Argonne Expands AI Inference With Tara, Minerva
The Argonne National Laboratory added two new computing systems to its AI inference service as the Energy Department pursues its Genesis Mission.
Updated on Nov. 5, 2026 in Artificial Intelligence
Live Poll
Should the federal government spend more on AI research for science?
Argonne National Laboratory expanded its AI inference service with new computing systems named Tara and Minerva, along with an expanded Metis platform. The service, launched in May, now hosts more than 50 models.
Why it matters
The expansion gives researchers on-premises access to large-scale AI computing as the Department of Energy builds out infrastructure for its Genesis Mission to accelerate scientific discovery. It also points to a broader federal push to have AI agents work alongside research data, supercomputers and laboratory instruments.
Tara comprises 672 compute nodes built on the HPE Cray Supercomputing EX4000 architecture, while Minerva is an eight-node NVIDIA DGX B200 cluster. Tara's 258 TB of combined GPU memory represents a 100-fold increase over the prior inference service's capacity.
The players
Argonne National Laboratory
A U.S. Department of Energy research facility near Chicago that operates large-scale supercomputing and AI infrastructure for scientific research.
Department of Energy
The federal agency overseeing national research laboratories and leading the Genesis Mission to apply AI to scientific discovery.
Nvidia
A semiconductor company whose GPUs and Superchips power the new Argonne systems and whose partnership with the DOE was announced in October 2025.
The details
Argonne expanded its Metis platform from two to six SambaRacks, totaling 96 SN40L processors. The laboratory is also developing tools that will let AI agents submit computing jobs and access data stored on its other systems.
Timeline
- October 2025: The Department of Energy and Nvidia announced five Argonne computing installations, including Tara and Minerva.
- May: Argonne's AI inference service launched.
- 2029: The Department of Energy targets roughly 100,000 persistent AI agents running in its infrastructure.
The Tech Race
The Argonne expansion extends the Department of Energy's Genesis Mission, which seeks to give AI agents access to research data, supercomputers and laboratory instruments. It places federal scientific computing alongside the commercial AI infrastructure buildouts now underway at major technology firms.
Researchers at national labs gain access to larger AI computing capacity, which could shorten the time needed to run scientific models and test hypotheses. Scientists outside Argonne may eventually use the same inference service through the Genesis Mission's common model access gateway.
The takeaway
Federal labs are becoming a testing ground for AI systems that work directly with scientific data, so progress on this infrastructure is worth following as the 2029 agent target approaches.
What happens next
Stevens is scheduled to discuss the Genesis Mission at the SC26 conference in Chicago, held Nov. 15-20.
Further reading
Explore more on the Artificial Intelligence section for context on how federal labs are deploying AI infrastructure.
Live Poll
Should the federal government spend more on AI research for science?
