Georgia Tech opens its AI supercomputer to researchers nationwide

Researchers across the United States can now apply to use Georgia Tech’s Nexus AI supercomputer in Atlanta. At least 90% of its computing capacity will be available to the national science and engineering community through peer-reviewed applications, opening a shared resource for projects too large or complex for a regular computer.

National access, allocated by review

Nexus is housed at Georgia Tech’s Coda data center, but its research audience extends beyond the campus. Cloud-based services let research teams access its computing resources from elsewhere in the country.

Researchers must submit applications, and access is awarded through a peer-reviewed allocation process coordinated by the National Science Foundation.

Researchers can apply through the NSF-funded Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support program, known as ACCESS.

By bringing together advanced computing capabilities and expertise from across the research community, Nexus will help broaden the range of scientific questions researchers can tackle and strengthen the nation’s capacity for AI-enabled discovery.

Irina Dolinskaya, NSF Directorate Head for Computer and Information Science and Engineering

AI models and simulations share a platform

Nexus combines high-performance computing, artificial intelligence and cloud-based services. Its practical distinction is the ability to develop AI models, run simulations and work with large datasets on the same platform.

The supercomputing side works by dividing a large problem into smaller tasks and processing many of them at once. That differs from making a single processor work through the entire calculation. Nexus is designed to combine this parallel approach with AI and support demanding projects that run over longer periods.

Hurricane modeling illustrates the approach. A supercomputer can divide the modeled space into small three-dimensional boxes and assign processors to simulate activity within them. Those processors share information with neighboring processors to build the larger picture.

A university host for a national resource

The system was developed with a $20 million NSF investment and a partnership between Georgia Tech and the University of Illinois Urbana-Champaign. It is the first national academic supercomputer Georgia Tech has hosted.

Potential research uses span drug discovery, neuroscience, energy, aerospace, robotics and advanced manufacturing. Georgia Tech President Ángel Cabrera said the resource would help researchers accelerate discovery and strengthen collaboration in fields where advanced computing can make a meaningful difference.

Nexus will also support the National Artificial Intelligence Research Resource, an NSF-led effort to expand access to AI research tools.