Mitsui and NVIDIA announced the Tokyo-1 project during the spring GTC 2023 conference. It is claimed to be the world’s first generative AI supercomputer designed specifically for the pharmaceutical industry. The capacity of the new system will be provided to Japanese customers, including pharmaceutical organizations and start-ups. The HPC complex will help accelerate the development of advanced drugs through the use of AI. Customers will also be able to run large AI models on Tokyo-1 using NVIDIA BioNeMo software and services.
At the initial stage, the supercomputer will combine 16 NVIDIA DGX H100 nodes, each of which will receive eight NVIDIA H100 accelerators. The system will be run by Xeureka, a Mitsui subsidiary specializing in AI-assisted drug development. Over time, the complex will include additional computing nodes, which will increase its performance. Commissioning of the supercomputer is scheduled for the end of 2023.
Image Source: NVIDIA
AI models based on Tokyo-1 will help in the creation of new molecular structures for drugs. The initiative will also include workshops and training courses on the application of accelerated computing in the pharmaceutical field. Customers will be able to access a dedicated server on the Tokyo-1 platform, as well as take advantage of technical support from Xeureka and NVIDIA. Customers will be able to purchase Xeureka software solutions for molecular dynamics, quantum chemistry and other calculations.
It is noted that using NVIDIA BioNeMo, researchers will be able to scale AI models to millions and billions of parameters in various applications, including protein structure prediction. Major Japanese pharmaceutical companies, including Astellas Pharma, Daiichi-Sankyo and Ono Pharmaceutical, have already announced their intention to use Tokyo-1 in their projects.
If you notice an error, select it with the mouse and press CTRL + ENTER. | Can you write better? We are always glad to new authors.
Source: