Nvidia and SK Group Bet on South Korea's AI Infrastructure
Nvidia and South Korea's SK Group announced an AI infrastructure partnership plan worth more than $500 billion, spanning a 2-gigawatt AI factory and…
On July 24, at an AI summit in San Francisco, Nvidia and South Korea's SK Group put a partnership spanning chips, memory, and data centers on the table: the two sides announced plans to advance an AI infrastructure collaboration valued at more than $500 billion. That figure represents the total scale of the partnership plan, and does not equal completed cash investment or binding orders.(investor.nvidia.com)
What Happened
Under the arrangement disclosed by Nvidia, SK Telecom plans to build a 2-gigawatt AI factory using Nvidia's Vera Rubin platform, with the first facility expected to come online in 2027. SK hynix will also establish a long-term partnership with Nvidia, focusing on supply and co-development of next-generation AI memory, including HBM.(investor.nvidia.com)
This is not a single GPU procurement agreement. It places several of the most common bottlenecks in AI computing on one roadmap: Nvidia provides the accelerated computing platform, SK Telecom handles network and data center operations, and SK hynix supplies high-bandwidth memory. For model companies and cloud providers, what is truly scarce is not just chip count, but also power, rack deployment capacity, memory supply, and the systems engineering needed to deliver continuously.
On the same day, Nvidia, NAVER, and Brookfield also announced plans to expand NAVER's AI factory in Sejong to 200 megawatts, more than tripling the previously disclosed 55-megawatt buildout. The project is positioned as sovereign AI infrastructure serving Korean companies, government agencies, and global AI cloud customers.(blogs.nvidia.com)
Why It Matters
Competition in AI infrastructure is shifting from "who can buy the most GPUs" to "who can reliably turn compute into usable services." A 2-gigawatt data center means the constraints will extend to grid access, land, cooling, capital, and construction timelines; the HBM partnership shows that memory is no longer just a standard component in a system, but a strategic resource that determines the speed and cost structure of AI system expansion.
For Nvidia, tying itself to SK Group strengthens system-level control from accelerators to memory to full-stack deployment. For South Korea, this is an industrial path that extends semiconductor manufacturing advantages into data centers, cloud services, and sovereign AI capabilities. For public markets, the impact is not limited to Nvidia and SK hynix, but also includes power equipment, data center operators, network infrastructure, and advanced packaging suppliers.
But the market should not interpret "more than $500 billion" as recognized revenue. The announcement is mainly based on letters of intent signed by both sides, and the project still faces execution conditions such as financing, permits, power supply, and delivery schedules.(investor.nvidia.com)
What to Watch Next
First, when will the 2-gigawatt project reach a final investment decision, and can the first facility come online on schedule in 2027? Second, will co-development of next-generation memory such as HBM4 translate into clear capacity allocations, pricing mechanisms, or long-term purchase commitments? Third, can Korea's sovereign AI factory secure real workloads from local model companies, government agencies, and international cloud customers, rather than remaining at the level of industrial-policy rhetoric?
More important to watch is that global AI capital expenditure is entering a phase of "infrastructure bundling": GPUs, memory, power, and data centers are no longer financed and built separately, but packaged together through strategic partnerships. If this model spreads, the competitive boundary of the AI industry will look more like long-term capital allocation across energy and semiconductor supply chains, rather than iteration among software models alone.
Sources
Information only. No investment, legal, tax, or financial advice.