(※1) Bare metal environment: An environment that directly utilizes physical servers without virtualization
【Background of Service Provision】
In recent years, demand for large-scale computational resources capable of advanced arithmetic processing has been increasing, driven by companies' development of proprietary LLMs (Large Language Models) and rapid evolution of AI and robotics in industrial sectors. To meet demands such as AI model scaling, real-time inference, and support for diverse generation formats, cloud infrastructure must also continuously update as a "dynamic technology foundation," making such updates the key to technological innovation.
GMO Internet received certification from the Ministry of Economy, Trade and Industry on April 15, 2024, for its supply security plan for specific critical materials "Cloud Program"(※2) under the Economic Security Promotion Act, and has been providing services adopting NVIDIA H200 GPU at the fastest pace in Japan as a cloud service provider since November 22, 2024(※3) .
Considering these market environments and technological requirements, we decided to review our additional investment plan for NVIDIA H200 announced on May 14, 2024, and switch to the higher-performance next-generation GPU " NVIDIA B300." This will enable "GMO GPU Cloud" to strengthen its response to the further computational performance, scalability, and low-latency(※4)
communication required for future AI development and operation, deploying as a more advanced infrastructure platform.
(※2) Reference: Ensuring Stable Supply of Specific Critical Materials under the Economic Security Promotion Act
https://www.meti.go.jp/policy/economy/economic_security/cloud/index.html
(※3) GMO Internet Group Launches “GMO GPU Cloud,” Ranked in the TOP500 Supercomputer Listhttps://internet.gmo/en/news/article/27/
(※4) Low-latency communication: A communication method that enables rapid processing with minimal delay in data transmission and reception
■ Key Features of "GMO GPU Cloud (NVIDIA B300)"
"GMO GPU Cloud" provides a high-performance computing infrastructure optimized for next-generation generative AI training and AI inference processing by offering NVIDIA-recommended configurations including Spectrum-X and storage, along with NVIDIA AI Enterprise.
The NVIDIA B300, equipped with 288GB HBM3e memory, 5th generation Tensor Core, and Dual-die structure, significantly accelerates inference processing for large language models and dramatically improves AI development efficiency.
■ Use Cases
• High-speed training and fine-tuning of large language models
• Acceleration of AI inference processing
• Training of computer vision models using large-scale datasets
【About "GMO GPU Cloud"】
(URL: https://gpucloud.gmo/)
(URL: https://gpucloud.gmo/)
The fastest-class GPU cloud service in Japan, "GMO GPU Cloud," began service in November 2024 adopting high-performance "H200 GPU." The newly introduced " NVIDIA B300" with NVIDIA's latest architecture "Blackwell (NVIDIA B300)" is a GPU designed for building ultra-large-scale AI factories, equipped with 5th generation Tensor Core supporting "FP4, (※5)" HBM3e memory(※6) reaching 2.3TB, ultra-high-density Dual-die structure, and high-speed communication via 5th generation NVIDIA NVLink(※7). This dramatically reduces large language model training time compared to the conventional H200, further significantly improving AI development efficiency.
Through this service, GMO Internet will provide enterprises and research institutions working in generative AI with high-performance computing environments that require no infrastructure tuning, contributing to reduced development time and costs while promoting the growth of Japan's AI industry.
(※5) "FP4": A new computational mechanism specialized for AI. It enables faster processing with less information than before, making AI operations faster and more energy-efficient.
(※6) "HBM3e": The latest high-speed memory (temporary data storage location). It can reduce processing time even for large AI models.
(※7) "NVLink": A component of high-speed multi-GPU communication that sends large datasets to models in a short time and enables high-speed data exchange between GPUs.
【Future Development】
GMO Internet will contribute to technological innovation in the rapidly evolving AI and robotics fields through its AI infrastructure strategy centered on "GMO GPU Cloud." We will continue to provide the latest AI computing infrastructure and flexible cloud environment construction according to customer needs, serving as an essential cloud service for Japan's AI industry and contributing to the creation of AI innovation originating from Japan.
■ Past Reference Releases
April 19, 2024:
GMO Internet Group to Provide Japan's Fastest GPU Cloud Service for Generative AI Adopting NVIDIA H200 Tensor Core GPU https://www.gmo.jp/news/article/8933/
June 11, 2024:
GMO Internet Group First Among Domestic Cloud Providers to Adopt NVIDIA Spectrum-X for Generative AI GPU Cloud Service https://www.gmo.jp/news/article/9005/
August 29, 2024:
GMO Internet Group "GPU Cloud Usage Survey" - Domestic Usage Rate Only 5.4%, Approximately 90% Use Overseas Services https://www.gmo.jp/news/article/9133/
September 26, 2024:
GMO Internet Group Demonstrates Performance of "NVIDIA H200 GPU" Equipped Environment https://www.gmo.jp/news/article/9164/
November 13, 2024:
GMO Internet Group Explains Infrastructure and Security for AI & Robotics Era at "NVIDIA AI Summit" https://www.gmo.jp/news/article/9233/
November 19, 2024:
GMO Internet Group's "GMO GPU Cloud" First Ranked in Global Supercomputer Ranking TOP500 https://internet.gmo/news/article/26/
November 22, 2024:
GMO Internet Group Begins Providing "GMO GPU Cloud" Ranked in Supercomputer TOP500 https://internet.gmo/news/article/27/
May 14, 2025:
GMO Internet Decides Additional Investment in "GMO GPU Cloud" https://internet.gmo/news/article/46/
Jun 11, 2025: GMO Internet’s “GMO GPU Cloud”
Ranked 34th in the world and 1st in Japan in the global power efficiency ranking “Green500” https://internet.gmo/en/news/article/50/
【About GMO Internet, Inc.】
GMO Internet will contribute to technological innovation in the rapidly evolving AI and robotics fields through its AI infrastructure strategy centered on "GMO GPU Cloud." Going forward, by providing the latest AI computing infrastructure and constructing flexible cloud environments tailored to customer needs, we will serve as an indispensable cloud service for Japan's AI industry and contribute to the creation of AI innovation originating from Japan.
