Nvidia deepens South Korea AI push

Nvidia has signed a wide set of artificial intelligence and data centre agreements with major South Korean companies, placing the country at the centre of its next phase of AI infrastructure expansion across Asia.

The agreements involve SK Telecom, SK Hynix, Naver, Doosan Group, LG Group, Hyundai Motor Group and continued discussions with Samsung Electronics, linking Nvidia’s graphics processors, AI factory architecture and robotics platforms with South Korea’s strengths in memory chips, telecoms, manufacturing, cloud services and industrial automation.

The deals were unveiled during Nvidia chief executive Jensen Huang’s Seoul visit, which has underlined how South Korea is moving from being a supplier of critical memory chips to a broader AI infrastructure hub. The push comes as governments and large companies seek sovereign computing capacity, tighter control over data, and dedicated facilities capable of training and running advanced AI models.

SK Telecom plans to build a gigawatt-scale AI cloud in South Korea using Nvidia’s DSX platform, with the first AI factory expected to come online in 2027. The infrastructure is designed to support enterprise AI, robotics, industrial automation and agentic AI services, using SK Telecom’s network, data centre and corporate technology base.

Naver, one of South Korea’s leading internet and cloud companies, is expanding its sovereign AI infrastructure with Nvidia, starting with 55 megawatts of capacity and a plan to move towards gigawatt scale. The collaboration is aimed at serving businesses, public-sector users and industries seeking locally controlled AI platforms.

SK Hynix, already a key supplier of high-bandwidth memory for Nvidia’s AI accelerators, has entered a multi-year technology partnership to develop next-generation memory for global AI data centres. The agreement strengthens the role of high-bandwidth memory as one of the most important bottlenecks in the AI hardware supply chain, with demand rising sharply as large models require faster data movement between processors and memory.

Samsung Electronics remains part of the wider conversation around next-generation foundry and memory cooperation. Its semiconductor leadership held talks with Nvidia on advanced chip manufacturing and future high-bandwidth memory, including technologies beyond the current HBM4 generation. Samsung is also involved in manufacturing AI accelerator chips for other customers, keeping it central to Nvidia’s broader supply chain calculations even as SK Hynix maintains a strong position in premium memory.

Doosan Group’s cooperation with Nvidia extends the agreements beyond data centres into physical AI. Doosan Robotics is working with Nvidia’s simulation, robotics and on-device AI platforms to develop an agentic robot operating system, while Doosan Enerbility is expected to support power infrastructure for AI factories. Doosan Corporation’s electronics materials unit is also positioned to support next-generation data centre hardware through advanced materials used in AI accelerators.

LG Group is working with Nvidia on an AI factory focused on physical AI, mobility and data centre technologies. The partnership is expected to support LG’s work in robotics, autonomous driving, smart manufacturing and GPU cloud services, giving the conglomerate a platform for applying AI across consumer electronics, industrial systems and vehicle technologies.

Hyundai Motor Group’s cooperation with Nvidia is centred on mobility, manufacturing and robotics, including support for Hyundai’s AI Valley project in Saemangeum. The group’s ownership of Boston Dynamics gives it a strong position in humanoid and industrial robotics, while Nvidia’s platforms can support simulation, model training and factory deployment.

South Korea’s national AI strategy gives the corporate agreements wider significance. The country had already announced plans to secure more than 260,000 Nvidia GPUs, with more than 50,000 intended for public AI infrastructure and the rest allocated to major companies including Samsung, SK Group, Hyundai Motor Group and Naver. The latest agreements suggest that those commitments are moving into more defined industrial projects.

The expansion also reflects a shift in Nvidia’s business model. The company is no longer selling chips only as components; it is promoting full-stack AI factories that combine processors, networking, software, reference architecture and operations. That approach gives Nvidia deeper influence over how data centres are designed and operated, while offering partners faster deployment at a time when AI capacity remains constrained.

South Korea’s Science and ICT Ministry is also seeking priority access to Nvidia’s Vera Rubin GPUs, amid expectations that next-generation chip supply could face delays. The effort shows how advanced AI processors have become strategic assets for countries competing in model development, cloud infrastructure and industrial automation.

The agreements carry risks as well as opportunity. AI factories require vast power supplies, expensive cooling systems and long-term commitments from corporate customers. Heavy dependence on one US chip supplier may also raise questions over supply security and bargaining power, even for a country with world-class semiconductor companies.

For Nvidia, the South Korean partnerships strengthen memory access, broaden its AI factory customer base and create new demand in robotics, cloud services and industrial AI. For South Korean companies, the deals offer a faster route into high-value AI infrastructure at a time when global competition is shifting from model development to the computing systems needed to run them at scale.