Chey Tae-won, Jensen Huang announce sweeping AI partnership spanning chips to data centers

SK Group and Nvidia announced a sweeping partnership Monday covering AI semiconductor design and manufacturing through to the construction and operation of AI data centers.

Under the agreement, SK Hynix will co-develop and supply next-generation AI memory chips to Nvidia, which will in turn provide its highest-performance AI superchips to SK Telecom to support the construction of what the companies described as Asia's largest AI data center.

The alliance, previously centered on SK Hynix's memory chip supply, has now expanded to encompass the full upstream and downstream scope of the AI business — a significant upgrade to the partnership.

For Nvidia, the deal secures a stable, long-term supply of SK Hynix memory at a time when the global memory shortage shows no sign of easing, while simultaneously opening a new sales channel for its superchips through SK Telecom — further cementing what Nvidia envisions as its "AI universe."

Nvidia CEO Jensen Huang called the announcement a "big partnership," saying it covers "not only the semiconductor sector but also the telecommunications sector for building AI factories in Korea" and described it as "the first case spanning multiple years, multiple platforms and technologies, and multiple businesses across SK Group."

"We are witnessing enormous demand around the world for more AI factories, and business is booming," Huang said. "That is why we formed this partnership. We are just at the beginning of building AI infrastructure, and the future is incredibly bright."

SK Group Chairman Chey Tae-won said the two companies would build AI factories together, explaining that the concept encompasses both SK Hynix's fabs and AI data centers. "A key part of this is Nvidia and SK jointly building an R&D roadmap so we can adapt more quickly to future demand," he said.

He added that while past cooperation had been focused primarily on memory, "from now on we will elevate the partnership to the group level."

On the fourth day of his visit to South Korea, Huang joined Chey in the lobby of SK Seorin Building in Jongno-gu, Seoul, on Monday morning to announce the new AI partnership.

Also present were SK Hynix CEO Kwak Noh-jung, SK Telecom CEO Jeong Jae-heon, SK Telecom CTO Jeong Seok-geun, Nvidia Senior Vice President Jeff Fisher, and Madison Huang — Huang's eldest daughter and Nvidia's senior director of Omniverse and robotics marketing.

SK Hynix agreed to co-develop and supply memory chips on a long-term basis for four Nvidia products: the "Vera Rubin" AI supercomputer, the "Vera" central processing unit, the "RTX Spark" personal PC superchip, and the "Jetson Thor" robotics computing platform.

These are the next-generation superchips Huang had earlier referred to as "four gifts for Korea." The agreement positions SK Hynix to align its product roadmap with Nvidia's AI superchip lineup — spanning AI data centers, personal AI and physical AI — and to enter new markets in the process.

"Our partnership will expand from AI supercomputers to CPUs, PCs and the very exciting field of robotics," Huang said. "SK Hynix will continue to be Nvidia's largest memory partner."

"We are already purchasing billions of dollars of memory from SK Hynix every year, and that will only grow," he added. "In a sense, you can buy their stock at a cheap price right now."

Huang had also alluded to the memory supply crunch at an earlier "samgyeopsal meeting," saying, "We need more HBM. The next-generation Vera CPU and the RTX Spark personal PC superchip also require large volumes of LPDDR5."

He reiterated the point Monday, saying, "Every country and every company will need AI infrastructure. AI startups around the world are booming — and that means demand for Nvidia computing is high, and demand for SK Hynix memory is high."

SK Telecom and Nvidia are pushing ahead with plans to build a gigawatt-scale AI factory, targeting its first operation in 2027. The gigawatt-class AI factory is a next-generation data center built on Nvidia DSX-based infrastructure, going well beyond the conventional data center model limited to general-purpose computing and data storage.

SK Telecom plans to join Nvidia's "Nvidia Cloud Partner" global ecosystem for high-performance cloud AI services, aiming to secure the lowest token costs and highest performance per watt — and through that, to grow into Asia's leading AI cloud provider.

SK Telecom plans to deploy Nvidia's latest superchip, Grace Blackwell, and then adopt the next-generation Vera Rubin superchip, set for release in the second half of this year.

The two companies also agreed to deepen cooperation in physical AI and robotics. Technology SK Telecom has built on Nvidia's digital twin platform Omniverse is already being applied to SK Hynix's semiconductor manufacturing processes, and robot simulation capabilities are being further developed using Nvidia's robotics platform Cosmos.

"We are at the beginning of the AI revolution — that is why we are building AI factories," Huang said. "Korea has almost no AI infrastructure. We will start building it here together with SK Telecom."

Monday's announcement marked the third meeting between Huang and Chey during Huang's four-day visit to South Korea. The two had previously met over samgyeopsal, a Korean pork belly dish, on Friday — the first day of his trip — and again over fried chicken on Sunday evening, appearing together in public nearly every day and putting their companies' deepening AI partnership on prominent display.

SK, 엔비디아와 AI 팩토리 구축…그룹 차원으로 파트너십 격상

https://biz.heraldcorp.com/article/10765799