[Edaily Reporter Kwon Oh Seok ] NH INVESTMENT & SECURITIES announced on the 1st that “it is time to increase exposure to companies engaged in the NeoCloud business,” recommending NAVER(035420)(Naver), SKTelecom(017670), NHN(181710), and Navius Group.

Ahn Jae-min, an analyst at NH INVESTMENT & SECURITIES, explained, “As AI (artificial intelligence)-based LLM (large language model) services rapidly permeate our daily lives, the use of GPUs (graphics processing units) for training and inference is expected to grow rapidly.” He added, “Consequently, we anticipate growth for NeoCloud companies—next-generation GPU-dedicated cloud service providers specialized in training large-scale AI models and processing high-performance inference workloads.”
The NeoCloud business is difficult to expand based solely on the will to build data centers. Since it requires large-scale land, power, and water resources, as well as approvals from the government and local authorities, it is not possible to build one whenever and wherever one wishes. Analyst Ahn explained, “Technical expertise in building and utilizing large-scale GPU clusters is also crucial. The Nebius Group is a company with experience in various platform businesses, including internet search engines, while CoreWeave and iRen are also companies that have worked with GPUs in the past through crypto mining,” he elaborated.
He added, “Purchasing GPUs requires substantial financial resources and the ability to secure GPUs from NVIDIA or AMD, as well as tenants who will occupy the AI data center once it is completed,” noting that “timing is also crucial. Even if investment to build a data center begins now, construction won’t be completed until 2028–2029 at the earliest. Given this, only companies that jump into the business quickly from now on will lead the market after 2028,” he concluded.
He further emphasized, “Since all these elements must be harmoniously integrated, the NeoCloud business has higher barriers to entry than one might think. If a company decides to build a data center in 2028—when the revenue of companies currently constructing facilities will become fully apparent—it will already be too late.” He added, “The growth of large language models (LLMs) is expected to lead to structural expansion in the NeoCloud market, which provides AI computing infrastructure.”
Furthermore, the firm noted, “As the scope of AI applications expands from simple question-and-answer systems to Agent AI services capable of complex reasoning and task execution, the demand for training to develop models and for inference required to operate actual services is increasing in tandem.” It added, “Consequently, not only are the number of AI users and usage frequency rising, but the computational load required per task is also increasing simultaneously, heightening the importance of securing stable computing infrastructure.”