Through this partnership, GMO Internet will provide Sakana AI with an environment that enables them to focus entirely on research and development by offering world-class AI computing infrastructure and robust technical support. GMO Internet is fully committed to supporting the future where Sakana AI’s breakthrough technologies lead the global AI landscape.
(※1）Based on an internal survey as of November 22, 2024, among commercially available domestic cloud services. See: GMO GPU Cloud ranked in TOP500 Supercomputer Rankings（URL：https://group.gmo/news/article/9266/）
【About Sakana AI】
Founded in 2023 by David Ha (CEO), Llion Jones (CTO), and Ren Ito (COO), Sakana AI is a Tokyo-based R&D company at the forefront of global AI research. The company is developing world-class technologies such as “AI Scientist,” “Evolutionary Model Merging,” “Continuous Thought Machines,” and “Darwin-Gödel Machines,” and is constantly pushing the boundaries of AI innovation.
In March 2025, Sakana AI also announced the formation of a Business Development Division to collaborate with some of Japan’s largest enterprises and public-sector organizations. The company is actively developing AI solutions tailored to Japan’s unique needs, particularly in the fields of finance and defense, while also working toward the democratization of AI.
【Key Evaluation Points of “GMO GPU Cloud”】
１．Comprehensive Technology Stack Optimized for Large-Scale Distributed Training
GMO GPU Cloud offers a complete environment for AI development out of the box, featuring high-speed storage, high-bandwidth interconnects, and a wide range of essential libraries (e.g., NCCL, CUDA, HPC-X(*2）) in multiple versions—reducing setup time and enabling immediate deployment.
２．Prompt and Comprehensive Technical Support
GMO Internet minimizes GPU idle time by responding quickly to technical inquiries, ensuring smooth continuity in research and development.
３．Advanced Monitoring Environment
A Grafana-based (*3) dashboard enables real-time visualization of GPU server operations, including metrics such as temperature and power consumption. This allows researchers to constantly monitor and optimize learning efficiency—maximizing resource utilization for an ideal development environment.
(※2）Product names mentioned are registered trademarks or trademarks of their respective companies.
(※3）Grafana: An open-source data visualization platform that intuitively visualizes GPU server usage and job status.
Comment from David Ha, CEO of Sakana AI
“In large-scale distributed training, performance is not just about powerful GPUs. The completeness of the software stack—including communication, storage, and job management—significantly affects research speed. GMO GPU Cloud excels in this balance and came pre-equipped with the exact environment we needed, allowing us to focus on our core research without spending time on setup.
The knowledge and responsiveness of the engineers supporting operations is extremely reliable. They quickly isolate and resolve issues, which lets us proceed with large-scale experiments with peace of mind.
I also highly value the monitoring infrastructure using tools like Grafana, which allows us to view GPU metrics in real time and constantly optimize model training efficiency.
I hope GMO GPU Cloud continues to evolve as Japan’s leading AI infrastructure and contributes to the growth of both research and industry.”
【Features of GMO GPU Cloud】
GMO GPU Cloud is a high-performance GPU cloud service that combines NVIDIA H200 Tensor Core GPUs, Japan’s first commercial deployment of the high-speed network NVIDIA Spectrum-X, and high-speed storage. It delivers the fastest performance among domestic commercial cloud services (as of November 22, 2024) in multi-node configurations.
In November 2024, GMO GPU Cloud ranked 37th in the world and 6th in Japan in the TOP500 list of the world’s fastest supercomputers—marking the first time it appeared in the rankings as a commercial cloud service.
Thanks to its exceptional performance and stability in multi-node configurations, GMO GPU Cloud provides an optimal environment for developing large language models (LLMs) and multimodal AI.
In addition, by the end of 2025, GMO Internet plans to launch a new cloud service powered by NVIDIA’s next-generation Blackwell Ultra GPU installed in NVIDIA HGX B300 (hereinafter “NVIDIA B300”).
Through this service, GMO Internet offers optimized infrastructure and customizable computing environments tailored to each customer’s workload. This contributes to shortening development cycles and reducing costs, thereby supporting the advancement of Japan’s AI industry.
（※4）Based on internal survey as of November 22, 2024. https://www.gmo.jp/news/article/9266/)
【Future Outlook】
AI technologies are transforming various industries—such as healthcare, manufacturing, agriculture, and energy—and are becoming essential in solving social issues such as aging populations, labor shortages, and depopulation in rural areas.
GMO Internet will continue to enhance the functionality and performance of GMO GPU Cloud and provide high-quality computing platforms to support AI companies and research institutions, including Sakana AI. Through this, the company aims to contribute to building a sustainable society and strengthening Japan’s international competitiveness in AI.
【About GMO Internet, Inc.】
As of January 1, 2025, GMO Internet, Inc. launched a new structure that combines the strengths of the Internet Infrastructure and Online Advertising and Media segments of the GMO Internet Group.
By leveraging the robust earnings foundation of the Internet Infrastructure business and the unique strengths of the Online Advertising and Media business, and under the corporate slogan “Internet for Everyone,” GMO Internet is committed to delivering “smiles” and “inspiration” to everyone it touches and creating new value for the future through AI.
