Nvidia and South Korea’s SK Group unveiled a business partnership exceeding $500 billion on July 24, 2026, in San Francisco. The initiative focuses on large-scale AI data centers and next-generation memory development to accelerate South Korea’s push toward becoming a global hub for AI innovation.
The scale of the agreement reflects a deep interdependence between Silicon Valley’s compute leadership and South Korea’s semiconductor industrial base. During a high-stakes gathering at the Midway venue in San Francisco, the partnership was framed not just as a corporate deal, but as a strategic alignment of infrastructure and talent.
The $500 Billion Infrastructure Push with SK Group
The centerpiece of the announcement is a more than $500 billion AI initiative. This collaboration targets two primary fronts: the construction of massive data centers and the securing of a long-term supply of next-generation memory.
A critical component of this plan involves SK Telecom, which intends to build a 2-gigawatt AI data center. According to Nvidia, this facility will be powered by Vera Rubin chips and SK Hynix’s HBM4 high-bandwidth memory, with the first site expected to be operational in 2027.
Beyond the hardware, Nvidia is entering a long-term partnership with SK Hynix to jointly develop high-bandwidth memory specifically for AI agents, physical AI applications, and AI training. This move secures the “fuel” for the AI revolution—HBM memory—which Nvidia CEO Jensen Huang noted was essential to the creation of the supercomputers powering today’s AI.
The San Francisco AI Declaration and National Ambition
The business deals occurred alongside a broader diplomatic effort. South Korean President Lee Jae Myung presented what he termed the San Francisco AI Declaration
, positioning South Korea as a primary engine of the AI revolution rather than a mere participant. President Lee compared the current emergence of artificial intelligence to humanity discovering fire anew
.
The President’s vision includes a multitrillion-dollar investment strategy targeting hyperscale data centers, AI semiconductors, and physical AI. This ambition is supported by a consensus among global tech leaders; OpenAI’s Sam Altman stated that the incredible AI revolution would not have been possible without Korea.
Expansion Into Automotive and Internet Portals
The cooperation extends beyond memory and data centers into consumer technology and software. Nvidia plans to jointly develop a self-driving Genesis sedan in partnership with Hyundai Motor Group.
Nvidia is also targeting the South Korean internet ecosystem. The company, alongside Brookfield, plans to expand the AI data center of Naver, the South Korean internet portal. Nvidia intends to make a large-scale investment in Naver to assist the company in expanding its presence both domestically and abroad.
Academic Integration and the “Korea AI” Project
To sustain this industrial growth, Nvidia is embedding itself in South Korea’s academic infrastructure. On July 23, Nvidia and the Korea Advanced Institute of Science and Technology (KAIST) announced a joint AI research lab at the KAIST Kim Jaechul Graduate School of AI in Seoul.
This lab—the first joint venture between a global tech company and a Korean university—will focus on advancing agentic AI. The project will utilize Nvidia Nemotron open models and AI Cloud partner computing. Furthermore, the Nvidia CEO stated a goal to build a Korea AI
, which involves developing a specific large language model for the country.
Strategic Stakes: The Full-Stack Dependency
This series of agreements signals a shift toward “full-stack” AI sovereignty. By integrating everything from the raw silicon (HBM4) and the chips (Vera Rubin) to the data centers and the localized language models, South Korea is attempting to insulate its AI future from supply chain volatility.
The urgency is evidenced by the appetite for hardware. SK Group Chairman Chey Tae-won noted that in every meeting with Jensen Huang, the request is simply for more chips
. This relentless demand creates a symbiotic loop: Nvidia requires Korean memory to build its supercomputers, and Korean conglomerates require Nvidia’s compute to power their national AI infrastructure.
While the financial commitments are massive, the success of the initiative depends on the 2027 timeline for the first 2-gigawatt data center. Whether the joint development of HBM4 and the “Korea AI” model can keep pace with the rapidly evolving demands of agentic AI remains the primary technical hurdle for this partnership.
Worth a look
