Industry partners launch joint demonstration of remote GPU computing resource provisioning platform.
Mitsubishi Corporation, Nippon Telegraph and Telephone Corporation, NTT Communications Corporation, Morgenrot Inc., and iPark Institute Co., Ltd. have jointly launched a demonstration experiment to remotely provide GPU computing power at Shonan Health Innovation Park and data centers operated by MC Digital Realty, Inc. The initiative aims to support AI drug discovery by utilizing NTT's IOWN APN technology and Morgenrot's virtualization technology to enable multiple tenant companies to simultaneously perform comfortable, flexible, and secure AI analysis. Through this demonstration, the partners aim to establish a safe AI analysis infrastructure construction model that can be used in various industrial clusters and contribute to strengthening Japan's industrial competitiveness.
The pharmaceutical and drug discovery industries face growing demand for more efficient research and development processes in response to competition and challenges with complex diseases. While AI has advanced in areas such as structural analysis of genes and proteins and structural generation of drug candidate compounds, cloud computing adoption has been limited. Key obstacles include lengthy data uploads and downloads for transferring learning and analysis data, as well as security concerns about depositing research data, including privacy information, in public clouds.
The demonstration experiment will connect Shonan iPark, the largest drug discovery research institute by land size in Japan, with the NRT10 data center operated by MC Digital Realty in Inzai City, Chiba Prefecture via high-speed, low-latency lines utilizing IOWN APN technology. By using Morgenrot's virtualization technology to create a private cloud environment within a closed network accessible only to limited users, research data can be analyzed securely without being stored in public clouds. Mitsubishi has installed servers equipped with NVIDIA H100 GPU in the NRT10 data center to provide the computing resources.
The joint verification efforts will assess three key areas. First, the experiment will verify that connecting user bases and remote data centers via high-speed, low-latency lines prevents network performance degradation such as delay and frame loss. Second, it will verify that the analysis infrastructure is suitable for various workloads specific to AI processes in different industries and contributes to business and economic efficiency for participating tenant companies. Third, it will verify that confidential computing using remote GPU computing power is feasible and meets security requirements required by the pharmaceutical and drug discovery industries.
Mitsubishi is providing servers equipped with NVIDIA H100 GPU and introducing NVIDIA BioNeMo to accelerate creation, customization, and deployment of AI models for drug discovery and molecular design, along with NVIDIA NIM inference microservice to assess computing power needs and evaluate economics. NTT is building on its IOWN APN technology to connect sites with end-to-end optical paths offering high speed and low latency. Additionally, NVIDIA provides expertise in the pharmaceutical and drug discovery industries and supports tenant companies in updating on latest industry trends in AI drug discovery and service use cases.