On July 24, 2026, NVIDIA and South Korea's SK Group jointly announced an AI infrastructure partnership with a total value exceeding $500 billion. The plan covers large‑scale AI data center construction and joint development and long‑term supply of next‑generation high‑bandwidth memory (HBM) to address the explosive global demand for AI compute. It is one of the largest AI infrastructure collaborations in Asia to date and further strengthens South Korea's central role in the global AI supply chain.
Project core details
1. Long‑term memory supply and joint development
- NVIDIA and SK Hynix signed a long‑term cooperation agreement to ensure stable supply of advanced memory chips.
- The two will jointly develop next‑generation high‑bandwidth memory (HBM4 and subsequent products) targeting AI training, autonomous agents, and Physical AI applications.
- SK Hynix is already NVIDIA's largest memory supplier; prior agreements were expected to run for more than two years and can be extended. The company also plans to double wafer capacity by 2030 to align with growing demand from AI servers, robotics, and supercomputers.
2. Construction of 2‑gigawatt AI data centers
- SK Telecom will build an AI data center with a capacity of up to 2 gigawatts, using NVIDIA's Vera Rubin processors and SK Hynix HBM4 memory.
- The first facility is planned to begin operations in 2027.
- The project expands on SK Telecom's earlier "gigawatt‑scale AI cloud" commitment, with funding potentially coming from global tech firms, overseas capital, and other strategic partners.
3. Collaboration with Naver and Brookfield
- NVIDIA, Naver and Brookfield Asset Management will jointly expand Naver's AI data center operations in South Korea.
- Brookfield previously launched a $100 billion AI infrastructure initiative and established a dedicated fund, with NVIDIA as a key partner.
The overall plan tightly binds Korea's memory manufacturing strengths with NVIDIA's GPU technology, creating a complete "chip + memory + data center" closed loop.
Strategic background: addressing AI compute and supply‑chain bottlenecks
High‑bandwidth memory has become a key bottleneck in the AI supply chain. Advanced processors require extremely high data transfer rates, while HBM capacity has long been constrained. Deeper cooperation between NVIDIA and SK Hynix can better align memory production with chip development cycles and reduce supply risks.
Meanwhile, global AI data center construction is accelerating, making power, cooling and land resources new challenges. Facilities at the 2‑gigawatt scale rank among the largest in terms of power and footprint, reflecting both parties' commitment to realizing the "AI factory" concept.
Potential impacts: multi‑dimensional industry effects
1. Impact on NVIDIA and SK Group
- Supply‑chain security: Long‑term agreements secure future memory supplies for NVIDIA's GPUs, reinforcing its leadership in AI hardware.
- Strengthening Korea's AI ecosystem: SK Group, leveraging NVIDIA's technology, can accelerate its "K‑AI" vision across chips, advanced models and AI factories.
- Investment returns: A partnership of this scale—over $500 billion—could drive business growth and potentially support stock performance for both parties.
2. Impact on global AI infrastructure
- Rise of an Asian compute hub: South Korea could become a key AI computing hub in the Asia‑Pacific region, attracting more international capital and projects.
- Supply‑chain reshaping: Expanded HBM capacity would ease global memory shortages and reduce AI training costs.
- Heightened competition: Other regions (such as the US, Middle East and Europe) may accelerate similar large‑scale collaborations, driving a global AI infrastructure race.
3. Impact on industry and users
- Enterprises and developers: Increased compute and memory availability will accelerate large model training and inference deployment, lowering adoption barriers.
- Physical AI and robotics: Dedicated memory development for Physical AI will advance applications such as autonomous driving and humanoid robots.
- Energy and sustainability: A 2‑GW data center entails massive power demand, likely prompting greater investment in clean energy and efficient cooling technologies in South Korea.
4. Potential risks and challenges
- Execution difficulty: Projects of this scale involve complex coordination of funding, land, power and policy, and have long lead times.
- Geopolitical risks: Deeper US‑Korea collaboration may raise concerns among other countries about concentrated supply chains.
- Market volatility: AI investment overheating remains a risk; demand must match capacity expansion.
Industry takeaway: hardware‑software coordination and long‑termism
The NVIDIA–SK Group partnership underscores the importance of tightly linking "chips + memory + infrastructure" in the AI era. GPU advantages alone are insufficient in the global compute race; long‑term strategic alliances with key supply‑chain partners are essential.
The over $500 billion plan is not only a merger of technology and capital but also a redefinition of AI infrastructure for the next decade. With the first facilities slated for 2027, the global AI compute landscape may shift significantly.
▶️ Like this article? Creating content is hard!
Follow our YouTube channel:
AI Rosetta
Your likes, subscriptions and shares are our greatest support! 🙏
For more AI news visit: https://www.airosetta.com