SK Hynix has built a nice business selling high-bandwidth memory to Nvidia Corp. (NVDA), but apparently that's not enough. As AI workloads keep expanding and straining the limits of existing storage technology, the South Korean chipmaker is now extending that partnership into solid-state drives designed specifically for AI systems.

The logic makes sense. AI computing is shifting from pure training workloads toward inference, where you need storage that's not just fast but also capacious and energy-efficient. SK Hynix is taking the same customer-centric design approach that worked for its HBM business and applying it to NAND flash storage, working directly with Nvidia to build SSDs that actually fit what large-scale AI systems need.

Three-Pronged NAND Strategy

SK Hynix isn't putting all its eggs in one basket here. According to a report by Chosun Biz on Tuesday, the company has divided its NAND development into three core areas. First, there's the Nvidia partnership focused on boosting SSD performance. Second, they're collaborating with SanDisk to establish standards for high-bandwidth flash (HBF), which is essentially a layered NAND architecture aimed at handling the intensifying demands of AI applications. Third, they're developing large-capacity storage that balances performance with cost efficiency.

The goal across all three initiatives is straightforward: reduce data bottlenecks, increase processing speeds, and support AI inference at scale. By combining higher speeds, greater capacity, and improved power efficiency, SK Hynix is trying to overcome the limitations that are holding back current AI memory solutions.

HBM4 Shipments Are Already Flowing

Meanwhile, the core memory partnership with Nvidia continues to deepen. Chosun Daily reported on Monday that SK Hynix has begun supplying its sixth-generation high-bandwidth memory, HBM4, to Nvidia. The company has delivered between 20,000 and 30,000 final sample units that reportedly meet all of Nvidia's specifications.

That memory is now integrated into Nvidia's next-generation Rubin GPU for final customer testing. Once Rubin moves into mass production, SK Hynix is expected to significantly ramp up HBM4 shipments. The company has already finalized overall supply volumes and pricing terms with Nvidia, which suggests this isn't just a test run but a long-term commitment.

All of this positions SK Hynix as more than just a memory supplier. They're becoming a broader infrastructure partner for AI systems, moving into storage just as that bottleneck becomes more apparent. It's a logical evolution for a company that's already deeply embedded in the AI supply chain.

NVDA Price Action: Nvidia shares were up 0.71% at $177.55 at the time of publication on Tuesday, according to market data.