NVIDIA is currently engaged in a competitive quest to secure DRAM for its AI accelerators and memory products, leveraging the company’s distinctive relationships with leading South Korean memory manufacturers.
NVIDIA’s CEO Hosts We Chairman for Dinner: Conversations on SOCAMM and HBM4
The demand for DRAM has skyrocketed, posing significant challenges for AI chip producers who have experienced shortages for several quarters. This surge in demand is primarily fueled by hyperscalers as well as an increased reliance on various general-purpose DRAM types, including LPDDR, DDR, and GDDR. However, HBM remains critical for NVIDIA and its peers in developing AI accelerators, making it imperative for the company to secure long-term supplies. Thanks to its proactive long-term agreements (LTAs) and the strategic diplomatic efforts of CEO Jensen Huang, NVIDIA is in a favorable position to mitigate risks related to memory shortages.
Reflecting on past engagements, it was noted that Jensen Huang met with executives from Samsung last October. The memorable atmosphere of a “fried chicken”dinner allowed for informal interaction, with images capturing the moment of Huang serving dishes to Samsung’s Jay Y. Lee.

The recent dialogues reportedly centered around SOCAMM, a low-power memory module crucial for the next-generation Rubin AI servers, and the production of HBM4 modules. These discussions highlight the importance of collaboration between NVIDIA and its partners.
When questioned about the potential impacts of memory shortages on NVIDIA, Huang referenced the early LTAs that place NVIDIA ahead in supply chain management compared to its competitors. This highlights how effective leadership and connection-building in the industry continuously position Team Green advantageously.
Leave a Reply