Performance improved 30 times! NVIDIA Corporation (NVDA.US) collaborates with SK Hynix to develop AI solid state drives, aiming for mass production in 2027.

According to reports, NVIDIA Corporation (NVDA.US) is collaborating with South Korea's SK Hynix to develop a solid-state drive (SSD) with artificial intelligence as its core, using NAND chips. This project is internally dubbed as "Storage Next" at NVIDIA Corporation and "AI-NP" (AI NAND Performance) at SK Hynix, aiming to break the current bottleneck in the storage end of the AI computational system. The jointly developed AI SSD boasts a significant upgrade in performance metrics. Its core goal is to achieve up to 1 billion input/output operations per second (IOPS), a value approximately 10 times faster than the most high-end enterprise SSDs currently available, and even up to 30 times faster in specific environments. The purpose of this device is to act as a key "intermediate layer" between high-bandwidth memory (HBM) and traditional large-capacity storage, ensuring that Beijing Vastdata Technology in AI inference processes can be instantly accessed by GPUs with extremely high throughput and low latency, significantly enhancing the response speed of generative AI and reducing overall energy consumption. In terms of technical support, SK Hynix is working with its partner SanDisk (SNDK.US) to advance a new standard called HBF (High Bandwidth Flash), a vertically stacked structure similar to HBM that will provide the necessary hardware foundation for this AI SSD. According to the timeline disclosed by both parties, the first product prototype is expected to be unveiled in the second half of 2026, with plans for mass production in 2027. At that time, this new storage solution is expected to be deeply integrated with NVIDIA Corporation's future Rubin CPX series GPUs and other cutting-edge hardware. This collaboration not only signifies NVIDIA Corporation's efforts to further strengthen its vertical integration capabilities within its AI ecosystem but also signals SK Hynix's attempt to establish new dominance in the AI NAND storage field after leading in the HBM memory market. However, this trend has also raised concerns within the supply chain, with industry predictions suggesting that as AI inference demand surges, the NAND flash memory market may experience a reoccurrence of supply shortages and price fluctuations similar to what happened in the DRAM market.