Nvidia will use DRAM memory chips from SK Hynix for its new Vera data-center CPU, CEO Jensen Huang said June 7 [1, 2].
The deal secures critical high-bandwidth memory for Nvidia as it expands its AI infrastructure. With the industry facing a projected memory shortage that could last for years, the partnership ensures a stable supply chain for the company's next-generation hardware [4, 5].
Huang made the announcement during a series of deals in Seoul, South Korea [2, 5]. The agreement is part of a multi-year technical partnership focused on the development of AI factories [2].
"We are excited to work with SK Hynix to secure memory for our next-generation AI infrastructure," Huang said [2].
The collaboration comes as demand for AI processing power accelerates globally. Huang said that the relationship between the two companies has already seen significant growth. "We had a very big year this year with SK Hynix, and we are preparing for a very, very large second half of the year and next year," Huang said [1].
Securing these components is a priority for Nvidia as it seeks to maintain its lead in the AI chip market. The company is moving to lock in supply agreements to avoid production delays caused by the scarcity of specialized memory [5].
"Memory shortage could last for years, so partnerships like this are critical," Huang said [5].
“"Memory shortage could last for years, so partnerships like this are critical."”
This partnership signals that the bottleneck for AI expansion has shifted from raw processing power to memory availability. By securing a multi-year deal with SK Hynix, Nvidia is mitigating the risk of supply chain disruptions that could otherwise stall the rollout of the Vera CPU and the broader build-out of AI factories.
