents a "symbolic turning point beyond the AI infrastructure limitations" and will be a crucial product for overcoming technological challenges.
[Copyright (c) Global Economic Times. All Rights Reserved.]
