Nvidia and SK Group Form Historic $500 Billion-Plus AI Alliance and Lock In HBM4 for Vera Rubin Platform
On July 24–25, 2026, NVIDIA Corporation (NVDA) and South Korean conglomerate SK Group announced a historic, multi-year comprehensive strategic partnership valued at over $500 billion. Formalized through letters of intent, the alliance represents the largest single AI infrastructure commitment to date. It directly ties Nvidia's next-generation Vera Rubin accelerated computing architecture to SK hynix's high-bandwidth memory (HBM) production, securing Nvidia's access to the critical memory substrate1 needed to bypass persistent hardware bottlenecks.
This deal introduces a massive sovereign-scale actor to the frontier AI infrastructure race and solidifies South Korea's position as a primary global AI manufacturing and computing hub.
The 2-Gigawatt Vera Rubin DSX AI Factory
A core pillar of the alliance is the construction of a massive 2-gigawatt (GW) AI cloud and data center infrastructure in South Korea by SK Telecom (SK Group's telecommunications arm).
- Technology Stack: The facility will deploy Nvidia's full-stack DSX™ platform and next-generation Nvidia Vera Rubin superchips, which are expected to begin shipping in 2026–2027.
- Memory Substrate: The superclusters will be powered by next-generation SK hynix HBM4 memory.
- Timeline: The first operational facility of the 2GW campus is scheduled to commence commercial operations in 2027.
Securing the HBM Supply Chain
High-bandwidth memory (HBM) has been the single most visible supply bottleneck of the current AI infrastructure cycle. SK hynix, which controls approximately 60% of the global HBM market and recently raised $28 billion in a record Nasdaq listing in June 2026, has entered a long-term supply and joint optimization agreement with Nvidia.
This long-term contract secures Nvidia's HBM4 supply for its Vera Rubin architecture and establishes a co-development roadmap for next-generation memory23, spanning applications from large language model (LLM) training to agentic and physical AI.
"In the AI era, competitiveness depends not just on how effectively AI is utilised, but on how much intelligence we can produce. By leveraging SK hynix's AI memory and SK Telecom's AI infrastructure capabilities, SK will collaborate with Nvidia to build a world-class AI factory..." — Chey Tae-won, Chairman of SK Group
The Naver-Nvidia-Brookfield Expansion
In tandem with the SK Group partnership, Nvidia, South Korean internet giant Naver, and Brookfield Asset Management announced a massive proposed expansion of South Korea's national AI factory infrastructure:
- The group plans to expand Naver's Nvidia DSX AI factory deployment to 200 megawatts (MW), more than tripling the previously planned 55MW project. Naver's long-term goal is to scale its total Nvidia AI infrastructure footprint in the country to 1 gigawatt (1GW).
- Nvidia will invest $1 billion in Naver to support this expansion.
- Brookfield has signed a non-binding term sheet to provide up to $9 billion in financing for the infrastructure project, with Naver covering the remaining capital requirements.
-
An instance of Astronomical fab fabrication costs force technology buyers to directly underwrite future chip capacity. — To guarantee access to the scarce physical components of advanced chips, the leading buyer must directly underwrite suppliers through historic, multi-billion-dollar alliances. ↩︎
-
An instance of Memory manufacturers have successfully broken the industry's historic boom-and-bust cycle. — SK hynix has locked in Nvidia with a massive multi-year supply agreement for next-generation memory, turning volatile DRAM capacity into a predictable, contracted pipeline. ↩︎
-
An instance of Generational compute capacity is fully monetized and locked in before the silicon even exists. — Nvidia is securing HBM4 memory allocations years before shipment to lock in the supply chains required for its next-generation hardware architectures. ↩︎