Key Details

The U.S. Department of Energy (DOE) and National Nuclear Security Administration (NNSA) have launched a next-generation supercomputer development project aimed at strengthening national security, artificial intelligence (AI), and scientific research. This initiative, totaling $370 million (approximately 514.3 billion KRW), will proceed through collaboration with major technology companies including NVIDIA, HPE, and Oracle.

Los Alamos National Laboratory (LANL) will develop the Mission and Vision supercomputers in partnership with HPE and NVIDIA. Argonne National Laboratory (ANL) will build the Solstice and Equinox systems together with NVIDIA and Oracle.

Los Alamos Laboratory's Mission & Vision Systems

The Mission supercomputer aims to assess and improve nuclear security without nuclear testing. Boasting processing speeds 4 times faster than the existing Crossroads system, it will provide advanced AI, modeling, and simulation capabilities for NNSA's national security programs.

The Vision supercomputer expands NNSA's AI capabilities based on the success of HPE's Venado system. It plans to accelerate research in national security, energy, and scientific fields by utilizing AI capabilities including Foundation Models and Agentic Models.

Both systems are based on the HPE Cray Supercomputing GX5000 architecture and feature Direct Liquid Cooling technology and NVIDIA Vera Rubin superchips.

Argonne Laboratory's Solstice & Equinox Projects

Solstice, being built by Argonne National Laboratory, will be the largest AI supercomputer across all DOE facilities. Equipped with 100,000 NVIDIA Blackwell GPUs, it will focus on solving critical challenges in energy, security, and discovery science.

Equinox, scheduled for completion in 2026, will be equipped with 10,000 GPUs. Both systems will operate integrated with DOE's scientific equipment and data networks.

System Comparison

Technical Significance

This project demonstrates that the U.S. government is making large-scale investments in AI-specialized supercomputing infrastructure for national security and scientific research. Particularly noteworthy is the mass adoption of NVIDIA's latest Blackwell GPUs and Vera Rubin superchips.

HPE Cray GX5000's direct liquid cooling technology is a key element in solving thermal management issues for high-density GPU clusters. As AI workloads increase, the importance of energy-efficient cooling solutions is growing.

National Security Importance

It is significant that the Mission supercomputer will verify nuclear weapon safety without nuclear testing. Under the Comprehensive Nuclear-Test-Ban Treaty (CTBT) regime, nuclear-armed states must rely on computer simulations to maintain the reliability of their nuclear arsenals.

Vision's support for agentic AI models represents an attempt to apply autonomously operating AI systems to national security. This suggests the potential for automating tasks such as threat analysis and scenario forecasting.

[AI Analysis] Future Outlook

This supercomputer project can be seen as part of the U.S. government's AI Sovereignty strategy. As China rapidly expands its own supercomputing capabilities, the United States is showing movement toward building computing infrastructure directly tied to national security under government leadership while reducing dependence on private companies.

Solstice, with its scale of 100,000 NVIDIA Blackwell GPUs, is likely to influence commercial AI model development as well. If foundation models developed by DOE laboratories are released as open-source for scientific research, significant ripple effects are expected in academia and industry.

After Equinox's completion in 2026, DOE is expected to accumulate operational expertise in integrating AI and traditional HPC (High-Performance Computing). This is anticipated to accelerate innovation in computation-intensive fields such as climate modeling, new materials development, and fusion research.

Additionally, if HPE's direct liquid cooling technology is validated, it could potentially become the standard cooling method for large-scale AI data centers in the future. This is expected to contribute to reducing data center operating costs and carbon emissions.