Naver to roll out AI agents, ramp up GPU investments

South Korean internet giant Naver is planning to introduce AI agents across key services like shopping and search next year, while investing over 1 trillion won ($691 million) to purchase graphic processing units to bolster infrastructure, ramping up its drive to become a global AI leader.

At the company’s flagship tech conference, Dan25, held at Coex in southern Seoul on Thursday, CEO Choi Soo-yeon outlined Naver's AI strategy in a keynote speech, focusing on artificial intleligence agent services and expanding infrastructure to support physical AI.

Naver plans to deploy AI agents in shopping in the first quarter of 2026, which will bring together “each user’s preferences, budget, search history and review data to offer the best possible choice from exploration to purchase,” Choi explained.

Later in the summer, Naver will introduce an AI tab for its search service.

“Instead of wondering what to type, users will simply talk to the agent, which will interpret their intent and connect them to the right content, products and services and even execute tasks,” said Choi.

Building on its ambition to establish Korea's largest and most advanced AI infrastructure, Naver said it will invest over 1 trillion won in GPUs — high-performance accelerators that power computation in AI data centers — next year.

Part of this investment includes Naver’s recent deal to purchase 60,000 units of Nvidia’s most advanced Blackwell GPUs, as part of the US chip giant’s plan to deploy a total of 260,000 GPUs to South Korean firms.

Naver Cloud CEO Kim Yu-won told reporters that the exact delivery time of Nvidia's GPU has yet to be finalized. "Naver has been purchasing tens of thousands of GPUs every year, but this time we are investing more aggressively," he said during a media briefing held on the sidelines of the conference.

He added that the GPUs will be used across Naver's full AI services, from data centers, to training its HyperClova X model and powering on-service AI, physical AI and vertical AI applications. "Even with 60,000 units, we believe it won't be enough (to meet the needs)."

Choi said Naver will begin testing its physical AI infrastructure later this year, connecting its headquarters in Seongnam, Gyeonggi Province, and its data center in Sejong.

Physical AI refers to AI embedded in physical systems and real world sites, such as robots and self-driving vehicles. Naver has been investing heavily in developing a next-generation physical AI platform, deploying AI infrastructure across key industries, including semiconductors, shipbuilding and energy.

“By adding Naver’s unrivaled AI software capabilities to the strong competitiveness of Korea’s core manufacturing industries, such as semiconductors, automobiles and shipbuilding, we will accelerate AI transformation and innovation across Korea’s industries,” said Choi.

“Based on our full-stack AI technology capabilities, we will contribute to Korea becoming one of the top three AI powerhouses.”

Naver Cloud, the core technology unit behind Naver’s AI ambitions, echoed this focus on physical AI applications in real industrial environments.

“Physical AI is about machines that can sense and understand space, as well as make judgement and act on their own,” said Kim during a keynote address at the forum. “At Naver’s 1784 headquarters and data centers, robots are collecting real-world data and connecting to the cloud platform and on-board AI systems to form a continuously evolving physical AI ecosystem.”

Kim added that Naver Lab, the company’s research and development arm, will unveil a 1-meter-tall humanoid robot at the end of this month, which will be “working hard” inside Naver’s headquarters.

Based on this technology and infrastructure, Kim said Naver Cloud will partner with firms in shipbuilding, energy and biotech to support AI integration across manufacturing processes and plan to expand the model internationally to markets such as Saudi Arabia, Thailand and Japan.