Summary of Key Points
NVIDIA and the South Korean SK Group have reached a cooperation agreement worth a total of $500 billion (not a one-time payment, but the overall scale of their long-term partnership). The main objectives of this collaboration are twofold: first, to jointly build AI factories and cloud computing centers; second, to develop the next generation of specialized AI memory (such as HBM). Both parties complement each other's strengths. NVIDIA needs a stable supply of memory to produce its AI chips, while SK Group aims to upgrade its AI infrastructure using NVIDIA’s technology, with the goal of meeting the global AI computing demands, particularly in the Asia-Pacific region. This represents an expansion of their previous collaboration.
1. What exactly does the $50 billion cooperation entail? The focus is on “AI infrastructure and memory”
The cooperation can be broken down into two key components:
- AI factories and cloud computing centers: SK Telecom, a subsidiary of SK Group, will build an AI cloud computing center with a capacity of 2 gigawatts (equivalent to the power of 2 million household air conditioners), which will support the operation of extremely large-scale AI models. This center will utilize NVIDIA’s Vera Rubin system and be equipped with SK Hynix’ next-generation HBM memory, with the first facility expected to be operational by 2027. It will provide AI services to companies, such as training AI models and offering cloud computing solutions.
- Next-generation AI memory development: SK Hynix and NVIDIA will collaborate on developing HBM (high-bandwidth memory, essentially the “superfast storage” for AI chips, which is several times faster than regular memory and becomes increasingly essential as AI models become more complex). This ensures that NVIDIA won’t face shortages of this critical component in its future chip production, given the current global demand for HBM and SK Hynix’s position as a leading supplier.
2. What are the motivations for both parties? The goal is to stay ahead in the “AI era”
- NVIDIA’s needs: AI chips like the H100 rely heavily on HBM, with each H100 chip requiring eight HBM modules. An unstable memory supply would hinder NVIDIA’s ability to mass-produce chips and potentially cause it to miss out on market opportunities. By partnering with SK Hynix, NVIDIA gains control of a vital supply chain component and the opportunity to jointly develop more advanced memory, maintaining its leadership in the AI chip industry.
- SK Group’s needs: As a major South Korean company with both memory (SK Hynix) and telecommunications (SK Telecom) businesses, SK Telecom aims to transition from traditional telecom services to AI-based solutions. Building a cloud computing center is a key step in this transformation. With NVIDIA’s technical support, it can quickly establish an AI infrastructure that competes with services provided by Amazon Web Services and Google Cloud. SK Hynix, on its part, will be able to better align its memory products with AI requirements, solidifying its dominant position in the global HBM market (currently holding over 60% of the market share).
3. This is not a sudden cooperation; it’s an upgrade of previous efforts
The $50 billion agreement builds on earlier collaborations:
- Last year’s APEC summit: NVIDIA reached an agreement with the South Korean government and companies to deploy 260,000 GPUs (AI chips), including a plan by SK Group to build an AI factory that will produce 50,000 of these chips.
- June this year: During Sung-Yun Park’s visit to South Korea, NVIDIA and SK Hynix signed a technical cooperation agreement to develop memory for use in AI supercomputers and robotics platforms. The current $50 billion deal consolidates these individual efforts into a comprehensive framework, moving from simple product purchases to joint research and development as well as infrastructure construction.
4. What impact will this have on the industry?
- Intensified AI competition: The combination of NVIDIA (AI chips), SK Hynix (AI memory), and SK Telecom (AI infrastructure) creates significant competitive pressure for other players. For example, AMD may need to find new memory partners to avoid falling behind. Other cloud service providers (such as Alibaba Cloud and Tencent Cloud) will also need to accelerate their investment in AI infrastructure.
- Accelerated development in the Asia-Pacific region: Both parties are committed to meeting the regional AI demands, with South Korea potentially becoming a hub for AI infrastructure. Companies in China and Southeast Asia may turn to SK Telecom’s cloud computing services for AI solutions, which could drive the adoption of AI applications in various industries (such as industrial and medical AI).
- Growing demand for HBM: As AI models become larger, the demand for HBM will continue to rise. SK Hynix is expected to secure more orders, potentially maintaining high prices for its products, while other memory manufacturers (like Samsung) may increase their HBM production efforts.
5. How will this impact ordinary people?
Although this cooperation may seem distant from everyday life, it will ultimately have a tangible impact on our lives:
- More widespread AI services: AI-generated content (such as articles and videos) will become faster and more affordable due to increased computing power.
- Enhanced efficiency in business: Companies will use AI to optimize processes (e.g., in manufacturing and logistics), which could lead to lower product costs.
- Improved AI capabilities in our devices: The AI algorithms used in smartphones (e.g., better photo editing and voice assistants) will become more advanced as the underlying models improve.
In summary, this collaboration is a prime example of the integration of “hardware, software, and infrastructure” in the AI era, marking a shift towards full-scale competition across the entire AI industry chain. The partnership between NVIDIA and SK Group will accelerate the adoption of AI technology, bringing benefits to consumers more quickly.