Aurora has officially joined the exascale club. In 2024, the system crossed that long-anticipated threshold and earned its place among the world’s three fastest supercomputers—alongside Oak Ridge’s Frontier and Lawrence Livermore’s El Capitan. It’s a milestone worth celebrating and one that reflects years of planning, preparation, and collaboration.
Aurora set a new standard for AI performance, topping the HPL-MxP mixed-precision benchmark—an important measure of a system’s readiness for AI and data-intensive workloads. It also delivered strong showings in both the Graph500 and HPCG benchmarks, further demonstrating the system’s well-rounded capabilities. For the second consecutive year, Aurora’s DAOS storage system led the production list for IO500, highlighting its capacity to concurrently support simulation, modeling, AI, and analytics at exascale.
And now we’re starting to see those capabilities in action. Over the last few years, the DOE’s Exascale Computing Project and ALCF’s Aurora Early Science Program laid the foundation for a robust data-science software stack. In this report, you’ll read about some of the exciting projects already making waves with Aurora, including the AuroraGPT project to develop powerful foundation models for science; the exascale-optimized HARVEY application that simulates blood flow to advance cancer research; and PHASTA, a computational fluid dynamics package now enhanced with machine learning. With both early science programs coming to a close, we’re excited to see how the 2025 INCITE allocations on Aurora enable researchers to make groundbreaking advances across diverse fields.
You’ll also read about our momentum across other key initiatives. Our ALCF AI Testbed continues to grow, bringing in new accelerator technologies and growing the community of researchers using them to explore how to automate and accelerate scientific discoveries and large-scale data analysis.
Our team also continues to advance the DOE’s vision for an Integrated Research Infrastructure (IRI), an initiative to connect supercomputing resources and experimental facilities to accelerate data-intensive science. In 2024, Argonne launched a fully automated pipeline that leverages ALCF resources to rapidly process data from the Argonne Photon Source for experiment-time analysis. Through Nexus, the Argonne-led initiative, we’re building a unified platform to manage high-throughput workflows across the HPC landscape, helping drive the technologies and methods that will enable a broader IRI.
Education and workforce development have long been a priority at ALCF, and in 2024, we continued to find new ways to make an impact. We hosted 22 training events that reached more than 1,600 participants. We also launched the ALCF Lighthouse Initiative to build and expand partnerships between our facility and U.S. universities. We welcomed our first class of Sprinterns, a national program that provides students with real-world experience through intensive micro-internships. And we’re continuing to grow the next generation of AI researchers through our “Intro to AI-driven Science on Supercomputers” student training series, which has now reached over 700 attendees since launching in 2021.
As always, I’m grateful for the opportunity to share these achievements and milestones with you in our Annual Report. They reflect the outstanding efforts of our team and the invaluable support from our program office and partners. While the technologies and systems we’ve built will drive breakthroughs in a wide range of scientific domains, none of it would be possible without the dedication of the ALCF staff and the strength of our collaborations across the research community. Together, we’re shaping the future of scientific computing, and I’m excited for what lies ahead.