Nvidia has signed a multi-year agreement with SK Hynix to co-develop and manufacture next-generation AI memory, strengthening its position ahead of the launch of the Vera Rubin platform. The deal gives SK Hynix a deeper role in Nvidia’s future roadmap and highlights how memory has become one of the most important constraints in the AI industry.
The partnership goes beyond a traditional supplier relationship. Nvidia and SK Hynix will work together on future high-bandwidth memory technologies designed for large-scale AI infrastructure, including the systems that will power upcoming AI factories and data centers.
Memory has become the biggest challenge in AI infrastructure
While GPUs have traditionally received most of the attention in the AI race, memory is increasingly emerging as the industry’s most significant bottleneck. Modern AI systems require enormous amounts of high-bandwidth memory to feed increasingly powerful processors.
Nvidia’s Vera Rubin platform is expected to rely heavily on HBM4 memory. Industry estimates suggest that SK Hynix could supply between 60% and 70% of the HBM4 volume allocated to Vera Rubin systems, placing the company ahead of Samsung and Micron in the race to support Nvidia’s next generation of AI hardware.
The agreement also gives SK Hynix greater certainty to expand production capacity as demand for advanced memory continues to rise. Analysts expect HBM supply constraints to remain a major challenge for the industry for several more years.
Nvidia recently confirmed that Samsung, SK Hynix and Micron have all qualified to supply HBM4 memory for Vera Rubin. However, the new co-development agreement signals a closer strategic relationship between Nvidia and SK Hynix than a standard vendor arrangement.
The announcement came during Nvidia CEO Jensen Huang’s visit to South Korea, where the company also unveiled new partnerships focused on AI infrastructure, cloud computing and industrial automation.
As AI systems continue to grow in scale, securing access to advanced memory is becoming just as important as developing faster processors. Nvidia’s latest move suggests the company is preparing for that reality years before its next generation of AI platforms reaches full deployment.
