[Edaily Reporter Kim Hyun-ah] #SamsungElectronics, #SKHynix, #NAVER, #SKTelecom, and domestic robotics companies are rapidly joining the global AI ecosystem centered around NVIDIA. The scope of cooperation is also expanding rapidly, ranging from High Bandwidth Memory (HBM) supply to AI factories, data centers, and physical AI.

However, concerns are growing that South Korea’s technological leadership in the AI industry could weaken as NVIDIA’s influence expands beyond AI semiconductors to encompass data center interconnect technologies, software platforms, and the broader robotics ecosystem. In particular, the competition between NVIDIA’s proprietary technology, “NVLink,” and the open standard “CXL (Compute Express Link)” is emerging as a new variable for the domestic semiconductor industry.

CXL Ecosystem Shaken by NVLink Expansion

According to industry sources on the 10th, NVIDIA is expanding its influence in the AI infrastructure market beyond just GPUs to include inter-GPU connectivity technologies.

The key is NVLink. NVLink is a technology that connects the hundreds to thousands of GPUs required for AI model training into a single system. NVLink switches also play a central role in NVIDIA’s latest Blackwell platform.

The issue is that as the NVLink-centric structure strengthens, the position of CXL—an open interconnect technology—could be undermined. CXL is a next-generation data center technology that connects CPUs, GPUs, and memory as a single resource pool, and it is an area in which Samsung Electronics, SK Hynix, and domestic startups have been actively investing.

Industry observers suggest that if AI data centers are reorganized around NVIDIA, the pace of CXL market expansion could slow down more than expected.

An official from a startup specializing in CXL switches said, “Just a few months ago, discussions regarding CXL switches were lively, but it is true that the market sentiment has cooled somewhat recently as the NVLink-centric structure has strengthened. We are redirecting our development personnel to other areas.”

Growing Concerns Over 'Technological Dependency' Amid Expanding Collaboration

Collaboration with NVIDIA clearly presents an opportunity for domestic companies.

Samsung Electronics and SK Hynix are benefiting from the expansion of HBM supply, while Naver and SK Telecom are pushing to expand their AI Factory (massive AI data center) businesses. Robotics companies such as LG and Doosan are also leveraging NVIDIA’s Physical AI platform to target the global market.

In fact, during his recent visit to South Korea, NVIDIA CEO Jensen Huang praised the country as a key hub in the Asian AI ecosystem and moved to expand cooperation with the semiconductor, telecommunications, and robotics industries.

On the other hand, some point out that if a closed ecosystem centered on CUDA—NVIDIA’s AI operating system—and NVLink—its AI highway—expands, it could limit the growth potential of Korea’s domestic AI semiconductor and software industries. This could weaken the competitive position of domestic NPU (neural processing unit) companies, as well as those involved in the CXL ecosystem and the development of robot foundation models.

"We Must Balance Leveraging NVIDIA with Securing AI Sovereignty"

Experts emphasize the need for a "two-track strategy" that combines entering the global market by leveraging the NVIDIA ecosystem with achieving self-reliance in core technologies.

Yoon Seok-bin, a special professor at Sogang University’s Graduate School of AI and Software, stated, “Global expansion through the NVIDIA ecosystem and the establishment of sovereign AI are tasks that must be pursued simultaneously.” He added, “Since Korea possesses competitiveness in semiconductor design, data, and content, it must also cultivate self-reliance in core technologies such as NPUs.”

He further emphasized, "While cooperation with NVIDIA is necessary, we must be cautious about ceding technological leadership."

An industry insider remarked, “In areas where we have secured a competitive edge, such as HBM, we must widen the gap further, while in strategic fields like AI semiconductors and next-generation connectivity technologies, we must continue our efforts to secure proprietary technologies,” adding, “Participation in the NVIDIA ecosystem and securing AI sovereignty are separate challenges.”

Analysts suggest that the challenge for the Korean industry in the AI era lies in finding a balance between leveraging NVIDIA’s massive ecosystem for growth while maintaining independent competitiveness in core technologies.