Expanding AI Infrastructure Services with the GPUaaS Model
Naver Builds an Ecosystem for Industry-wide AI Adoption
Naver Cloud has announced its commitment to lead the subscription-based Graphics Processing Unit (GPU) cloud service (GPU as a Service, or GPUaaS) market, leveraging its artificial intelligence (AI) infrastructure.
Sangjun Lee, Chief Information Officer and Head of Operations at Naver Cloud, is explaining Naver's AI infrastructure competitiveness at the Tech Meetup held on the 27th at 'Gak Sejong,' Naver Data Center in Jiphyeon-dong, Sejong City. Provided by Naver Cloud
Sangjun Lee, Chief Information Officer at Naver Cloud, stated at the Tech Meetup held on the 27th at 'Gak Sejong,' Naver Data Center in Jiphyeon-dong, Sejong City, "The competitiveness of AI infrastructure depends not only on how many GPUs you secure, but also on how stably and efficiently you operate the resources you have acquired." He added, "We will advance our accumulated AI infrastructure operation capabilities into the GPUaaS model to create an ecosystem where domestic companies can easily utilize AI."
This means providing infrastructure designed for AI model training and inference as a service. Lee emphasized, "Naver was the fastest company in the world to commercialize Nvidia's supercomputing infrastructure 'SuperPOD' in 2019 and has experience in directly designing and operating ultra-high-performance GPU clusters." He continued, "Based on our global-level GPU operation capabilities, we will enhance AI utilization across industries."
He further explained, "In addition to the GPUaaS business, we are also working on creating vertical AI models tailored to each industry, company, and business based on HyperCLOVA X, as well as developing work applications that can be used internally within companies." Hyundai Motor Company, Samsung Electronics, Korea Hydro & Nuclear Power, and the Bank of Korea are among the organizations using related Naver services.
Regarding the recent formation of a consortium with Samsung SDS to participate in the national AI computing center construction project, Lee expressed confidence in their competitiveness from a risk hedging perspective. He said, "We will not only fulfill our role as a service provider but also leverage our growing internal AI workloads to execute the project seamlessly."
Meanwhile, as awareness of data center management has increased following the fire at the National Information Resources Service, Lee noted that Gak Sejong can operate without interruption even in the event of a failure. Noh Sangmin, Head of Naver Cloud Data Center, said, "In the event of a fire, we operate both fire suppression gas and sprinklers. We also run a significant number of leased IDCs in the Seoul metropolitan area, Chungcheong, and Gyeongnam regions, ensuring rapid recovery through continued investment."
The center is designed with redundant power, cooling, and server operation systems, and has reconfigured uninterruptible power supply (UPS) and power distribution equipment to structurally block the spread of failures. For example, by duplicating the main incoming power lines, stable power is ensured even if one line fails. The power system supplying the server rooms is built with both active and standby lines, so if all lines experience issues, the UPS provides backup power for 15 minutes. During this time, emergency generators are activated to supply power for up to 72 hours.
Opened in November 2023, Gak Sejong is a hyperscale data center utilizing up to 270 MW of power capacity. It is designed as a high-density GPU computing space where AI training and inference occur simultaneously, going beyond the storage and processing functions of traditional internet data centers (IDCs). The second and third phases of expansion are scheduled for completion in 2028 and 2029, respectively.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
