NHN计划明年新增至少30兆瓦AI数据中心容量
NHN在第二季度财报电话会议上表示，计划到明年新增至少30兆瓦AI数据中心容量，以加速AI基础设施业务扩张。该公司正积极投资建设数据中心，以满足AI计算需求。此举将增加对服务器、存储及网络设备等电子元器件的需求，对供应链产生积极影响。
原文深度解读
NHN在8月11日第二季度财报电话会议上宣布，计划到明年新增至少30兆瓦（MW）AI数据中心容量，以加速AI基础设施业务扩张。其中20MW通过租赁设施获得，10MW通过自建数据中心（预计明年下半年运营）。NHN Cloud的GPU即服务（GPUaaS）业务（基于Nvidia B200 GPU）是主要增长动力，第二季度营收1598亿韩元，同比增长85%。公司还计划在日本获取数据中心容量，并推出AI全栈品牌Factory X。此举将增加对服务器、存储、网络设备及液冷系统等电子元器件的需求，对供应链产生积极影响。
- NHN计划到明年新增至少30MW AI数据中心容量，其中20MW通过租赁设施，10MW通过自建数据中心（预计明年下半年开始运营）。
- 新增容量约等于NHN现有数据中心（杨平、板桥、光州）的总容量。
- NHN Cloud第二季度营收1598亿韩元，同比增长85%，环比增长49%，主要得益于基于Nvidia B200 GPU的GPUaaS业务，该业务于3月底在首尔杨平地区全面商业运营。
- NHN Cloud的NIPA-CL1 GPU集群在6月底TOP500超算排名中韩国第一、全球第20，Rmax性能137.4 PF；NIPA-CL2韩国第四、全球第40，性能68.42 PF。
- NHN Cloud正在讨论新的长期合同，并主要瞄准海外市场而非国内客户。
- NHN Cloud正在日本获取数据中心容量，以准备为全球客户扩展服务。
供应链影响
- 新增30MW数据中心容量可能增加对服务器、存储、网络设备及液冷系统等电子元器件的需求，具体取决于建设进度和租赁设施的设备配置。
- 基于Nvidia B200 GPU的GPUaaS业务扩展可能带动对高端GPU、相关服务器组件（如电源、散热）的需求，但受全球GPU供应和出口政策影响。
- 液冷服务器的部署可能增加对液冷组件（如冷却液、管路、泵）的需求，但取决于技术路线和供应商选择。
- 日本数据中心容量的获取可能涉及当地供应链布局，可能影响区域电子元器件采购和物流，但具体取决于合作模式。
- 公司计划分阶段扩展基础设施，可能使元器件采购呈现阶段性增长，但受市场需求和合同签订情况影响。
系统解读，仅作为判断线索，不构成备货、出货、涨价或投资建议。
NHN表示，计划到明年新增至少30兆瓦（MW）的AI数据中心容量，以加速其AI基础设施业务的扩张。
在8月11日的第二季度财报发布和电话会议中，NHN将其技术业务——尤其是NHN Cloud的AI图形处理器（GPU）业务——视为盈利改善的最大贡献者，并概述了其基础设施扩张战略。
在计划的30MW中，20MW将通过租赁设施获得，10MW通过公司自建数据中心获得，该数据中心计划于明年下半年开始运营。此次扩张规模大致相当于NHN现有数据中心（杨平、板桥和光州）的总容量。
NHN的技术业务在第二季度实现营收1598亿韩元。NHN Cloud的营收同比增长85%，环比增长49%，这得益于其基于Nvidia B200 GPU的GPU即服务（GPUaaS）业务，该业务在首尔杨平地区于3月底全面商业运营后开始贡献收入。
NHN Cloud首席执行官Kim Dong-hoon在5月介绍公司AI全栈品牌Factory X。（照片：THE ELEC）
NHN Cloud表示，正在就新的长期合同进行讨论，这使得明年的营收前景比此前预期更为乐观。公司补充称，其GPUaaS业务主要瞄准海外市场而非国内客户。
“我们看到海外公司的需求日益增长，一些客户预计将在近期签约，”NHN Cloud首席执行官Kim Dong-hoon表示。
Kim补充说，提前确保大规模数据中心基础设施对于快速响应未来增长机会至关重要。
“在GPU业务中，在客户需要时立即提供GPU的能力极为重要，”他说。
据Kim称，NHN Cloud还在努力获取日本的数据中心容量，以为全球客户扩展服务做准备。
Kim将NHN Cloud描述为韩国部署B200液冷服务器最多的运营商，并表示公司积累的运营经验使其区别于其他GPUaaS提供商（即neoclouds）。
他还在电话会议中指出，NHN的NIPA-CL1 GPU集群在6月底发布的最新TOP500超算排名中位列韩国第一、全球第20。该系统实现了137.4千万亿次浮点运算（PF）的Rmax性能，而NIPA-CL2在国内排名第四、全球第40，性能为68.42 PF。
Kim表示，公司的基础设施和积累的GPU运营经验构成了NHN于5月推出的AI全栈品牌Factory X的基础。Factory X不仅仅是提供GPU，还提供集成平台，包括GPU管理能力和支持AI代理环境的服务。
NHN还表示，通过利用其国内GPUaaS业务获得的运营经验，正在持续评估和寻求额外的数据中心空间，以满足长期AI需求。公司计划随着需求增长分阶段扩展基础设施。
首席财务官Ahn Hyun-sik表示，NHN的战略不同于全球超大规模云提供商，它通过较小的增量扩展，而不是承担较大的短期风险，并补充说这种方法预计将支持云业务的利润率。
Ahn表示，公司更重视自营GPUaaS业务的盈利能力，而不是政府项目，在政府项目中公司获得GPU使用权但仍承担运营成本。他补充说，由于AI数据中心成本持续上升且市场状况高度波动，利润率难以量化，但NHN打算在谨慎管理风险的同时保持高盈利能力。
Ahn还表示，随着杨平地区开始产生可观的收入，第二季度折旧费用显著增加，预计第三季度将保持类似水平。然而，由于公司计划额外购买GPU和进一步租赁数据中心，折旧将保持波动。他表示，较高的折旧应伴随着相应的收入和利润增长。
关于NHN Cloud的多重上市和潜在首次公开募股（IPO）的监管问题，公司采取了谨慎态度。Ahn表示，NHN正在与金融投资者IMM Investment进行讨论，但任何决定都需要股东同意。
他补充说，必须首先建立GPUaaS业务的稳定盈利能力，以支持NHN Cloud的估值，并表明一旦业务达到足够规模并产生可持续的营业利润，将考虑IPO。
NHN said it plans to secure at least 30 megawatts (MW) of additional AI data center capacity by next year as it accelerates the expansion of its AI infrastructure business.
