As a leading company in the field of memory chips, SK Hynix is ​​accelerating the expansion of its production capacity to meet the industry's rising chip demand. The company is currently preparing new technologies in an effort to further enhance the AI ​​workload processing capabilities on mobile and edge devices.

According to multiple Korean media reports, SK Hynix is ​​developing new computer memory to accelerate local AI computing. This new technology, called High Bandwidth Storage (HBS), is an expansion of the previous High Bandwidth Flash (HBF) solution. It merges mobile DRAM and NAND flash memory components into the same device and is specifically designed to accelerate the AI ​​load of mobile devices such as smartphones and tablets. The chip can vertically stack up to 16 layers of DRAM and NAND, and achieve inter-layer interconnection through a unique vertical wire fan-out structure (VFO).

SK Hynix has already used VFO DRAM in Apple Vision Pro products, but HBS technology further integrates NAND flash memory. When released in 2023, the company emphasized that VFO not only improves packaging efficiency, but also improves heat dissipation and chip size reduction. Compared with the traditional curved wire connection method, VFO can reduce the space requirement for inter-layer electronic transmission by 4.6 times, increase the overall energy efficiency by 4.9%, improve heat dissipation by 1.4%, and reduce the chip thickness to 73% of the traditional solution.

Different from the HBF technology developed in cooperation with SanDisk, HBS does not require through silicon vias (TSV), the manufacturing process is simpler, the yield rate is higher, and the production cost is lower, which is conducive to widespread promotion in the semiconductor industry.

The new DRAM+NAND stack module will be directly co-packaged with the application processor (AP), significantly improving the data processing speed of smartphones and SoC terminals. SK hynix’s goal with this innovation is to further improve mobile AI performance. Although the specific performance has yet to be tested in practice, the company plans to officially release the technology from 2029 to 2031. Currently, due to the booming sales of new-generation chips in 2026, SK Hynix's production capacity has become tense.