Nvidia and SK hynix have joined forces on a new storage technology expected to revolutionize AI infrastructure. The two companies are developing a new solid-state drive (SSD) specifically designed for AI inference processes, offering performance up to 10 times faster than existing enterprise SSDs.
The SSD Revolution is Coming
Announced by SK hynix Vice President Kim Cheon-seong, this partnership extends the successful collaboration between the two giants in high-bandwidth memory (HBM) to the NAND flash storage sector. The project is codenamed “Storage Next” by Nvidia and “AI-NP” (AI NAND Performance) by SK hynix.
This next-generation SSD aims to achieve 100 million input-output operations per second (IOPS), setting it apart from standard storage units. This performance level, far exceeding the capacity offered by the most powerful existing enterprise SSDs, aims to eliminate the sharp boundaries between memory and storage in AI systems.
The constant need for access to datasets and parameters, especially in large-scale AI models, poses a bottleneck for traditional HBM and DRAM technologies. The new architecture fundamentally solves this access problem by creating a NAND flash-based “half-memory” layer.
The technology is currently in the proof-of-concept (PoC) phase. The first physical prototype, developed as part of the partnership, is planned to be completed and tested by the end of 2026.
Fully operational and ready-for-mass production is targeted for market launch in 2027. SK hynix is also working with SanDisk on a new standard called “High Bandwidth Flash” (HBF) to support this process.
{{user}} {{datetime}}
{{text}}