[One Year After the Kkanbu Gathering] 1: Memory Boom, Broader Cooperation Must Boost Korean Manufacturers' AI Competitiveness

Jensen Huang assesses Korea's manufacturing capabilities

Pledge to bring in 260,000 GPUs

Catalyst for cooperation across Korea's manufacturing sector beyond semiconductors

Counter China's technological advances and investment push

At a chicken restaurant in Samseong-dong, Gangnam-gu, Seoul, on the evening of October 30 last year, NVIDIA CEO Jensen Huang clinked beer glasses with Samsung Electronics Chairman Lee Jaeyong and Hyundai Motor Group Executive Chair Euisun Chung. Citizens and reporters crowded inside and outside the restaurant, and cheers erupted when Huang shouted, "Everybody, dinner is free!" It was what came to be known as the "Kkanbu gathering."

Lee Jaeyong, Chairman of Samsung Electronics; Jensen Huang, CEO of Nvidia; and Euisun Chung, Executive Chair of Hyundai Motor Group, pose for a commemorative photo after their chimaek gathering at a chicken restaurant in Samseong-dong, Seoul, on the evening of October 30 last year. Photo by Cho Yongjun

At the time, the semiconductor industry was focused on whether Samsung Electronics could increase supply after its high-bandwidth memory (HBM) passed NVIDIA's quality validation. As Korea marks the gathering's first anniversary on October 30, the country's industry faces another challenge: building on the competitive edge gained through memory supply and extending it to products and manufacturing technologies that use artificial intelligence (AI).

Huang cited Korea's manufacturing capabilities at the time and laid out a vision of exporting "intelligence," in addition to semiconductors, automobiles, ships, and electronics. Plans by the government and companies to bring in more than 260,000 graphics processing units (GPUs) were also announced. Since then, companies have expanded cooperation from semiconductor design and production to robotics and autonomous driving. However, it remains necessary to assess whether the announced investments have translated into actual operations and business results.

◆ From memory to factories and robots...Companies broaden cooperation= Samsung Electronics began commercial shipments of HBM4 in February this year and is now pursuing an "AI Factory" initiative with NVIDIA to apply AI to semiconductor design and production. The plan is to increase supplies of AI semiconductors while using them to improve manufacturing processes.

In June, SK hynix announced a multiyear technology collaboration with NVIDIA to jointly develop next-generation memory. The collaboration covers not only memory for AI accelerators but also semiconductor design simulation and the digital transformation of factories.

In the automotive industry, cooperation is expanding into robotics and autonomous driving. At CES 2026, the world's largest electronics and IT trade show, in January this year, Hyundai Motor Group unveiled its industrialization strategy for Atlas, a humanoid robot, and said it would deploy the robot at its U.S. factories starting in 2028. The group is also expanding its autonomous driving collaboration with NVIDIA. It plans to use automotive factories as sites to train robots and validate their performance.

LG is connecting its proprietary AI model, Exaone, with robotics, smart factories, and data center technologies. Doosan is seeking business opportunities in autonomous construction equipment, robotics, and power generation equipment for AI data centers. Korean companies that have supplied semiconductors to meet AI demand are now also pursuing businesses that use AI to improve the performance of factories and machinery.

Huang also visited Korea again in June and met with leading business figures, including SK Group Chairman Chey Tae-won. At Seoul National University, he emphasized that mechanical engineering and manufacturing would become increasingly important in the AI era.

Jang Byung-tak, a professor at Seoul National University, said, "If we validate and accumulate data at industrial sites, we can also be sufficiently competitive in physical AI."

However, these investments and research and development (R&D) efforts did not begin with the Kkanbu gathering. Since the meeting helped expand cooperation on existing businesses, it is now important to assess what results have been achieved in areas such as product development and productivity improvements.

◆ 260,000 GPUs: Between announced plans and actual deployment= To establish a foundation for cooperation, it is also necessary to distinguish between the announced number of GPUs and the number actually available for use.

According to parliamentary audit materials that Choi Hyung-doo, a People Power Party lawmaker on the National Assembly's Science, ICT, Broadcasting and Communications Committee, received from the Ministry of Science and ICT and the National IT Industry Promotion Agency (NIPA), about 81% of the 260,000 GPUs, or 210,000 units, are included in private companies' deployment plans. The plans comprise more than 60,000 units for Naver Cloud, more than 50,000 each for SK Group and Samsung Electronics, and 50,000 for Hyundai Motor Group.

The government and companies have no separate implementation agreements or commitments for these units, and deployment and operational figures by company and year have not been disclosed. However, the absence of an agreement with the government does not indicate whether individual companies have contracts with NVIDIA.

Progress on the government's share of 50,000 GPUs also varies depending on when they were procured. About 13,100 units purchased last year are in service. Of the 9,704 units contracted for this year, 2,608, or 26.9%, had actually been delivered as of October 8. By contrast, the budget execution rate stood at 79.8% as of the end of August. In other words, budget execution does not necessarily mean that equipment has been delivered or services have begun.

The average monthly utilization rate for existing government GPUs rose from 26.8% in March to 51.5% in September. Choi said, "GPU competitiveness is determined not by how many units have been contracted for, but by when they arrive and how quickly they are put to use."

For GPU deployments to support companies' product development and factory innovation, it is important to ensure that the equipment is brought online on schedule and that follow-up measures enable companies and researchers who need it to use it.

◆ Memory boom must translate into manufacturing competitiveness= The memory boom, fueled by AI demand, is giving Korean companies room to make follow-on investments. The Korea Development Institute (KDI) forecast economic growth of 3.2% this year, citing the semiconductor boom.

However, market expectations cannot be sustained by improving earnings alone. Concerns are also growing over China's technological advances and the profitability of massive AI investments. On October 15, Samsung Electronics and SK hynix shares fell 9.48% and 11.06%, respectively. Regardless of the current boom, companies now need to demonstrate their future competitiveness and the results of their investments.

Korea's lead in manufacturing AI is not guaranteed either. According to the International Federation of Robotics (IFR), about 59% of new industrial robots installed worldwide last year were concentrated in China. China is also accelerating its pursuit of Korea in memory and AI.

To secure a competitive edge, Korean companies need to develop their accumulated factory data, design capabilities, and experience validating robots in real-world settings into proprietary technologies. Even when using the same NVIDIA platform, results can vary depending on how effectively companies use data from their operations.

Data center investments should be expanded in line with demand and profitability, while research into next-generation memory, industrial AI, and robotics should continue steadily. When profits from the memory boom and the computing capacity secured through investment translate into growth for domestic equipment, parts, and software companies, competitiveness across the manufacturing sector can also improve. A year after the Kkanbu gathering, it is time to look beyond the scale of announced investments and assess the results in actual factories and products.

This content was produced with the assistance of AI translation services.