Lincoln Lab Launches TX-GAIN, Most Powerful AI Supercomputer at a U.S. University, to Advance Generative AI and Scientific Research

The Lincoln Laboratory Supercomputing Center (LLSC) has unveiled TX-GAIN, the most powerful AI supercomputer at any U.S. university. Ranked among the world’s top systems by the biannual TOP500 list, TX-GAIN marks a major leap in computational capability for academic research. The system is designed to accelerate generative AI, physical simulation, and large-scale data analysis across multiple scientific and engineering domains. Jeremy Kepner, Lincoln Laboratory Fellow and head of the LLSC, emphasized the system’s transformative potential. "TX-GAIN will enable our researchers to achieve scientific and engineering breakthroughs," he said. "It will play a large role in supporting generative AI, physical simulation, and data analysis across all research areas." TX-GAIN is powered by over 600 NVIDIA graphics processing unit accelerators optimized for AI workloads, combined with traditional high-performance computing hardware. It delivers a peak performance of two AI exaflops—two quintillion floating-point operations per second—making it the leading AI system in the Northeast and the most powerful at any U.S. university. The system is especially tailored for generative AI, which goes beyond simple classification tasks to create new content. While traditional AI identifies patterns—like distinguishing a dog from a cat—generative AI synthesizes novel outputs, such as text, images, or molecular structures. At Lincoln Laboratory, researchers are using this capability to analyze radar signals, fill in missing weather data, detect anomalies in network traffic, and model complex chemical interactions for drug and material discovery. Since going online this summer, TX-GAIN has already enabled significant advances. Rafael Jaimes, a researcher in the Counter-WMD Systems Group, noted that the system allows modeling of far more complex protein interactions and larger proteins than previously possible, a critical step in biological defense research. What sets the LLSC apart is its focus on interactive computing. The center has long developed user-friendly software that lets researchers run complex models without deep expertise in parallel programming. “The LLSC has always tried to make supercomputing feel like working on your laptop,” Kepner said. “People can run their model and get answers quickly from their workspace.” TX-GAIN supports not only Lincoln Laboratory’s internal projects but also collaborations across MIT. These include the Haystack Observatory, the Center for Quantum Engineering, Beaver Works, and the Department of Air Force–MIT AI Accelerator, which is applying AI to optimize global flight scheduling for military operations. The LLSC’s data center in Holyoke, Massachusetts, is designed for energy efficiency. Researchers there are actively developing methods to reduce AI’s environmental footprint. One tool they created can cut the energy required to train an AI model by up to 80 percent. All LLSC systems follow the “TX” naming tradition, honoring the TX-0 and TX-2 computers from the 1950s—pioneering machines that helped lay the foundation for modern computing and AI. With TX-GAIN, the LLSC continues that legacy of innovation, delivering cutting-edge capabilities in a cost-effective and sustainable way.