SK hynix reportedly merges DRAM and NAND into unified High-Bandwidth Storage package to boost on-device AI performance

A pair of SK hynix chips (Image source: SK hynix)

SK hynix is reportedly developing a new High-Bandwidth Storage (HBS) package that merges DRAM and NAND chips into one ultra-efficient module. The innovation promises faster data processing, better heat control, and enhanced AI performance for next-generation smartphones and tablets.

David Odejide, Published 11/12/2025

AI CPU Smartphone Storage

SK hynix has reportedly achieved a major milestone with an innovation that could change how mobile devices handle AI. The company’s new hybrid architecture, High-Bandwidth Storage (HBS), merges DRAM and NAND, enabling higher data speeds, lower latency, and improved energy efficiency.

According to ETNews, SK hynix expects its new tech to boost smartphone AI performance.

Next-gen memory for AI-powered mobile devices

The key part of SK hynix’s latest innovation is Vertical Wire Fan-Out (VFO) technology. VFO is a method that allows up to 16 layers of DRAM and NAND to be stacked vertically while they are connected in a straight line. With this process, SK hynix can reduce signal loss and transmission distance because there is no need for the curved wiring common in traditional packages.

The outcome is an HBS package with a shorter data processing time and increased efficiency. This brings bandwidth improvements to mobile devices, similar to how High-Bandwidth Memory (HBM) boosted performance in GPUs and AI servers.

Smaller, faster, and cooler

According to SK hynix, its new VFO process cuts wiring requirements by 4.6 times. It reduces power consumption by 5 percent and improves heat dissipation by 1.4 percent. The South Korean company also slashed the packaging height by 27 percent, allowing mobile devices to be even slimmer and to operate at lower temperatures.

SK hynix is also saving on production costs because HBS does not require Through-Silicon Via (TSV) manufacturing, which vertically connects stacked chips by drilling and filling microscopic holes through the silicon wafers, unlike HBM. In addition, this helps to increase yield, which is crucial for scaling production.

Bringing AI performance to mobile chips

When paired with the application processor (AP), SK hynix’s HBS module will enable smartphones and tablets to handle AI workloads better. The company has already used VFO-based DRAM packaging in Apple’s Vision Pro.

Source(s)

TrendForce, ETNews

Loading Comments

Comment on this article

TrimUI Smart Pro S debuts as new re...

Xiaomi 17 Ultra: Details of first l...

David Odejide - Tech Writer - 343 articles published on Notebookcheck since 2024

I am a writer and software developer with a background in Mechanical Engineering and a distinction MSc in Environmental Management (Energy). My career spans fullstack development (building desktop, web, and Android apps for enterprise clients), Content Strategist/Business Developer in the solar industry, and writing research-driven articles on electric vehicles, renewable energy, and consumer tech. I previously ran WindowsFoneFans (remember when Windows phones were a thing?) as a social media hub for Windows phone enthusiasts. My work includes hundreds of video scripts on EVs and sustainability, with over 100 million views on YouTube. I also write about chipsets, mobile hardware, and emerging IT trends, drawing on years of hands-on development experience. Whether it is decoding Snapdragon nodes or analyzing offshore wind policy, I aim to deliver content that is both technically sound and widely accessible. In my spare time, I manage one of the largest LinkedIn groups for wind energy professionals and enjoy getting lost in Scotland’s natural beauty.

Please share our article, every link counts!