Nvidia Closes Six Deals in Korea with SK Hynix, SK Telecom, LG, Hyundai, Naver, and Doosan
Jensen Huang was in Seoul to announce a package ranging from HBM memory to humanoid robots, with a gigawatt data center from SK Telecom slated for 2027 as the most visible piece.
Seoul as Nvidia's Third Hub
Jensen Huang arrived in Seoul on Monday and closed six deals with major Korean companies in a single day. The array covers five simultaneous fronts and positions Korea in a role comparable to that of Taiwan in manufacturing and India in IT services: Nvidia's third strategic hub outside the United States.
At the core of the package is a partnership of over two years with SK Hynix, currently Nvidia's largest memory supplier, to develop the next generations of HBM for AI data centers. "We already buy billions and billions of dollars from SK Hynix every year, and that is going to grow substantially," Huang stated. The aggregate value of the six agreements was not disclosed.
The Most Relevant GPU Cluster Depends on This
SK Telecom simultaneously announced plans to build a gigawatt-scale AI data center, with the first site expected to go live in 2027. Gigawatt is the capacity benchmark that xAI is already using in Memphis and that OpenAI sought in the Stargate project, thus positioning the Korean operator as a cloud provider for Asian customers looking for an alternative to American hyperscalers. Naver and Doosan will also utilize Nvidia's technology in their own facilities.
LG entered the arena through humanoid robots in a collaboration that utilizes Nvidia's Isaac platform. Hyundai deepened its work in autonomous mobility and AI-assisted manufacturing, expanding on the existing relationship in driver model training. Doosan, which supplies materials for Blackwell chips, closed a dual arrangement: Nvidia adopts the company’s energy solution in data center platforms, while Doosan licenses Nvidia's physical AI technology for its own industrial operations.
The Geopolitical Reading is What Matters to the CIO
The Korean choreography is not coincidental. The agreements were announced at a time when Washington is discussing the export ceiling for Blackwell chips to China and when the European Union is debating sovereignty criteria for AI infrastructure. Korea enters the scene as a partner that combines world-class memory (HBM4 and the next generation), industrial capital for gigawatt data centers, and a robotics industry that adopts the Isaac stack without regulatory disputes.
The two-year agreement with SK Hynix signals that the marginal price of memory for AI accelerators will remain pressured until 2028, directly affecting the TCO of any on-premises GPU platform. SK Telecom's gigawatt data center in 2027 offers a third supplier for workloads that currently run only on AWS or Azure's Asian operations, with competitive latency for Japan and Southeast Asia. Clients of Hyundai and LG on the assembly line gain a clearer roadmap for adopting physical AI, which impacts consultancies advising supply chains in automotive manufacturing from the UK to Brazil’s hubs of Camaçari and Betim.
What Wasn't Said
No deals were announced with stated values, and this is information in itself. Nvidia has already made public commitments of $100 billion with OpenAI, $6.3 billion with CoreWeave, and $5 billion with Intel, all detailed in press releases or filings. Keeping Korea’s figures undisclosed preserves the political sensitivity of the relationship with Washington and allows room to negotiate with Beijing without exposing a flank. The omission also suggests that at least one of the contracts has components contingent on regulatory approval still underway, both from the Seoul government and the U.S. Department of Commerce.
What remains unknown for investors is the crucial number: how much of this package is a firm revenue commitment versus a letter of intent. Until Nvidia’s next earnings call in August, the market will need to speculate.