지코어(Gcore)는 4일 서울 시청 더플라자 호텔에서 기자간담회를 개최해 국내 최초 엔비디아 H100 기반 AI 퍼블릭 클라우드 서비스를 한국 시장에 공개했다.
▲Jikor Jeong Hyeon-yong, Korea Branch Manager
NVIDIA Collaboration H100 Acquisition… Future Adoption of H200·GB200
NHN Cloud-Zicore Join Hands… Support for Localization and Customized Services
Zicor is about to open a data center with 320 NVIDIA H100s installed. Zicor has expressed its ambition to expand the number of GPUs to over 1,000 by the end of this year and implement the best AI cloud service that reflects the characteristics of Korean AI customers through a partnership with NHN Cloud.
On the 4th, Gcore held a press conference at the Double Plaza Hotel in Seoul City Hall and unveiled Korea’s first NVIDIA H100-based AI public cloud service to the Korean market. The event began with a congratulatory address by Jacques Flis, the first ambassador of Luxembourg to Korea, followed by an announcement by Jeong Hyeon-yong, CEO of Gcore Korea, and Kim Jin-yong, head of the pre-sales team at Gcore, about the opening of a domestic data center and new services. Kim Dong-hoon, CEO of NHN Cloud, introduced the partnership with Gcore Korea.
■ Zicore's new data center opens in Incheon on April 15 Zicor's new domestic data center, located in Gajwa-dong, Incheon, will open on April 15. As Korea's first GPU-only data center, 40 servers, each equipped with eight NVIDIA H100 GPUs, recognized as high-performance GPUs essential for AI learning, are scheduled to be installed.
Zicore said, “Zicore’s new region was selected in Incheon after consultation with NHN Cloud, considering its proximity to Seoul and future scalability, and entered the bid in the form of colocation.” Hosting global cloud services in a domestic region is expected to result in faster speeds. Zicore aims to secure at least 10-15% of its sales from Korean data centers.
Zicor's H100-based AI public cloud service provides high-performance computing performance by configuring servers with eight H100 GPUs of the SMX5 type in large clusters. President Jeong said, "In order to participate in the domestic AI demand, we will expand the domestic region to over 1,000 GPUs by the end of this year and secure the successor model H200 as well as the recently released GB200." He added, "Through this, we will provide customers with the most cost-effective generative AI computing, thereby serving as a bridgehead for expanding the base of the AI market and advancing into the global market." In the mid to long term, Zicor aims to globally expand its edge AI service specialized in inference.
Zicor's AI cloud service based on NVIDIA H100 features 'InfiniBand NDR (400Gbps)' between GPUs, providing a bandwidth of 3.2Tbps per server. It also provides a large-capacity data processing and high-performance computing environment with 2TB of memory and 112 CPU cores per server. It is supported by power capacity of up to 12KW per server, fast network, and secure design.
“With the H100, which has more than three times the performance of NVIDIA’s A100, customers can benefit from Zcore’s high-performance cloud services without waiting for dozens of weeks or using GPU resources from other countries,” said Kim Jin-yong, head of Zcore’s pre-sales team. “Zcore supports every stage of the AI adoption process, from building proof-of-concepts to training and deployment, with our AI platform, IPU, and NVIDIA GPU-based ML acceleration cloud infrastructure.”
Zicore has its own global IT infrastructure, with a global average response time of 30ms. It has cloud and edge AI know-how based on more than 160 PoPs with a capacity of more than 110Tbps worldwide. With the competitive edge of being able to provide a stable network that connects data and processing, it provides stable cloud services by using global infrastructure along with domestic infrastructure.
■ NHN Cloud, Joint Operation of Stable Data Center
▲ NHN Cloud CEO Dong-Hoon Kim
Zicore jointly operates a data center with NHN Cloud, a Korean cloud company. NHN Cloud plays a major role in localizing Zicore's services and configuring them to suit customers by sharing the needs of domestic AI customers.
NHN Cloud CEO Dong-Hoon Kim said, “NHN Cloud, which has its parent company in Hangame, is a leading CSP operator in the Asia-Pacific region. The meeting between NHN Cloud and Zcore, which possesses open stack-based cloud development and operation technology, is an essential collaboration for securing and building AI data centers, AI edge, 5G, and GPU infrastructure.”
NHN Cloud has built more than 1,000 H100s in its data centers, including Gwangju and Pangyo. It has a national AI data center in Gwangju and provides public cloud services, while its Incheon data center with Zigcore will have general customers.
NHN Cloud CEO Dong-Hoon Kim said, “The data center we will operate with Zicor will provide a quick and rapid response to domestic startups, enterprises, research institutes, finance, manufacturing, and others who need AI resources more publicly.”
Team Leader Kim said, “Through our collaboration with NVIDIA, we have secured enough H100s, and as the AI inference area has recently expanded, we are also exploring the introduction of SAPEON’s NPU ‘X330’, which NHN Cloud is partially introducing. This is being reviewed as applicable to ZigCor’s 160 global edge networks.”