FTFuture Technology
AI

SK Group and NVIDIA Announce a 500 Billion Dollar AI Infrastructure Partnership

· 3 min read

Key takeaways

  • SK Group and NVIDIA announced a partnership valued at over 500 billion dollars
  • The deal covers AI factories and next-generation memory, tying SK Hynix's HBM roadmap to NVIDIA's GPU pipeline
  • SK Hynix is one of the world's leading suppliers of high-bandwidth memory, a critical constraint for AI workloads
  • The partnership positions both companies to offer a vertically integrated AI infrastructure stack across Asia-Pacific

Half a trillion dollars. That is the number SK Group and NVIDIA have put on the table in what is shaping up to be one of the largest technology partnerships ever announced. The two companies confirmed plans to build out AI infrastructure at a scale designed to meet what they describe as the surging global demand for AI compute, covering everything from AI factories to next-generation memory chips.

For context, SK Group is not just one company. It is a South Korean conglomerate that spans SK Hynix, one of the world's leading memory chip manufacturers, SK Telecom, and SK E&S, among dozens of other subsidiaries. That breadth matters here. A partnership with NVIDIA that spans AI factories and next-generation memory is not a single procurement deal. It is a vertical integration play that could knit together the hardware stack from memory all the way up to the GPU systems that run large AI workloads.

Why Memory Is the Bottleneck Everyone Is Talking About

If you have been paying attention to AI infrastructure over the past two years, you will know that GPUs get most of the headlines, but memory is quietly becoming the real constraint on AI performance. High-bandwidth memory (HBM), which SK Hynix currently dominates, is the type of memory stacked directly onto AI chips. Demand has consistently outpaced supply, and every major AI lab has felt that squeeze.

By tying SK Hynix's next-generation memory roadmap directly to NVIDIA's GPU development pipeline, this partnership could give both companies a meaningful advantage over competitors who have to negotiate memory supply on the open market. It also raises interesting questions about what this means for other GPU makers who rely on SK Hynix memory, given that the Korean firm supplies a significant share of the HBM market.

NVIDIA, for its part, is clearly focused on ensuring that its Blackwell and Vera Rubin architecture chips have a guaranteed and optimised memory supply chain. The Vera Rubin NVL72 racks, which are currently ramping at partners like CoreWeave and Google, are power-hungry systems that demand enormous memory bandwidth. Locking in a deep partnership with the world's leading HBM supplier is a logical move.

The future, in 3 minutes a day. The biggest tech story explained every morning, free. Get the briefing →

AI Factories: The New Data Centres

The other major pillar of this partnership is AI factories, a term NVIDIA has been pushing hard as a rebranding of what most of us would call data centres. The framing is deliberate. A data centre stores and retrieves information. An AI factory, in NVIDIA's telling, actively produces intelligence, running inference and training workloads continuously.

SK Group has the capital, the energy infrastructure, and the land to build at scale, particularly across South Korea and potentially broader Asia-Pacific regions. Pairing that with NVIDIA's hardware and software stack creates a compelling offering for enterprises and governments in the region that want sovereign AI compute capacity without having to negotiate directly with American hyperscalers.

South Korea has been aggressive about positioning itself as an AI infrastructure hub. This partnership, combined with similar moves from Japan (more on that separately), suggests that Asia is not content to simply consume AI services built on Western infrastructure. The ambition is to own meaningful portions of the supply chain and compute layer.

What to Watch

A deal of this size and ambition will play out over years, not months. The 500 billion dollar figure is a commitment envelope, not a cheque being written today. The phasing, the specific projects, and the governance of the partnership will determine whether this delivers on its promise or becomes another grand announcement that quietly scales back over time.

That said, the strategic logic is tight. SK Group has the memory. NVIDIA has the GPUs and the software ecosystem. Together they can offer a vertically integrated AI infrastructure stack that neither could credibly offer alone. For anyone watching the global AI infrastructure race, this partnership is one of the clearest signals yet that the competition is not just between companies. It is between industrial alliances.

More from Future Technology