NVIDIA, SK Hynix sign multi-year AI chip deal
What's the story
NVIDIA and South Korea's SK Hynix have announced a multi-year partnership to design and manufacture advanced memory chips for artificial intelligence (AI). The collaboration is expected to give a major boost to SK Hynix as it prepares for a major expansion into high-bandwidth memory (HBM4). This comes as NVIDIA CEO Jensen Huang confirmed that Samsung, SK Hynix, and Micron Technology have been cleared to supply this crucial product.
Chip production
NVIDIA's Vera Rubin systems to use HBM4 memory chips
NVIDIA's Vera Rubin, its most powerful accelerator, is now in full production. Deliveries for this advanced system will begin in the third quarter of this year. These systems are built around clusters of Vera central processing units (CPUs) and Rubin graphics cores, with terabytes of HBM4 in each server system. The partnership will also see SK Hynix supplying a full range of DRAM, high-bandwidth memory and solid-state drives to NVIDIA.
Market expansion
Long-term supply agreements expected to help SK Hynix expand capacity
The long-term supply agreements with NVIDIA are expected to help SK Hynix expand its capacity and gradually increase market share. The company is likely to supply more generations of HBM for AI inference systems, along with HBM for AI training GPUs. Sales of DRAM for CPUs and SSDs used to store inference data are also expected to grow over the next few years.
Strategic alliance
Partnership will strengthen SK Hynix's position as long-term supplier
NVIDIA's AI infrastructure technology partnership will strengthen SK Hynix's position as a long-term supplier of memory products across multiple AI platforms. "Together, we will co-develop the next generation of memory for AI factories and support the accelerating global expansion of AI infrastructure — from frontier model training to agentic and physical AI," Huang said in the statement.