NVIDIA announced a strategic cooperation with the U.S. Department of Energy (DOE) to build a total of seven ExaFLOPS-level AI supercomputers, two of which—Solstice and Equinox—will be built by Oracle and deployed at Argonne National Laboratory.
According to NVIDIA’s official statement, the Solstice system will be equipped with a record-breaking 100,000 Blackwell GPUs with a total power of 200MW to support the Department of Energy’s AI applications in the fields of defense security, science and energy; another Equinox system will be equipped with 10,000 Blackwell GPUs and is expected to go online in 2026. The two systems are interconnected through NVIDIA network technology, providing a combined computing performance of up to 2,200 FP4 ExaFLOPS, and are expected to become the most efficient AI platform in the U.S. scientific research system.
NVIDIA founder and CEO Jensen Huang said: "AI is the most powerful technology of our time, and science is its biggest frontier. Working with Oracle, we are building the largest AI supercomputer in the United States for the Department of Energy, allowing researchers to use the most advanced infrastructure to drive progress in fields such as medical, materials and energy."
On the other hand, the U.S. Department of Energy has previously announced that it has signed a $1 billion cooperation agreement with AMD to build two supercomputers, "Lux" and "Discovery" at Oak Ridge National Laboratory (ORNL), equipped with the latest MI355X AI chips. This shows that the Department of Energy is simultaneously promoting cooperation with multiple industries to comprehensively expand the computing power of AI scientific research in the United States.
Nvidia and partners to build seven AI supercomputers for the U.S. gov’t with over 100,000 Blackwell GPUs —combined performance of 2,200 ExaFLOPS of compute