NVIDIA announced a broad collaboration with the U.S. Department of Energy (DOE)'s national laboratories and several American companies to build the nation's AI infrastructure. NVIDIA founder and CEO Jensen Huang described this as "the dawn of the AI industrial revolution" and called it America's "Apollo moment of this generation."
Huang emphasized, “The next wave of invention, discovery, and progress will depend on our nation’s ability to scale up its AI infrastructure. Together, we are building the most advanced AI infrastructure ever.”
Accelerating National Laboratory Science: Seven New Systems Introduced at Argonne and Los Alamos National Laboratories
NVIDIA will provide AI infrastructure to the U.S. Department of Energy’s Argonne National Laboratory and Los Alamos National Laboratory (LANL) to accelerate the deployment of seven new systems:
Solstice (Argonne National Laboratory):
• Built in partnership with Oracle and the U.S. Department of Energy, it will become the U.S. Department of Energy’s largest AI scientific research supercomputer.
• Powered by a record-breaking 10 NVIDIA Blackwell GPUs.
Equinox (Argonne National Laboratory):
• Equipped with 1 NVIDIA Blackwell GPUs, it is expected to be online in 2026.
• Solstice and Equinox will be interconnected via NVIDIA networking technology, providing a total of 2200 exaflops of AI computing power.
Tara, Minerva, Janus (Argonne National Laboratory):
The three NVIDIA-based systems will expand access to AI-driven computing resources for researchers nationwide.
Mission & Vision (Los Alamos National Laboratory):
• Built by HPE, it will use the NVIDIA Vera Rubin platform and Quantum-X800 InfiniBand network architecture.
• Mission is the fifth-generation Advanced Technology System (ATS5) of the U.S. Nuclear Security Administration (NNSA), used for classified applications, and is expected to be operational by the end of 2027.
• Vision is based on the existing Venado supercomputer at Los Alamos National Laboratory and is used for unclassified research.
NVIDIA AI Factory Research Center and Omniverse DSX Gigawatt-Scale Blueprint
NVIDIA also announced that it will establish an AI Factory research center within the Digital Realty data center in Virginia.
• Hardware:Adopting NVIDIA Vera Rubin computing platform.
• aims:Accelerate breakthroughs in generative AI, scientific computing, and advanced manufacturing, and serve as the foundation for research in digital twins and large-scale simulations.
Omniverse DSX Blueprint:
This center will lay the foundation for NVIDIA Omniverse DSX, which will use the NVIDIA Omniverse library to plan a blueprint for multi-generational, gigawatt (GW)-scale AI Factory construction. The goal is to integrate virtual and physical systems to create smart facility models that continuously optimize performance, energy efficiency, and sustainability.
NVIDIA is developing Omniverse DSX with several partners, including:
• Engineering and Construction Partners:Bechtel, Jacobs (integrating digital twins into designs)
• Power, Cooling & Energy Partners:Eaton, GE Vernova, Hitachi, Mitsubishi Electric, Schneider Electric, Siemens, Tesla, Trane, Vertiv (establishing grid interaction models, liquid cooling systems, etc.)
• Software and Agent-Based AI Partners:Cadence, Emerald AI, Phaidra, PTC, Siemens, Switch (building digital twin solutions, AI agent optimization of power loads, etc.)
Expanding U.S. AI Infrastructure Investment: System Makers, Cloud Providers, and Model Builders Respond
Several leading US companies also announced expanded investments in NVIDIA AI infrastructure:
System Manufacturer:
Cisco, Dell Technologies, HPE, and Supermicro will integrate NVIDIA GPUs and AI software to launch full-stack systems, including the NVIDIA AI Factory for Government reference design.
Cisco will also launch the Nexus N9100 switch series powered by the NVIDIA Spectrum-X Ethernet switch chip and provide an NVIDIA Cloud Partner-compatible AI Factory reference architecture based on this switch.
Cloud providers and model builders:
• Akamai:Launched Akamai Inference Cloud, a distributed platform powered by NVIDIA RTX PRO servers.
• CoreWeave:CoreWeave Federal is established to provide FedRAMP-compliant NVIDIA GPU AI cloud services to the U.S. government.
• Global AI (NCP):The purchase of 128 NVIDIA GB300 NVL72 racks (over 9000 GPUs) will become the largest GB300 NVL72 deployment in New York.
• Google Cloud:Provide A4X Max (GB300 NVL72) and G4 (RTX PRO 6000 Blackwell) VMs, and import the Blackwell platform into Google Distributed Cloud.
• Lambda:Establishing a new 100+ MW AI Factory in Kansas City, Missouri, initially deploying more than 1 NVIDIA GB300 NVL72 GPUs.
• Microsoft:NVIDIA RTX PRO 6000 Blackwell GPUs are available on Azure. We recently announced the deployment of a large-scale GB300 NVL72 cluster for OpenAI and will add support for RTX GPUs on Azure Local.
• Oracle:Recently launched OCI Zettascale10, using NVIDIA AI infrastructure.
• Together AI:In partnership with 5C, we have already operated the B200 AI Factory in Maryland and will soon open the new GB200/GB300 factory in Tennessee.
• xAI: is building its massive Colossus 2 data center in Memphis, Tennessee, which will house more than 50 NVIDIA GPUs.
Enterprise AI Factory:
• Lilly: Build the pharmaceutical industry's most powerful AI Factory, using NVIDIA DGX SuperPOD powered by DGX B300, Spectrum-X networking, and Mission Control software to develop large-scale biomedical foundational models.
• Mayo Clinic: Establishing an AI Factory powered by DGX SuperPOD and Mission Control powered by DGX B200 to leverage its vast database of medical records and pathology images to advance medical research, digital pathology, and personalized medicine.
