What Happened

Nvidia announced two major deals in South Korea on Friday, coinciding with President Lee Jae Myung's visit to Silicon Valley.

Naver Investment: Nvidia will invest $1 billion in Naver, South Korea's largest internet company, to finance an AI data center currently under construction. The investment will expand the facility from 55 megawatts to 200 megawatts — a nearly 4x increase. US private equity firm Brookfield has signed a nonbinding term sheet to fund up to $9 billion for the broader project. The site will run on Nvidia's AI computing hardware.

SK Group Partnership: Separately, Nvidia announced a long-term commercial partnership with SK Group valued at more than $500 billion. The figure includes Nvidia's purchases of high-bandwidth memory (HBM) chips from SK Hynix — the world's largest HBM supplier — as well as SK Group's purchases of Nvidia supercomputers. SK Telecom will build more than 2 gigawatts of AI data centers on the Korean Peninsula, with the first facility scheduled to open next year.

HBM Design Collaboration: Nvidia will also help SK Hynix design future high-bandwidth memory chips, a move aimed at guaranteeing access to the component that remains in critical shortage due to the global AI data center buildout. Only SK Hynix and Samsung can produce the advanced HBM variants that Nvidia's processors require.

Jensen Huang told Bloomberg Television that South Korea is in a "golden age," citing its semiconductor and industrial businesses.

Why It Matters

This is Nvidia's sovereign AI strategy executed at unprecedented scale. The company has now committed more than $40 billion to AI equity positions in the past year — including $30 billion in OpenAI — and the pattern is consistent: Nvidia invests capital in companies that subsequently become customers for its hardware.

Critics have labeled this "circular financing," and the structure is now being replicated with sovereign partners, not just neoclouds and data center operators. The SK Group deal is the clearest example yet: Nvidia buys HBM from SK Hynix, SK Group buys Nvidia supercomputers, and Nvidia invests in Naver's data center that runs on Nvidia chips.

The strategically critical element is the HBM design partnership. SK Hynix just completed a $26.5 billion US IPO in July 2026 — the second-largest US share sale on record — and is already the dominant HBM supplier. A design collaboration with Nvidia gives SK Hynix a structural advantage over Samsung, its only competitor in advanced HBM production. Samsung, meanwhile, announced its own $647 billion domestic chip investment in June, signaling an escalating Korean semiconductor arms race.

For the broader market, this means HBM supply constraints — the tightest bottleneck in AI compute — will persist through at least 2027. Nvidia is addressing this not through procurement alone, but by vertically integrating its supply chain through equity and co-design.

Who Is Affected

GPU cloud providers and neoclouds should plan for continued allocation-constrained access to next-generation Nvidia accelerators. The HBM bottleneck isn't resolving soon, and Nvidia's design partnership with SK Hynix may actually tighten supply for non-Nvidia customers.

Enterprise IT leaders evaluating Asian data center capacity gain a new option: Naver's 200MW facility, plus SK Telecom's broader 2GW pipeline. These could provide regional alternatives to US-based hyperscalers for Korea-adjacent workloads.

AI hardware competitors (AMD, Intel, custom silicon teams) face a deepening moat. Nvidia isn't just locking in HBM supply — it's co-designing the next generation of memory chips with the world's leading producer.

Strategic Implications

For AI startup founders: HBM-driven supply tightness on next-gen Nvidia accelerators will persist through 2027. Plan hardware roadmaps accordingly. The Naver facility expansion could become a viable sovereign AI cloud option in Asia — worth evaluating when it comes online, particularly if you have Korean or Japanese customers with latency requirements.

For developers/operators building with AI APIs: Short-term impact is minimal — this deal doesn't change API pricing or model availability. But the 200MW Naver facility and SK Telecom's 2GW pipeline mean new Asian compute capacity entering the market starting next year, which could improve regional latency and capacity for Korea-adjacent inference workloads.

For non-technical business owners evaluating AI tools: Nvidia's $500B SK Group partnership signals sustained infrastructure investment, meaning compute supply will grow substantially over the next 2-3 years. Expect gradual cost stabilization, but not dramatic price reductions in the next 12 months — the buildout timeline means capacity arrives incrementally.

What to Watch Next

Monitor whether Samsung responds with its own Nvidia partnership or pivots toward AMD or custom silicon partnerships. Also watch for Brookfield's term sheet converting from nonbinding to binding — the $9 billion financing commitment is not yet locked. SK Telecom's first data center opening next year will be the first concrete test of whether Korea's sovereign AI infrastructure can compete on cost and performance with US-based hyperscalers.

Frequently Asked Questions

Q: How much is Nvidia investing in South Korea?

A: Nvidia is investing $1 billion directly in Naver for an AI data center expansion, and has announced a separate $500 billion commercial partnership with SK Group that includes memory chip purchases, supercomputer sales, and over 2GW of data center buildout.

Q: Why is Nvidia partnering with SK Hynix on HBM chip design?

A: High-bandwidth memory (HBM) is the tightest supply bottleneck in AI compute. Only SK Hynix and Samsung can produce the advanced HBM variants Nvidia's processors require. Co-designing future chips with SK Hynix guarantees Nvidia's component pipeline while giving SK Hynix a structural advantage over Samsung.

Q: What is Nvidia's 'circular financing' model and why does it matter?

A: Nvidia invests capital in companies (data center operators, neoclouds, sovereign partners) that then purchase Nvidia hardware. Critics argue this creates dependency loops. Nvidia has committed over $40 billion to AI equity positions in the past year, including $30 billion in OpenAI, using this pattern.