Conference: TANGO@SC25

As research workflows become more data-intensive and complex, there is a need for greater acceleration, and GPUs are a popular choice. Federated GPU-heavy systems are fully subscribed, and this practice is costly in terms of energy and hardware. However, innovation that employs next-generation arithmetic, CXL, and RISC-V technologies can produce greater precision with fewer bits, tackling both energy and storage challenges for AI/ML workflows, while paving the way for quantum computing, which will rely on next-generation math and standards. The efforts in this area will support the nation’s leadership in AI and quantum, and strengthen the cyberinfrastructure ecosystem. This project, in collaboration with STEM-Trek Nonprofit and the Conference on Next-Generation Arithmetic (CoNGA), will host the TANGO workshop ahead of the Supercomputing Conference, SC25, in St. Louis, Missouri November 16-22, 2025. Participants will learn how Compute eXpress Link (CXL), RISC-V, and Next-Gen Math drive the revolution toward more energy-efficient supercomputers. With such tools, it is possible to render artificial intelligence and machine learning (AI/ML) workflows with greater precision and fewer bits. A greater understanding of next-gen math is also essential for those who develop embedded systems and quantum arithmetic circuits. This is the 6th workshop in the series and co-funded by industrial support. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.