Redazione RHC : 30 October 2025 07:01
The U.S. Department of Energy (DOE) has entered into a strategic collaboration with Nvidia and Oracle to build seven next-generation AI-powered supercomputers , set to revolutionize scientific research and the development of intelligent agents.
Two of these systems will be installed at Argonne National Laboratory in Illinois, which will house the largest AI supercomputing infrastructure ever built by the Department of Energy.
At NVIDIA’s GPU Technology Conference, CEO Jensen Huang announced Project Solstice , described as the largest AI supercomputer ever built for the DOE. The system will be equipped with 100,000 Blackwell GPUs and built in collaboration with Oracle and Argonne Lab.
Solstice will be joined by Equinox , a second supercomputer under development with 10,000 Blackwell GPUs , also hosted at Argonne . The interconnection between the two systems will allow for a combined computing power of 2,200 exaFLOPs , marking a new milestone in AI computing.
Huang emphasized that this infrastructure “will become America’s engine of discovery,” providing researchers with advanced tools to accelerate studies in fields ranging from medicine to materials science.
In addition to providing increased computing power for scientific experiments, Solstice and Equinox will be an integral part of Argonne Laboratory’s “Developing Agent Scientists” program. This initiative aims to introduce autonomous agent-based artificial intelligence systems into research, with the goal of increasing productivity and fostering new discoveries over the next decade.
According to initial forecasts, Equinox should enter into operation next year , while an official date for the completion of Solstice has not yet been communicated.
In parallel, NVIDIA announced a new collaboration with Palantir Technologies , which will integrate NVIDIA accelerated computing technologies— including CUDA-X libraries and Nemotron AI models—into its AI platform. The goal is to increase data processing speed and introduce customizable AI agents within Palantir systems.
The combination of the two technologies has already been tested by Lowe’s , the US home improvement chain, which used a digital twin of its supply network to optimize logistics flows through artificial intelligence models.
Argonne Lab also plans to expand its infrastructure with three additional NVIDIA GPU-based systems: Tara , Minerva , and Janus . These supercomputers will be made available to the scientific community to foster shared access to AI computing resources.
Los Alamos National Laboratory will also be part of the initiative and will receive two new systems based on the NVIDIA Vera Rubin platform. The first, called Vision , will be operational in 2027 and dedicated to unclassified research projects—from national security to materials science to biomedical research. The second, called Mission , will be dedicated to classified workloads, replacing the Crossroads supercomputer, which began operations in 2023.
” The Mission and Vision systems represent a crucial investment in our scientific capabilities and national security ,” said Tom Mason, director of Los Alamos Laboratory. “They are designed for supercomputing in the AI era.”
Redazione