Nvidia and South Korea’s SK Group have put a huge number on the board: more than $500 billion for a broad AI infrastructure push that spans data centers and next-generation memory. The plan includes a long-term collaboration with SK Hynix on high-bandwidth memory, a 2-gigawatt AI data center led by SK Telecom, and a separate effort with Naver and Brookfield to expand Naver’s AI data center in South Korea.
This is not just another chip announcement. It is a supply-chain strategy, a power-infrastructure play, and a regional AI land grab all rolled into one. The deal was announced during President Lee Jae Myung’s AI summit visit to San Francisco, which also featured a wider wave of Korea-U.S. AI and semiconductor deals.
Key Insights
- SK Telecom plans to build a 2-gigawatt AI data center using Nvidia Vera Rubin chips and SK Hynix HBM4 memory, with the first facility expected in 2027.
- Nvidia and SK Hynix will work on long-term supply and co-development of next-generation HBM for AI training, AI agents, and physical AI workloads.
- Nvidia, Naver, and Brookfield are also moving to expand Naver’s AI data center in South Korea.
- Nvidia’s own newsroom says the parties signed letters of intent, which means the package is serious but still subject to execution steps and closing details.
- The broader Korea AI wave is accelerating fast: Reuters also reported a separate $950 billion set of South Korean AI-related deals announced around the same summit.
The big signal here is scale: this is about securing compute, memory, land, and power before the next AI demand wave hits.
The Big Picture
The headline number grabs attention, but the real story is strategic control. Nvidia wants more than chip sales. SK Group wants to anchor Korea’s AI buildout around its own memory business, cloud footprint, and telecom infrastructure. Naver wants to turn its AI ambitions into something bigger than a software layer by tying itself to physical infrastructure and global capital.
That is why the memory piece matters so much. HBM has become one of the most valuable bottlenecks in modern AI systems, and SK Hynix sits at the center of that market in this deal. Nvidia’s emphasis on next-generation HBM shows that the AI race is now as much about memory bandwidth and supply resilience as it is about model quality.
How We Got Here
The July announcement did not appear out of nowhere. Reuters reported in June 2026 that Naver was already preparing to build gigawatt-scale AI factories using Nvidia technology. This latest deal turns that earlier ambition into a more explicit infrastructure and financing program.
Nvidia also had already been working with SK Group on an AI factory effort before this week’s expansion, including plans tied to semiconductor research, development, production, and cloud infrastructure for AI agents and digital twins. The new package broadens that relationship from a partnership into a much larger regional platform strategy.
The Critical Shift
The most important shift is that AI competition is moving from software bragging rights to physical capacity. If a company cannot secure enough power, memory, and data center space, its AI roadmap can stall no matter how good its models are. That is exactly why a 2-gigawatt buildout matters: it signals industrial-scale ambition, not a pilot project.
It also shows how national AI policy and private-sector infrastructure are now overlapping. South Korea is positioning itself as a serious AI manufacturing and deployment hub, while Nvidia is using its ecosystem power to shape where the next generation of compute gets built. Brookfield’s role adds another layer: the project is not just about technology, but about financing hard infrastructure at a scale few companies can manage alone.
What Lies Ahead
The near-term test is execution. Letters of intent and headline figures are impressive, but the real work will be in financing closure, power delivery, construction timelines, and chip supply. Those are the pressure points that decide whether a megadeal becomes real infrastructure or just a press release with a big number attached.
For the broader tech market, the message is clear: the next AI winners will not only be the best model makers. They will be the companies that can lock in memory, energy, land, and capital fast enough to keep up with demand. Korea is now one of the most important test cases for that new reality.
0 Comments