South Korea rolls out 10,000 Nvidia GPUs for AI projects

South Korea has begun distributing 10,000 Nvidia GPUs to universities, research institutes, and AI projects as part of a planned supply of 260,000 units through 2030.

The initiative was announced last October, with the goal of expanding AI infrastructure across the country.

The first batch includes 4,336 units, consisting of H200 and B200 chips, with additional shipments expected.

The GPUs are allocated free of charge to academic and public research institutions, while private companies, including startups and SMEs, are required to pay a reduced fee of 5 to 10 percent of market price.

Some units are reserved for national AI projects, such as the sovereign AI foundation model.

The government also plans to deploy another 15,000 GPUs by 2028 via the National AI Computing Center.

🔗 Source: The Korea Herald

🧠 Food for thought

Implications, context, and why it matters.

South Korea’s GPU rollout links government and industry, with one estimate at $7.8 billion to $10.4 billion

The effort runs as a public-private partnership, rather than only a South Korean government program.
The 260,000-GPU plan sets aside about 52,000 GPUs for the South Korean government, 50,000 each for Samsung, SK Group and Hyundai Motor Group, and 60,000 for Naver Cloud 1.
One estimate cited by Light Reading from The Chosun Daily values the program at about 14.8 trillion won (about $7.8 billion to $10.4 billion), using an assumed $30,000 to $40,000 unit price for Nvidia’s Blackwell B200 GPU 2.
Nvidia says the agreements would lift South Korea’s use of Nvidia AI chips to more than 300,000 from about 65,000 3.

The rollout backs AI sovereignty and lowers reliance on a single supplier

The initiative sits within Nvidia’s global sovereign AI push, and Nvidia has said sovereign AI deals would likely add more than $20 billion to fiscal year 2026 revenue 3.
Yahoo Finance reported that these deals also arrive as US-China trade frictions and export controls have cut into Nvidia’s China business 3.
South Korean officials have also pushed for domestically developed neural processing units (NPUs), specialized chips designed to accelerate AI workloads, to curb long-term dependence on Nvidia hardware 1.
Local startup Rebellions has announced a collaboration with Wert Intelligence to build an on-premises AI infrastructure for patent analytics using Rebellions’ ATOM-Max NPU 4.