South Korea Announces $950 Billion AI Initiative as SK Hynix Partners with Nvidia
0xBroomberg
South Korea has launched a $950 billion AI initiative spanning Samsung, SK Group, and major U.S. tech firms — with the SK–Nvidia partnership alone exceeding $500 billion — marking the deepest binding yet between Korean chipmakers and the American AI ecosystem.
$950 billion — where is all that money going?
South Korea unveiled the $950 billion AI initiative at an AI summit in San Francisco, involving Samsung Electronics, SK Group, and several U.S. tech companies.
SK Group has signed $750 billion in agreements. Its Nvidia partnership alone tops $500 billion, covering AI data-center construction and next-generation memory-chip supply.
Samsung signed a memorandum with Broadcom worth over $200 billion, committing to supply memory and foundry chips through 2030.
This means → South Korea is betting its semiconductor future on locking in American AI demand, with Samsung and SK each securing different U.S. anchor clients.
What exactly will SK and Nvidia build together?
SK Telecom plans to build a 2-gigawatt AI cloud in South Korea, powered by Nvidia's DSX platform and Nvidia Vera Rubin accelerated computing driven by SK Hynix HBM4 chips.
In plain terms = SK provides the power and real estate, Nvidia provides the chips and platform, and together they build an "AI factory" — a massive data center purpose-built for running AI models. The first facility is set to go live in 2027.
SK Hynix will also co-develop next-generation AI memory solutions with Nvidia, including HBM — high-bandwidth memory, a technology that stacks multiple memory-chip layers to dramatically increase data throughput — targeting workloads from large-model training to agentic AI.
Why are U.S. companies scrambling for Korean chips?
SK Group Chairman Chey Tae-won revealed that Broadcom CEO Hock Tan told him Broadcom's memory-chip demand may exceed available supply; OpenAI has asked SK to supply chips directly; Anthropic has signed supply agreements with both Samsung and SK Hynix.
This reflects a fundamental shift: global AI compute is expanding faster than the chip supply chain can keep up — the question is no longer "who wants to buy" but "who can get the chips."
Nvidia separately announced a plan to invest $1 billion for a 4.5% stake in Korean internet giant Naver. This means → Nvidia is not just partnering at the chip level but buying directly into Korea's AI application ecosystem.
What role is the Korean government playing?
South Korean President Lee Jae-myung attended the San Francisco summit, unveiling the "San Francisco AI Declaration" outlining Korea's AI cooperation strategy.
During the summit he held separate meetings with OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, and Broadcom CEO Hock Tan — putting Korea's chip-supply capabilities on the CEO-level negotiating table.
In plain terms = this is not just a set of corporate contracts — the Korean government has elevated it into state-level AI industrial diplomacy.
What should investors make of this?
$950 billion is the commitment envelope, not booked orders — the gap between memoranda and actual capacity delivery still needs to be tracked.
But the fact that multiple top-tier U.S. AI companies are simultaneously competing for Korean memory-chip capacity is itself a powerful signal: in the AI compute arms race, whoever controls memory-chip capacity holds pricing power.
This means → near-term, order visibility for SK Hynix and Samsung improves markedly; longer-term, the key checkpoint is whether these massive pledges convert into real capacity and revenue on schedule.
Content is for reference only, not financial advice.