SK Group & Nvidia Sign $500 Billion Deal: South Korea's Big AI Infrastructure Bet

By Imran Khan (AI Tech Safar)

Nvidia just signed a $500 billion partnership — and its stock barely moved. Shares closed down 0.8% the day after the SK Group deal was announced, sitting roughly 10% below their 52-week high, with a relative strength index of 52.6 that signals neither excitement nor concern. For a deal this size, that's a strange reaction, and it tells you something the headline number doesn't: Wall Street has started treating half-trillion-dollar AI infrastructure deals as routine.

The deal itself is real and genuinely significant — South Korea's SK Group and Nvidia have signed letters of intent worth more than $500 billion, covering new AI factories, next-generation memory development, and large-scale cloud computing capacity. It's the largest single AI infrastructure commitment Nvidia has made in Asia-Pacific. It's also, as it turns out, just one piece of a much bigger week for Nvidia's capital deployment.

Deal Breakdown: What's Included

Component	Company Involved	Details
AI Cloud Infrastructure	SK Telecom	2GW AI cloud using Nvidia's DSX platform & Vera Rubin accelerators
AI Memory Supply	SK hynix	Long-term HBM4 supply & co-development; holds ~70% of Nvidia's HBM4 allocation for Vera Rubin
First Facility Timeline	SK Group + Nvidia	Operations begin in 2027
Core Architecture	Nvidia	Full-stack DSX design combining compute, systems & software
National Context	South Korea	Part of a broader ~$950 billion national AI infrastructure push

What Happened: Inside the SK-Nvidia Partnership

The collaboration was announced July 24–25, 2026, during South Korean President Lee Jae-myung's Silicon Valley visit, with SK Telecom set to deploy a large-scale AI cloud built on Nvidia's newest DSX architecture, paired with the Vera Rubin accelerated computing platform. Vera Rubin entered full production in June 2026, with customer shipments beginning in Q3 — meaning this deal is essentially first in line for Nvidia's next-generation hardware, not a commitment to older technology.

Separately, SK hynix locked in a long-term supply agreement with Nvidia to jointly optimize next-generation HBM4 memory. According to TrendForce industry data, SK hynix currently holds roughly 70% of Nvidia's total HBM4 allocation for the Vera Rubin platform, with Samsung and Micron splitting the remainder — meaning this deal isn't just a partnership, it's a formal acknowledgment of a supply dominance SK hynix already had.

Why It Matters: The Part the Headline Number Hides

Here's what most coverage of this deal missed: it's not Nvidia's only major bet from that same week. Just days before the SK Group announcement, Nvidia revealed a roughly $5 billion investment in Safe Superintelligence, Ilya Sutskever's AI startup, giving it early access to Vera Rubin. Around the same period, Nvidia also committed $1 billion to Naver's Korean AI data center expansion, with Brookfield agreeing to fund up to $9 billion more of that same project.

Add it up, and Nvidia deployed capital across at least three separate major AI bets — a Korean conglomerate, a rival AI lab founded by a former OpenAI chief scientist, and a Korean internet company — within roughly the same two-week window. That's a company spreading enormous sums across very different types of partners simultaneously, not making one big strategic bet and waiting to see how it plays out.

💡 AI Tech Safar Insight

The flat stock reaction is the most interesting data point in this entire story, and it's worth sitting with. When Nvidia's stock barely moves on a $500 billion announcement, one of two things is true: either the market genuinely believes these Korea-scale AI infrastructure deals are now Nvidia's baseline pace of business rather than a special event, or investors are starting to discount headline deal values because so much of the money represents future purchase commitments rather than cash changing hands today — the $500 billion here includes projected memory purchases and hardware sales stretched over years, not money sitting in an account.

The more revealing signal is the HBM4 allocation number: SK hynix's ~70% share isn't something this deal created — it's something this deal made official. That distinction matters for how you read Nvidia's Korea strategy going forward. Nvidia isn't diversifying its memory supply chain by signing this deal; it's doubling down on the supplier that already dominates the component. If SK hynix ever faces production problems, a disproportionate share of Nvidia's next-generation AI hardware roadmap sits exposed to a single company's ability to deliver — a concentration risk that gets easy to overlook amid the headline enthusiasm about a "golden age" for Korea.

Zoom out further, and the Safe Superintelligence and Naver deals happening in the same window suggest Nvidia is running a portfolio strategy, not a single strategic pivot: back the country that makes the memory, back the lab that might build the next frontier model, back the internet company that needs the compute — hedging across the entire stack rather than betting on any one relationship carrying the next phase of the AI buildout alone.

Frequently Asked Questions (FAQs)

Q1: If SK hynix faces a production delay, can Samsung or Micron fill the gap?
Not easily, at least in the short term. With SK hynix holding roughly 70% of Nvidia's HBM4 allocation, Samsung and Micron's combined remaining share leaves little spare capacity to absorb a major supply disruption without delaying Vera Rubin shipments industry-wide.

Q2: Why didn't investors react more strongly to a $500 billion deal?
Beyond the headline figure being a multi-year projected commitment rather than upfront cash, Nvidia has now signed several deals at this scale in recent months — meaning the market may simply be pricing in mega-deals as Nvidia's new normal rather than one-off news.

Q3: Does Nvidia investing in a rival AI lab like Safe Superintelligence create a conflict of interest?
It's an unusual position — Nvidia profits from selling chips to nearly every major AI lab, including competitors to its other customers, which means its financial success is increasingly tied to the industry's overall growth rather than any single lab winning.

Q4: Could South Korea's heavy reliance on Nvidia hardware become a strategic risk?
Potentially. Building national AI infrastructure so closely around one supplier's architecture and roadmap ties South Korea's AI ambitions to Nvidia's own execution — if Nvidia faces production setbacks or export restrictions, Korea's 2027 timeline moves with it.

What Do You Think?
Does Nvidia spreading billions across SK Group, Safe Superintelligence, and Naver in the same two weeks look like smart diversification, or overextension? Share your take in the comments below!

Source: Reporting based on Bloomberg, Yahoo Finance / GuruFocus, Startup Fortune, and ad-hoc-news.de, with HBM4 allocation data from TrendForce.