Catenaa, Tuesday, October 28, 2025- Nvidia announced on Tuesday that it’s teaming up with the US Department of Energy to develop seven AI supercomputers at two major US labs.
The Argonne National Laboratory and the Los Alamos National Laboratory will see the first phase of the rollout with Nvidia and Oracle partnering to build two new supercomputers called Equinox and Solstice.
Equinox will run on 10,000 Nvidia Blackwell GPUs, while Solstice will use 200 megawatts worth of chips.
Both systems, Nvidia says, will be used to build one trillion-parameter AI models, simulation, and scientific computing.
Additional Argonne supercomputers will include Tara, Minerva, and Janus. The Los Alamos supercomputers, called Mission and Vision, will be built by HPE and run on Nvidia’s upcoming Vera Rubin superchips.
Mission, the company said, will power classified applications for the National Nuclear Security Agency, while Vision is built for unclassified AI applications.
Nvidia also said it’s working with nine US national labs and a host of additional companies on quantum computing technologies via its new NVQLink technology, which pairs a classic GPU supercomputer to a quantum computer to help address quantum error correction.
Rather than the bits represented by ones and zeros that classic computers use, quantum computers use qubits. But qubits, unlike standard bits, are susceptible to outside interference, which can impact the results of quantum calculations.
That has made it difficult to build a practical quantum computer that can pump out the correct answer to an equation without fail. Quantum error correction is meant to address errors caused by that interference.
In addition to the AI supercomputer, Nvidia is developing an all-American, AI-native wireless stack for 6G connectivity. The company said it’s working with US firms, including T-Mobile and Cisco, to ensure that America can deploy 6G networks built using US technology.
