Biophysical Society Bulletin | February 2026

Public Affairs

States Build AI Supercomputers to Power Next-Generation Science As AI research accelerates, states are stepping in to give universities the computing muscle they cannot afford on their own. New York’s Empire AI—a $500 million, 10-year initiative—has quickly become a model, providing research ers access to cutting-edge Graphics Processing Units (GPUs) needed for modern AI systems. Its first machine, Alpha, is already supporting more than 350 scientists; its second, Beta, arriving soon, will be among the most powerful academic AI supercomputers in the nation.

While fusion researchers have long advocated for their own office, arguing that the field is expanding rapidly with the rise of well-funded private fusion companies, concern remains that the new offices may be created by carving programs out of DOE’s Office of Science, the nation’s largest supporter of the physical sciences and steward of major national labs and research facilities. The new AI office may absorb the Office of Science’s advanced scientific computing research program, potentially shifting DOE’s focus toward AI-optimized machines that are not as strong at traditional high-precision simulations. Some national labs—such as the Princeton Plasma Physics Laboratory— could be reassigned to the new fusion office. With staffing shortages at DOE and a long-standing bipartisan commitment to basic research, many fear the move could weaken support for fields that fall outside today’s favored technologies. cloud costs mean many scientists still face long waits. Empire AI plans to keep scaling: a third system, Gamma, is slated for 2027 and will be 10 times more powerful than Beta, with a fourth, Delta, to follow. Meanwhile, federal agen cies are exploring larger investments, including the partner ships of the US Department of Energy with NVIDIA and AMD to build nine new AI-focused supercomputers. As AI becomes central to scientific discovery—from climate prediction to cancer diagnostics—states are building the infrastructure needed to keep US research competitive, collaborative, and publicly accessible. New York is not alone in this push. California is launching CalCompute, an AI-ready supercomputing program at the University of California. Federal initiatives such as the National Science Foundation’s CloudBank and the pilot National AI Research Resource help, but limited access to GPUs and high

DOE Reorganization Prioritizes AI and Fusion, Raising Concerns for Basic Research The US Department of Energy (DOE) is restructuring its research portfolio, launching a new Office of Fusion and an Office of Artificial Intelligence and Quantum. The shift comes as the US administration sharpens its focus on emerging technologies. On November 24, President Donald Trump signed an executive order creating the Genesis Mission, an AI-accelerated science initiative to be led by DOE Undersec retary for Science Dario Gil . The effort envisions a unified plat form connecting US supercomputers, AI systems, quantum computers, and scientific instruments into what Gil calls “the most complex and powerful scientific instrument ever built.” Science is already advancing. University at Buffalo research ers are using Empire AI to build SWAXSFold, an AI system that predicts how proteins change shape—an essential leap beyond AlphaFold’s static structures. Columbia scientists are training models to forecast flash floods at neighborhood scales, while New York University neuroscientists are running virtual brain networks in a fraction of the time. At Weill Cor nell, researchers are training AI to identify a dangerous sub type of prostate cancer—work impossible on local clusters.

February 2026

6

THE NEWSLETTER OF THE BIOPHYSICAL SOCIETY