Computing

NVIDIA and Korea Expand Collaboration to Advance Sovereign AI Infrastructure with NVLink Fusion

Tendela Briefing ·
3 views · 0 Comments
NVIDIA and Korea Expand Collaboration to Advance Sovereign AI Infrastructure with NVLink Fusion

Image: NVIDIA · Source

NVIDIA teams with Korea's Ministry of Science and ICT, Rebellions, and SK Telecom to integrate Korean AI semiconductors with NVIDIA's NVLink Fusion platform, aiming to build sovereign AI ecosystem including chips, data centers, and AI models.

NVIDIA has announced an expanded partnership with South Korea's Ministry of Science and ICT (MSIT), AI semiconductor startup Rebellions, and telecom leader SK Telecom to develop a sovereign AI ecosystem tailored to Korea's language, culture, and industrial needs. This collaboration focuses on integrating Korea's next-generation AI chips, data center infrastructure, talent development, and AI models using NVIDIA's NVLink Fusion open infrastructure platform.

A core part of the initiative involves Rebellions adopting NVIDIA's NVLink Fusion technology to connect its Rebel R200 neural processing units (NPUs) within rack-scale systems based on NVIDIA’s MGX architecture. These systems aim to provide low-latency and energy-efficient AI inference capabilities optimized for frontier models, including multitrillion-parameter mixture-of-experts. SK Telecom plans to deploy these integrated systems in its AI data centers, leveraging its proprietary foundation models and public AI services under Korea’s government-backed AI for All project.

NVLink Fusion extends NVIDIA’s AI platform by allowing custom AI processors, or XPUs, such as Rebellions’ NPUs, to tightly couple with NVIDIA GPUs via high-bandwidth, low-latency networking. This integration enables heterogeneous computing strategies where different AI tasks are assigned to the most suitable processors, such as GPUs handling prefill phases and NPUs managing decode processes, thereby enhancing inference throughput and power efficiency.

Alongside infrastructure deployment, NVIDIA and Korean partners will foster ecosystem growth through training, certification, hackathons, and technical seminars to strengthen GPU and AI resource utilization nationwide. The collaboration also lays groundwork for other Korean AI semiconductor companies to join NVIDIA's NVLink Fusion ecosystem, further expanding the country's sovereign AI capabilities.

South Korea’s Deputy Prime Minister and MSIT Minister Bae Kyung-hoon emphasized the strategic importance of this effort, noting the government’s active role in supporting collaborations that enable domestic AI chipmakers to compete alongside global leaders like NVIDIA. The partnership envisions a comprehensive sovereign AI stack combining Korean silicon innovation, NVIDIA’s infrastructure technology, and SK Telecom’s deployment expertise.

Rebellions expects to have R200-based systems ready for deployment by 2028, marking a significant milestone in Korea's ambition to create a sovereign AI computing foundation. This initiative represents a coordinated push to harness national AI talent and technology in ways that serve Korea’s unique linguistic and industrial needs, potentially positioning the country as a global AI leader.

Sources and original reporting

Comments (0)

No comments yet. Start the discussion.

Write a comment

Comments are published after moderation. Your name and comment will be visible publicly. Account