South Korea’s $500B AI Bet: NVIDIA and SK Group Are Building the World’s Next Compute Hub
The SK Group NVIDIA AI partnership announced on July 25 is not a routine supply deal. It is a structural bet that South Korea can move from being the world’s best chip manufacturer to becoming one of its most important compute exporters — and that NVIDIA wants a guaranteed stake in both sides of that transition.
The two companies signed letters of intent for a $500 billion-plus initiative that spans two distinct but deeply connected areas: AI factory construction and next-generation memory co-development. Together, the agreements touch SK Telecom, SK hynix, and NVIDIA in ways that will reshape Korea’s AI infrastructure posture for at least the next decade.
A 2-Gigawatt AI Factory Powered by Vera Rubin and SK hynix HBM4
The headline infrastructure commitment comes from SK Telecom, which will build a 2-gigawatt-scale AI cloud in Korea. This facility will run on the NVIDIA DSX full-stack AI factory platform, deploying Vera Rubin accelerated computing — NVIDIA’s next-generation GPU architecture — powered by SK hynix HBM4 memory. The first AI factory is scheduled to come online in 2027.
Two gigawatts is a significant number. For context, most large hyperscale data centers currently operate in the range of 100 to 500 megawatts. A 2-gigawatt deployment represents the kind of sovereign-scale compute infrastructure that governments have begun treating as a strategic asset, comparable in importance to national energy grids.
The NVIDIA DSX platform integrates accelerated computing, systems software, and partner technologies to deliver what NVIDIA describes as the lowest token cost at maximum energy efficiency — a critical metric as inference costs increasingly define the economics of AI deployment.
SK hynix and NVIDIA: From Supplier to Long-Term Co-Developer
The memory component of this deal carries arguably more long-term significance than the factory announcement. SK hynix is entering a formal long-term AI memory partnership with NVIDIA, upgrading what was already a major supplier relationship into a co-development framework.
Under the agreement, the two companies will jointly develop and optimize next-generation AI memory solutions, including HBM, targeting workloads that span large language model training, agentic AI, and physical AI. NVIDIA gains a more secure supply chain for its highest-demand components. SK hynix gains a locked-in strategic customer and a seat at the table for roadmap decisions.
This matters because HBM has become one of the genuine bottlenecks in AI infrastructure scaling. NVIDIA certified SK hynix, Samsung, and Micron to supply HBM4 for Vera Rubin, but a formal co-development agreement with SK hynix goes further. It signals that memory architecture decisions will increasingly be made jointly rather than handed downstream as a specification.
The agreement also extends across NVIDIA’s broader product portfolio — including Vera CPUs, RTX Spark PCs, and Jetson Thor robotics platforms — suggesting the partnership covers physical AI and edge compute as well as data center workloads.
What Korea Gains — and Why NVIDIA Needs This
SK Group Chairman Chey Tae-won framed the partnership in explicitly sovereign terms. The goal is not simply to build AI services for Korean consumers, but to position Korea as a global hub that exports AI intelligence, not just the chips that power it.
That framing aligns with a broader pattern NVIDIA is executing across Asia. Earlier this month, Japan announced a national physical-AI system backed by 27,500 Rubin GPUs. The SK Group deal extends NVIDIA’s Asia-Pacific infrastructure network significantly, adding Korea’s world-class semiconductor manufacturing base and SK Telecom’s dense domestic network infrastructure to the mix.
For NVIDIA, the partnership solves two problems simultaneously. First, it secures stable supply of the memory components that most constrain Vera Rubin GPU throughput. Second, it creates a major committed customer for AI factory capacity at a time when compute demand continues to outpace available infrastructure. Jensen Huang noted that South Korea has world-class networks, data centers, chip technology, and industrial scale — the full stack for turning NVIDIA’s hardware into a national AI output.
Furthermore, the partnership accelerates what NVIDIA calls sovereign AI services — cloud infrastructure owned and operated by a single country, rather than routed through US hyperscalers. That is an increasingly important distinction for governments evaluating data residency, security, and strategic technology control.
The Broader AI Infrastructure Race
This announcement did not arrive in isolation. NVIDIA also revealed a partnership with NAVER and Brookfield on the same day to expand Korea’s national AI factory buildout further. The coordination suggests NVIDIA is deliberately assembling a comprehensive Korean AI ecosystem rather than making isolated infrastructure bets.
The timing also connects directly to global capital market dynamics. South Korea’s chip stocks suffered significant turbulence earlier this month, with SK hynix caught in a broader sell-off driven by AI rotation trades. A $500 billion-plus commitment from NVIDIA, formalizing multi-year demand visibility for SK hynix’s HBM output, provides a structural counterweight to the short-term volatility that has unsettled chip investors throughout 2026.
The letters of intent signed today are not a binding purchase order. The commercial terms underlying that $500 billion-plus headline remain subject to finalization. However, the public commitment from both Jensen Huang and Chairman Chey Tae-won, combined with the specificity of the DSX platform selection and the 2027 factory timeline, signals that both parties have moved well past exploratory discussions.
South Korea’s path from chip exporter to AI infrastructure power is now tied directly to whether NVIDIA’s Vera Rubin generation delivers on its efficiency promises — and whether SK hynix can scale HBM4 production fast enough to feed it.
Frequently Asked Questions
What is the SK Group and NVIDIA $500 billion partnership about?
SK Group and NVIDIA signed letters of intent for a $500 billion-plus initiative covering two areas: SK Telecom will build a 2-gigawatt AI factory in South Korea running NVIDIA’s DSX platform and Vera Rubin GPUs, and SK hynix will enter a long-term co-development agreement with NVIDIA for next-generation HBM memory.
When will the first SK Telecom AI factory come online?
The first AI factory is planned to come online in 2027. It will use the NVIDIA DSX full-stack platform and be powered by SK hynix HBM4 memory paired with NVIDIA Vera Rubin accelerated computing.
What does this mean for SK hynix’s HBM memory business?
The agreement upgrades SK hynix from a certified HBM4 supplier to a formal co-development partner for NVIDIA’s AI memory roadmap. This gives SK hynix demand visibility across multiple product generations and a direct role in shaping memory architecture for AI infrastructure, from LLM training to agentic and physical AI workloads.
Is the $500 billion figure a confirmed spend commitment?
The $500 billion-plus figure refers to the scope of the initiative as outlined in the letters of intent signed by both parties. Letters of intent are formal expressions of agreement that precede binding commercial contracts, so the final committed spend will depend on subsequent agreements.
