Nvidia and SK Group expand AI ties into co-development

Nvidia and SK Group expanded their partnership from HBM supply into AI infrastructure and chip co-development.

Nvidia CEO Jensen Huang and SK Group Chairman Chey Tae-won used a June 8 press conference in Seoul to say their companies are moving well beyond a buyer-supplier relationship. The new plan reaches into AI memory, cloud infrastructure, semiconductor design, and factory automation.

From memory supply to full-stack AI work

The biggest signal in the announcement is the shift in scope. For years, SK hynix has been tied to Nvidia mainly through high-bandwidth memory, the kind of component that keeps modern AI accelerators fed with data. Now the two sides are talking about joint roadmaps for next-generation AI memory and infrastructure.

Huang was blunt about the scale of the relationship. He said SK hynix will remain Nvidia’s “largest memory partner,” and he framed the next phase as a multi-platform, multi-technology effort. That language matters because it suggests closer technical planning, not just purchase orders and supply forecasts.

He also made clear that Nvidia’s own product roadmap will keep pulling SK hynix deeper into the stack. Huang said SK hynix memory products will be used across future Nvidia platforms, including Vera Rubin AI supercomputers, Vera CPUs, RTX Spark, and Jetson Thor.

• SK hynix already sits inside Nvidia’s memory supply chain
• The new scope includes AI memory design and infrastructure planning
• The partnership touches data center, workstation, and edge AI products

Korea wants more than chip exports

Huang spent part of the event arguing that Korea needs more AI infrastructure if it wants to keep up with its semiconductor strength. His point was simple: memory and chip manufacturing are not enough if the country wants to host the systems that train and run large AI models.

“Korea already possesses world-class semiconductor manufacturing capabilities and an advanced AI ecosystem, but AI infrastructure is still insufficient,” Huang said. “Just as the semiconductor industry needed fabs, AI also needs factories.” That line neatly captures Nvidia’s pitch: AI compute clusters are becoming industrial infrastructure, not just IT gear.

“Korea already possesses world-class semiconductor manufacturing capabilities and an advanced AI ecosystem, but AI infrastructure is still insufficient,” Huang said. “Just as the semiconductor industry needed fabs, AI also needs factories.”

The quote also explains why Nvidia keeps pushing into cloud partnerships and local deployments. If AI training capacity is scarce, the company can sell more than chips. It can sell the blueprint for the entire stack, from silicon to software to data center architecture.

SK Telecom is part of that plan. The telecom company and Nvidia said they will build an AI cloud platform based on Nvidia’s DSX architecture, with commercial operation targeted for 2027. The platform is expected to scale to gigawatt-class infrastructure and later expand into broader Asian markets.

The cloud plan is as important as the memory deal

The cloud announcement may end up mattering as much as the memory supply story. SK Telecom will act as an Nvidia Cloud Partner, which means it will use Nvidia’s latest infrastructure and software stack to offer training and inference services. That puts SK Telecom in a position to sell AI compute, not just connectivity.

The hardware path is also clear. The first phase will use Nvidia Blackwell GPUs, then move to the company’s next-generation Vera Rubin platform, which Nvidia says is scheduled for release later this year. That sequencing gives the partnership a near-term anchor and a longer-term upgrade path.

• AI cloud launch target: 2027
• Initial GPU platform: Blackwell
• Next platform: Vera Rubin, scheduled for later this year
• Infrastructure target: gigawatt-class scale

For Korea, this is a strategic bet. If the buildout works, SK Telecom could become a regional AI cloud provider with enough scale to serve enterprises that do not want to buy their own clusters. For Nvidia, it creates another outlet for its chips, software, and platform strategy in Asia.

Semiconductor design and factory automation are next

The partnership does not stop at cloud infrastructure. SK hynix said it plans to use Nvidia CUDA-X and PhysicsNeMo to improve semiconductor process simulation, and the companies are also considering broader work in electronic design automation, or EDA.

That matters because chip design is one of the most expensive and time-consuming parts of the semiconductor business. Better simulation can shorten iteration cycles, reduce waste, and help engineers test more ideas before a wafer ever reaches a fab. If Nvidia’s software stack becomes part of that workflow, the company gets deeper into the engineering process itself.

SK hynix is also pushing AI into manufacturing operations. The company is using Nvidia Omniverse and OpenUSD to build digital twins of semiconductor fabs, while also working on autonomous mobile robots and AI-based production optimization. That combination points toward a more automated factory model, where software watches equipment, predicts bottlenecks, and adjusts workflows in real time.

Chey Tae-won described the goal in direct terms: the companies want to jointly develop next-generation memory for AI factories while applying AI to semiconductor design and manufacturing. He said, “Based on full-stack AI infrastructure competitiveness, we aim to grow into a leading AI cloud operator representing Asia.”

That is a bold ambition, but the pieces are at least visible now. SK Group has memory, telecom, and manufacturing assets. Nvidia has the chips, software, and system design know-how. Put together, they can build something bigger than a supply contract.

What this means for the AI supply chain

This deal also says something broader about where AI infrastructure is heading. The bottleneck is no longer just model quality. It is memory bandwidth, power delivery, packaging, cloud capacity, and factory-level automation. Nvidia and SK Group are trying to own more of that stack together.

That makes the partnership more durable than a normal procurement relationship. It also raises the stakes for rivals in memory, cloud, and semiconductor tooling, because the competition is no longer happening in one product category at a time.

For readers watching the chip business, the practical takeaway is simple: SK hynix is moving from being one of Nvidia’s key suppliers to being a partner in how the next generation of AI systems gets built, deployed, and manufactured. If the 2027 cloud target holds and the Vera Rubin transition lands on schedule, this alliance could become one of the clearest examples of how AI infrastructure is now a joint engineering project.