During its second-quarter earnings announcement and conference call on Aug. 11, NHN identified its technology business—particularly NHN Cloud's AI graphics processing unit (GPU) business—as the largest contributor to its earnings improvement and outlined its infrastructure expansion strategy.
Of the planned 30 MW, 20 MW will be secured through leased facilities and 10 MW through a company-owned data center, which is scheduled to begin operations in the second half of next year. The planned expansion is roughly equivalent to the combined capacity of NHN's existing data centers in Yangpyeong, Pangyo and Gwangju.
NHN's technology business generated 159.8 billion won in revenue during the second quarter. Revenue at NHN Cloud increased 85% from a year earlier and 49% from the previous quarter, reflecting contributions from its GPU-as-a-service (GPUaaS) business powered by Nvidia B200 GPUs after its Seoul Yangpyeong region entered full-scale commercial operation in late March.
Kim Dong-hoon, Chief Executive Officer of NHN Cloud, introduces the company's AI full-stack brand, Factory X, in May. (Photo: THE ELEC)
NHN Cloud said discussions are underway for new long-term contracts, making next year's revenue outlook more favorable than previously expected. The company added that its GPUaaS business is primarily targeting overseas markets rather than domestic customers.
"We are seeing increasing demand from overseas companies, and some customers are expected to sign contracts in the near future," NHN Cloud Chief Executive Officer Kim Dong-hoon said.
Kim added that securing large-scale data center infrastructure in advance is essential to respond quickly to future growth opportunities.
"In the GPU business, the ability to provide GPUs immediately when customers need them is extremely important," he said.
According to Kim, NHN Cloud is also working to secure data center capacity in Japan as it prepares to expand services for global customers.
Kim described NHN Cloud as the operator with the largest deployment of B200 liquid-cooled servers in South Korea and said the company's accumulated operational expertise differentiates it from other GPUaaS providers, or neoclouds.
He also noted during the conference call that NHN's NIPA-CL1 GPU cluster ranked first in South Korea and 20th globally in the latest TOP500 supercomputer rankings released at the end of June. The system achieved an Rmax performance of 137.4 petaflops (PF), while NIPA-CL2 ranked fourth domestically and 40th globally with 68.42 PF.
Kim said the company's infrastructure and accumulated GPU operating expertise form the foundation of Factory X, NHN's AI full-stack brand launched in May. Rather than simply supplying GPUs, Factory X provides an integrated platform that includes GPU management capabilities and services supporting AI agent environments.
NHN also said it is continuously reviewing and pursuing additional data center space to meet long-term AI demand by leveraging operational expertise gained through its domestic GPUaaS business. The company plans to expand infrastructure in phases as demand grows.
Chief Financial Officer Ahn Hyun-sik said NHN's strategy differs from global hyperscale cloud providers by expanding in smaller increments rather than assuming large short-term risks, adding that this approach is expected to support cloud business margins.
Ahn said the company places greater emphasis on the profitability of its self-operated GPUaaS business than on government projects, where it receives GPU usage rights but still bears operating costs. He added that margins are difficult to quantify because AI data center costs continue to rise and market conditions remain highly volatile, but NHN intends to maintain high profitability while carefully managing risk.
Ahn also said depreciation expenses increased significantly in the second quarter as the Yangpyeong region began generating meaningful revenue and are expected to remain at a similar level in the third quarter. However, depreciation will remain fluid as the company plans additional GPU purchases and further data center leases. He said higher depreciation should be accompanied by corresponding revenue and profit growth.
Regarding regulatory issues surrounding multiple listings and a potential initial public offering (IPO) for NHN Cloud, the company adopted a cautious stance. Ahn said NHN is in discussions with financial investor IMM Investment, but any decision will require agreement from shareholders.
He added that stable profitability from the GPUaaS business must first be established to support NHN Cloud's valuation, indicating that an IPO will be considered once the business reaches sufficient scale and generates sustainable operating profit.