Nvidia has entered a multi-year agreement with South Korean chipmaker SK Hynix to secure high-bandwidth memory supplies and develop large-scale AI data centers. The partnership also involves an expansion into cloud services through SK Telecom and a $1 billion investment by Nvidia into the cloud provider Naver.
The Strategic Partnership
Nvidia has formalized a significant supply agreement with SK Hynix to guarantee access to critical high-bandwidth memory (HBM). This partnership, announced at an AI summit in San Francisco, is projected to reach a total value of $500 billion over a multi-year period. By securing this supply, Nvidia aims to mitigate the global memory shortage that has hindered the production of its advanced graphics processing units (GPUs). Raj Mirpuri, Nvidia's enterprise vice president, highlighted that the collaboration includes a co-development arrangement for next-generation AI memory, which is essential for maintaining a stable supply chain for the company's proprietary systems.
Data Center Expansion and Infrastructure
Beyond memory procurement, the deal outlines a massive expansion in physical data center infrastructure. The project targets a staggering 2 gigawatts of total power capacity, necessitated by the deployment of hundreds of thousands of Nvidia's advanced GPUs. As part of the broader agreement, SK Hynix's affiliate, SK Telecom, will launch a new cloud computing venture utilizing Nvidia's Vera Rubin systems. This move signals a shift in the AI industry where, rather than relying solely on traditional hyperscalers, massive conglomerates and government-supported entities are beginning to build their own dedicated AI computing environments to ensure they remain competitive in the face of rapidly rising demand.
Additional Investments and Regional Focus
The influence of South Korean tech firms in the global AI landscape was further underscored during the summit, which was attended by President Lee Jae Myung. In a move to broaden its ecosystem, Nvidia also announced a $1 billion investment into Naver, a prominent Korean cloud service provider. This funding is specifically earmarked for the construction of data centers centered around Nvidia’s hardware. This initiative is designed to offer both domestic and international customers early access to AI computing power. These data centers are expected to provide 200 megawatts of capacity, further cementing the region's role as a key hub for global AI infrastructure development.
Broader Industry Implications
The announcement coincided with separate industry developments, indicating a wider trend of consolidation and partnership among tech giants. Samsung Electronics disclosed a memorandum of understanding with Broadcom valued at approximately $200 billion to foster advancements in memory and foundry technologies. These concurrent deals reflect a market-wide race to secure hardware components as the artificial intelligence boom continues to strain existing supply chains. Analysts view SK Hynix's recent Nasdaq listing, which occurred earlier this month, as a tactical step to provide the necessary capital to meet these massive infrastructure requirements, highlighting the immense financial stakes involved in the current AI transition.
⚖ The Balanced View
Supporting view
The partnership is seen as a necessary move by Nvidia to ensure a stable supply of HBM, which is a critical bottleneck in current GPU production, while SK Hynix gains long-term capital and a guaranteed massive customer for its leading-edge memory technology.
Concerns & criticism
The massive scale of the buildout, requiring 2 gigawatts of power, underscores the extreme resource intensity of modern AI infrastructure and the potential strain on energy and supply chains as conglomerates race to monopolize capacity.
→What's next
The planned data center projects are expected to begin coming online in 2027. Future developments will focus on the progress of the joint co-development of next-generation memory and the scaling of the 2-gigawatt infrastructure project.