South Korea has announced a $919 billion AI investment plan to build 18.4 gigawatts of data center computing power by 2035.Article author and source: SemiAnalysis
South Korea is elevating artificial intelligence to a national strategy, with an investment scale that has drawn widespread market attention. But in this nationwide race, who will truly benefit? The answer is far more complex than it appears.
In July this year, the South Korean government announced plans to invest $919 billion to build 8.4 gigawatts of data center computing power by 2029, ultimately expanding the total capacity to 18.4 gigawatts by 2035.
This initiative will open up a vast sovereign AI market for NVIDIA, with SK Group committing to purchase 2 gigawatts of Rubin systems and Naver signing an order for 200 megawatts.
However, this is not purely beneficial for Samsung and SK Hynix. South Korea lacks domestic suppliers capable of independently producing cutting-edge AI accelerators. According to SemiAnalysis, NVIDIA may have already entered into a long-term agreement with SK Hynix for SOCAMM and secured favorable pricing for HBM in 2027.
Meanwhile, South Korea’s government-led competition for an “independent AI foundation model” is underway, but it has made a baffling decision at the critical elimination stage—excluding Motif Technologies, a small startup ranked first in the benchmark tests.
This outcome reveals an inherent contradiction in the sovereign AI race: there is often a significant gap between governments' initial goal of achieving technological independence and the actual logic of resource allocation.
For investors, South Korea’s trillion-dollar commitment signifies a significant expansion of NVIDIA’s customer base, but for Samsung and SK Hynix shareholders, this massive construction plan may not yield equivalent returns.
Why NVIDIA Supports Open Source and Sovereign AI
NVIDIA CEO Jensen Huang recently published an open article outlining the importance of open-source AI, which was subsequently endorsed by nearly all major AI companies (excluding Anthropic). Behind this statement lies a clear business logic.
Selling frontier model tokens at API prices is the application with the highest return on computational power per unit. This means Anthropic and OpenAI will capture an increasingly larger share of net new computational demand, exposing NVIDIA to the risk of over-concentration among a small number of mega-buyers.
Even when including hyperscale cloud providers in the statistics, there are only seven true core customers, and all of them except SpaceX are actively developing their own chips to reduce dependence on NVIDIA.
Sovereign AI is a crucial breakthrough for NVIDIA in diversifying its customer base. The classic technology competition strategy of “commoditize your complements” applies here as well: AI models and applications are complements to NVIDIA’s GPUs; the thriving ecosystem of open-source models means greater demand for computing power.
NVIDIA’s recent announcement of a $500 billion memorandum of understanding with the world’s largest capital allocator is a financial extension of this strategy—transforming “NVIDIA AI Factory compute” into an investable asset class, enabling pension funds, insurance companies, and private credit capital to flow into large-scale compute procurement.
Data center construction boom: NVIDIA benefits, HBM landscape uncertain
South Korea has confirmed three under-construction data center sites, totaling 4.4 gigawatts. SK Group, GS Group, and Naver are respectively responsible for the initial construction of 5 gigawatts, 2.4 gigawatts, and 1 gigawatt, with SK Group also set to handle the remaining 10 gigawatts in the second phase. President Yoon Suk Yeol has explicitly positioned AI infrastructure as a historic endeavor comparable to the broadband expansion of the 1990s.
However, this large-scale construction plan does not bring only straightforward benefits to South Korea's domestic semiconductor companies.
South Korea currently lacks domestic suppliers capable of independently producing cutting-edge AI accelerators. While Rebellions and FuriosaAI have made progress, they still lag significantly behind NVIDIA in terms of deployment scale, software maturity, and commercial availability. Therefore, no amount of investment can enable South Korea to build a competitive accelerator ecosystem solely through domestic manufacturing.
This is precisely where the strategic value lies in NVIDIA deepening its relationships with Samsung and SK Hynix. Samsung is reportedly planning to build an AI factory equipped with over 50,000 NVIDIA GPUs; SK Telecom’s 2-gigawatt DSX AI factory is expected to deploy Vera Rubin systems powered by SK Hynix’s HBM4.
According to SemiAnalysis, NVIDIA may have reached a long-term agreement with SK Hynix on SOCAMM, securing favorable pricing and large-scale procurement commitments for HBM in 2027, although the final terms may still differ from current estimates.
If this assessment is accurate, South Korea’s current wave of infrastructure investment, while strengthening NVIDIA’s market position, may yield returns for Samsung and SK Hynix shareholders that are far more complex than their apparent financial figures suggest.
South Korea's "Squid Game": The Rise and Fall of a Sovereign AI Tournament
In June 2025, the South Korean government officially launched the "Independent AI Foundation Model" project, aiming to develop a large language model that Korean institutions can train, modify, and operate independently, without relying on foreign AI labs.
The most distinctive feature of this project is its competition format. Rather than pre-selecting a "national team," the government employs a knockout tournament model: all participants receive subsidies in computing power, data, and researchers, and are evaluated every six months, with resources from losing teams reallocated to the winners.
A total of 15 consortia registered for the competition; after an initial document review, 10 were shortlisted. In August 2025, the final five teams were confirmed: Naver Cloud, LG AI Research, SK Telecom, NC AI, and Upstage.
In the first round, the South Korean government provided each team with approximately 3,000 H100-equivalent computing units, invested around $45 million to purchase data, and allocated approximately $1.4 million per company for recruiting overseas talent.
However, an unexpected change occurred at the end of the first round: NC AI, which ranked last, was eliminated, but Naver Cloud was also disqualified.
The South Korean government subsequently added Motif Technologies as the fourth participant; the company was spun off from South Korean AI foundational software firm Moreh in February 2025.
The best open-source model is unexpectedly eliminated, sparking questions.
Four teams publicly released their latest models in July 2026, with evaluation results announced on August 18. Data shows that two startups, Motif and Upstage, significantly outperformed teams backed by large conglomerates, with their models requiring less than half the parameters—making the performance gap even more striking.
According to SemiAnalysis data, Motif 3 leads Upstage by 10 points on the Artificial Analysis intelligence index, making it the top Korean model, surpassing the two best U.S. open-source models, Inkling and Nemotron 3 Ultra.
Motif is a startup with fewer than 30 employees, having raised only $17 million in funding and access to less than 768 B200 GPUs. The total compute cost for training Motif 3, including experimental phases, is approximately $15 million.
It was therefore surprising that Motif was eliminated. According to the competition scoring criteria, benchmark tests accounted for 40%, expert evaluations for 35%, and user testing for 25%. Motif ranked first in the benchmark tests but came last in the other two categories, ultimately leading to its elimination.
The specific methodology for the review lacks transparency. In user testing, the test tasks and evaluation criteria were not disclosed, and the tests were not conducted blind, potentially favoring more well-known large companies.
It is understood that the expert review panel also expressed concerns that Motif’s technology might leak to foreign investors—yet this directly contradicts the competition’s requirement that all technical components must be open source.
In addition, certain evaluation criteria, such as "ecosystem impact" and "future roadmap," structurally favor large enterprises like LG and SKT.
Now, Motif may be forced to relocate outside Korea to secure the capital and computing power needed to continue its research—an outcome rich in irony for the most surprising contender in the nation’s AI race.