As part of the deal, Oracle is also providing immediate access to AI computing resources built on NVIDIA’s Hopper and Blackwell architectures. The goal is to get more computing power into scientists’ hands faster while the larger systems are built.

“Winning the AI race requires new and creative partnerships,” said Energy Secretary Chris Wright, calling the collaboration a “commonsense approach” that brings private innovation directly into DOE’s mission. NVIDIA founder and CEO Jensen Huang described the project as “an AI factory that will serve as America’s engine for discovery.”

The DOE has a long history of partnering with industry on supercomputing projects, but this marks one of its most ambitious AI pushes yet. Officials say the partnership reflects a new model for shared investment between government and industry, one designed to bring new computing capacity online quickly and cement U.S. leadership in artificial intelligence.

According to Paul Kearns, director of Argonne National Laboratory, the new systems will “accelerate a broad set of scientific AI workflows” and connect seamlessly to facilities like Argonne’s Advanced Photon Source. Researchers will use the infrastructure to develop large-scale reasoning models using NVIDIA’s Megatron-Core and TensorRT platforms, forming the foundation for next-generation scientific AI.

Together, Solstice, Equinox, and Oracle’s interim resources mark what the DOE calls a three-phase strategy to shrink the gap between idea and discovery. By merging national lab expertise with industry’s most powerful AI systems, the partnership aims to keep the United States at the forefront of innovation and redefine what “supercomputing” means in the age of artificial intelligence.