High-performance computing conference highlights emerging links between quantum, HPC and AI
The Department of Energy’s Oak Ridge National Laboratory joined other national labs in hosting DOE Undersecretary of Science Dario Gil at the 2025 International Conference for High Performance Computing, Networking, Storage and Analysis (SC25). His fireside chat at the DOE booth was one of several talks and presentations at the annual event.
Held in St. Louis, Missouri, SC25 saw thousands of attendees over the four-day event with several major industry figures and academic institutions who are investing in supercomputing, AI and quantum. Gil’s fireside chat with Deb Gracio, Director of Pacific Northwest National Laboratory, was titled "Accelerating Discovery: Building the AI Ecosystem for the Future of Science" and stressed the importance of investing in AI as a nation moving forward.
Director of ORNL’s Quantum Science Center Travis Humble also gave a featured talk at the DOE booth where he discussed driving high performance computing innovation using quantum computers and integrating this technology into existing HPC systems.
“SC25 clearly showcased the rapidly developing connections between quantum computers, which are seen to be accelerating their development, with the broader field of HPC, especially AI-driven systems,” said Humble.
The conference’s focus on emerging AI-driven workflows continued as additive manufacturing researcher Stephen Dewitt demonstrated his latest work creating an artificial intelligence agent that makes decisions as parts are made.
"Supercomputing is a unique event that brings together the whole computing community,” said Dewitt. “It was the perfect place to learn about the new ways computing facilities will support autonomous science, meet with industry partners and share my team's work on using agentic AI (where AI agents are able to act autonomously) to power advanced manufacturing."
SC25 clearly showcased the rapidly developing connections between quantum computers, which are seen to be accelerating their development, with the broader field of HPC, especially AI-driven systems.
ORNL teams earn Best Paper, Best Student Paper and multiple SC25 honors
ORNL researchers brought home Best Paper and Best Student Paper awards, and multiple projects were finalists for the prestigious Association for Computing Machinery’s Gordon Bell prizes given annually at the conference. Best Paper went to ORNL’s ORBIT (Oak Ridge Base Foundation Model for Earth System Predictability) team for its work on the next-generation AI-driven weather prediction tool, ORBIT-2. The ORBIT team includes ORNL researchers Prasanna Balaprakash, Dan Lu, Xiao Wang, Jong Youl Choi, Isaac Lyngaas, Hong-Jun Yoon, Dave Pugmire, Ming Fan, Nasik Muhammad Nafi, Dali Wang, Peter Thornton, Takuya Kurihana and Moetasim Ashfaq.
“Winning the Best Paper Award at SC25 is an important recognition of how foundational AI and exascale computing can accelerate scientific discovery and strengthen national security,” said Lu. “ORBIT-2 demonstrates that we can scale Vision Transformers in ways that were previously impossible, enabling new capabilities for high-resolution, data-driven modeling across complex systems critical to the nation.”
Best Student Paper, titled "X-MoE: Enabling Scalable Training for Emerging Mixture-of-Experts Architectures on HPC Platforms," highlighted how AI models deliver strong model quality through fine-grained expert segmentation and large top-k routing. (Top-k routing is a decision-making mechanism used within Mixture of Experts models to decide which parts of the neural network should process a specific piece of data.) The paper included ORNL’s Feiyi Wang and Sajal Dash.
ORNL researchers William Godoy, Pedro Valero-Lara, Wael Elwasif, Philip Fackler, Keita Teranishi, Rafael Ferreira da Silva and Jeffrey Vetter were part of a team that won Best Paper at the SC25 Twelfth Workshop on Accelerator Programming and Directives (WACCPD), as well as University of Tennessee student and ORNL intern Tatiana Melnichenko. The team did a comprehensive study on the novel Mojo language’s vendor-neutral GPU capabilities for HPC science workloads. Melnichenko presented the paper at WACCPD and was also named a best poster finalist in the ACM undergraduate research poster competition.
ORNL’s Woong Shin, Renan Souza, Daniel Rosendo, Frederick Suter, Feiyi Wang, Prasana Balaprakash and Ferreira de Silva also earned the distinguished paper award at the SC25 Workshop on Workflows in Support of Large-Scale Science (WORKS).
This marks the third consecutive year a team of ORNL co-authors has taken home a best paper recognition from an SC workshop.
Gordon Bell finalists demonstrate ORNL’s leadership across HPC science
Gordon Bell award finalists with contributions from ORNL researchers included BerkeleyGW: Investigating a fundamental concept of quantum physics, which includes ORNL Distinguished Research Scientist Trey White; HACC: Calculating the largest-ever simulation of the universe, which includes ORNL Applied Mathematician Antigoni Georgiadou and co-authors ORNL Applied Mathematician Audrey Prokopenko and ORNL Senior Computational Scientist Damien Lebrun-Grandie; MFC/IGR: Unleashing the largest-ever CFD simulation of high-speed compressible fluid flow, with ORNL scientific liaison Reuben Budiardja providing assistance; and QuaTrEx: Exploring the nanoscale world of next-generation transistors.
HPCwire award highlights ORNL’s open-source leadership
ORNL researchers were also honored with at SC25 with the 2025 HPCwire Readers’ Choice Award for Best HPC Collaboration (Academia/Government/Industry), recognizing the laboratory’s leadership and partnership in advancing high performance computing technologies that support national priorities in energy, security and scientific innovation. Leading ORNL’s contribution in this collaboration is Damien Lebrun-Grandie, a senior computational scientist whose team plays a key role in the High-Performance Software Foundation (HPSF). The foundation unites government, academic and industry partners to develop and sustain critical open-source software that underpins the HPC and AI ecosystem. ORNL is a founding member of HPSF, working alongside Lawrence Livermore National Laboratory and Sandia National Laboratories to establish the organization.
“We are honored to receive this HPCwire Readers’ Choice Award, which underscores the crucial importance of community-driven efforts like the High-Performance Software Foundation,” said Lebrun-Grandie. “Our collaboration ensures that the essential open-source software underpinning both traditional HPC and AI-driven science is robust and sustainable. This recognition belongs to all the HPSF partners who are collectively driving the future of scientific discovery.”
UT-Battelle manages ORNL for the U.S. Department of Energy’s Office of Science, the largest supporter of basic research in the physical sciences in the United States. The Office of Science is working to address some of the most pressing challenges of our time. For more information, visit energy.gov/science. — Mark Alewine
