▲ Nvidia CEO Jensen Huang greets Naver Board Chairman Lee Hae-jin during his visit to Naver's 1784 headquarters in Seongnam, Gyeonggi Province, on Monday.
Naver is teaming up with Nvidia to build a gigawatt (GW)-scale, ultra-large global artificial intelligence (AI) factory.
Naver and Nvidia announced on Monday that they have agreed on a joint global AI factory project and will gradually expand their AI infrastructure business, starting with the operation of a 55-megawatt (MW) infrastructure in 2027.
An AI factory refers to next-generation AI infrastructure designed to train and run inference on large-scale AI models and apply them to actual services and industrial sites.
Simply put, it is an intelligent infrastructure that inputs power and data to mass-produce tokens—the core unit of AI—and is a next-generation computing facility designed to go beyond traditional data centers to optimize AI training, inference, and deployment.
In manufacturing, in particular, it is used to design solutions for automated quality inspection, predictive equipment maintenance, and process optimization, enabling AI to make decisions and perform control tasks.
Naver explained that this collaboration goes beyond simple GPU supply or technical alliance, representing a partnership that covers the entire value chain, including global demand discovery, infrastructure construction, and capital cooperation.
Naver plans to participate as a core global partner sharing business outcomes and risks with Nvidia, aiming to build an AI infrastructure ecosystem spanning the Asia-Pacific region, the Middle East, and Europe.
Lee Hae-jin, chairman of Naver's board, and Nvidia CEO Jensen Huang met on Monday at Naver's second headquarters, 1784, to discuss the roadmap for the AI factory business currently being pursued by both companies and plans for joint entry into the global market.
At the meeting, Chairman Lee emphasized, "Naver is the world's first company to build a 'SuperPOD' with GPUs," adding, "We stably operate the largest service in Asia and have built and operated our own data centers, so we are already fully prepared for the AI factory business."
He continued, "It is not a matter of us building a cloud and an AI factory in the future; Naver is the only company capable of meeting the demand of the GPU and AI markets, where demand is currently surging most rapidly."
Lee added, "In that sense, Nvidia chose Naver to expand AI factories not only to South Korea but also to Asia. A 1GW-scale AI factory is an immense scale, making this a huge opportunity for Naver."
Naver plans to first utilize "Gak Sejong," South Korea's largest hyperscale data center, as the outpost for its AI factory business.
The strategy is to start operating a 55MW infrastructure in the first half of 2027, and expand the scale of overseas infrastructure to 100MW within the same year and to 200MW by 2028.
In the long term, the companies have decided to pursue the construction of GW-scale infrastructure.
One GW is equivalent to approximately four times the maximum capacity of Gak Sejong, a scale capable of accommodating hundreds of thousands of Nvidia's latest GPUs.
Naver plans to secure a leading position in the global AX (AI transformation) market by establishing an integrated partnership with Nvidia to build AI factories, covering all stages of the value chain from global demand discovery to capital cooperation.
Naver and Nvidia will also expand their technical cooperation.
The plan is to combine Naver's capabilities in building and operating its own large-scale GPU clusters and its data center operational know-how with Nvidia's next-generation high-performance infrastructure platform to enhance data center operational efficiency and business viability.
In the AI model sector, Naver will utilize Nvidia's open large language model (LLM) Nemotron technology to further advance the performance of its hyperscale AI model, HyperCLOVA X.
Naver expects that combining its accumulated capabilities in building and operating large-scale proprietary GPU clusters and data center know-how with Nvidia's next-generation high-performance infrastructure platform, "DSX," will maximize data center operational efficiency.
Naver also plans to initiate full-scale cooperation in the field of physical AI, combining its proprietary spatial modeling technology and Street View data with Nvidia's world foundation model platform, Cosmos, to develop the "Seoul World Model."
At the same time, the company will pursue the development of next-generation technologies in the field of spatial intelligence.
Previously, on June 2, Naver Cloud announced at the Nvidia Cloud Partner Summit held in Taiwan that it is pursuing cooperation with Nvidia to build global AI factories.
Chairman Lee said, "It is highly encouraging that through this alliance, we can present a concrete alternative for regions and countries around the world to build their own sovereign AI capabilities," adding, "This has provided an opportunity for Naver's technological infrastructure competitiveness to take a leap forward in the global market."
(Photo: Yonhap News)
※ Please note: This article was translated by AI and may contain errors.