Naver, Nvidia declare alliance to build global AI factories

Full-stack AI cooperation spans infrastructure, models and services

Lee Hae-jin, Jensen Huang to meet Friday ahead of Nvidia campus visit Monday

Partnership deepens from physical AI to AI factory buildout

From left: Choi Su-yeon, Naver CEO; Lee Hae-jin, Naver board chairman; Jensen Huang, Nvidia CEO; Jay Puri, Nvidia executive vice president; and Kim Yu-won, Naver Cloud CEO, pose for a photo after a meeting at Nvidia's Taiwan office last year. [Naver Cloud]

Naver is teaming up with Nvidia to formally launch a global AI factory construction initiative.

The two companies aim to strengthen their global competitiveness across the entire AI ecosystem — spanning infrastructure, models and services — deepening an already wide-ranging partnership.

Further details are expected following a meeting between Naver board chairman Lee Hae-jin and Nvidia CEO Jensen Huang scheduled for Friday. The collaboration, which has expanded from physical AI to AI factory buildout, appears set to grow well beyond a strategic alliance.

Naver Cloud said Monday it shared the direction of the partnership at the Nvidia Cloud Partner Summit held in Taiwan.

"Naver Cloud has robust full-stack technology capabilities covering everything from AI infrastructure to services, making us a perfect fit for Nvidia's AI factory platform strategy, which encompasses energy, chips, infrastructure, models and applications," Naver Cloud CEO Kim Yu-won said, explaining the rationale for the partnership.

Huang had officially named Naver Cloud a key partner in the global AI ecosystem at his keynote address for Nvidia GTC Taipei 2026 on Sunday. He placed a heart symbol and the Nvidia logo next to the Naver Cloud name — rendering it as "Nvidia ♥ Naver Cloud" — to underscore the closeness of the relationship.

Naver Cloud CEO Kim Yu-won speaks at the Nvidia Cloud Partner Summit held in Taiwan on Monday. [Naver Cloud]

At the heart of the partnership is deep technical cooperation across the full AI stack — from infrastructure and AI models to physical AI.

Naver Cloud will use Nvidia's open large language model Nemotron 3 Ultra to advance its HyperCLOVA X model. The two companies also plan to jointly research optimization techniques and foundational technologies for large-scale language models.

Cooperation on physical AI is also intensifying. In March, Naver Cloud unveiled the Seoul World Model — a digital replica of Seoul built using real-world data and powered by Nvidia's physical AI platform Cosmos — and has been refining the model's performance since.

The model drew attention for reconstructing Korea's actual road environments and spatial structures, trained on 1.2 million panoramic images collected across Seoul using domestic map data.

"Our collaboration with Nvidia goes beyond a simple GPU supplier-customer relationship — it is a strategic decision to jointly develop AI technology and expand the global AI ecosystem," Kim said. "We will establish ourselves as a key supplier supporting strong AI demand in the Asian market and as an unrivaled AI infrastructure hub."

Nvidia CEO Jensen Huang introduces global AI-native cloud partners during his keynote address at Nvidia GTC Taipei 2026 in Taipei, Taiwan, on Sunday, naming Naver Cloud — a subsidiary of Naver. He placed a heart symbol and the Nvidia logo next to Naver Cloud's name, rendering it as "Nvidia ♥ Naver Cloud," to highlight the two companies' partnership. [Naver Cloud]

Following the meeting between Lee and Huang on Friday in Korea, the two companies plan to formally unveil the detailed execution plan for the global AI factory initiative on Monday. Huang will visit Naver's second headquarters, "1784," in Bundang-gu, Seongnam, Gyeonggi Province, on that day.

The detailed cooperation plan to be released Monday is expected to cover robotics and digital twin collaboration, commercialization of physical AI using Nvidia platforms, and upgrades to AI cloud and GPU infrastructure.

The two companies have built up a track record of collaboration over several years. Team Naver became the world's first adopter of Nvidia's SuperPOD supercomputing infrastructure in 2019. At the APEC CEO Summit held in Gyeongju, North Gyeongsang Province, last October, Lee and Huang agreed to jointly develop a physical AI platform. More recently, Madison Huang, Nvidia's senior director of Omniverse and robotics product marketing, visited Naver's 1784 headquarters to discuss technology and business cooperation with Naver CEO Choi Su-yeon and Naver Cloud CEO Kim.

The closeness of the relationship is also reflected in AI infrastructure investment. Naver Cloud has been allocated 60,000 Nvidia GPUs — 10,000 more than the allocations given to Samsung Electronics, SK Group and Hyundai Motor Group respectively. The figure underscores Naver's position as a core Nvidia partner as it pursues HyperCLOVA X, cloud infrastructure and sovereign AI initiatives.

rim@heraldcorp.com