Description

The Advanced Scientific Computing Research (ASCR) program focuses on developing computational and networking capabilities to analyze, model, simulate, and predict complex phenomena relevant to the Department of Energy. ASCR works on enhancing computing systems and architectures to support exascale science. The program is divided into two subprograms: Mathematical, Computational, and Computer Sciences Research, which develops mathematical models, methods, and algorithms for complex systems, as well as software for advanced networks and computers; and High Performance Computing and Network Facilities, which provides cutting-edge computational and networking capabilities. ASCR aims to address national challenges in energy and the environment while advancing U.S. competitiveness in information technology.

Budget

FY 2025 budget increases:

Mathematical, Computational, and Computer Sciences Research: This funding includes investments in “trustworthy and responsible AI,” developing foundation models for scientific applications, and expanding AI partnerships across industry, academia, and government. The funding supports AI testbeds and research efforts to increase participation in AI research, particularly through programs like EPSCoR, CSGF, FAIR, and RENEW. Beyond AI, the increase supports ongoing research in computer science and applied mathematics, addressing challenges in data-intensive science, quantum information science, and emerging technologies. It also funds the development of new scalable energy-efficient algorithms and software, and supports Microelectronics Science Research Centers.

High Performance Computing and Network Facilities: This increase is partly allocated to AI-related infrastructure and capabilities. This includes providing testbed resources to explore emerging AI technologies and deploying software critical for AI operations on exascale systems.

Programs

The Mathematical, Computational, and Computer Sciences Research subprogram supports research to meet the Department of Energy’s high-performance computing and computational science needs. The subprogram includes several activities relevant to AI:

• Applied Mathematics Research: Supports methods for building foundation models for trustworthy and privacy-preserving AI/ML.
• Computer Science Research: Focuses on software infrastructure essential for integrating AI techniques and real-time exploration of complex data.
• Computational Partnerships: Through the SciDAC program, accelerates progress in scientific computing, including AI/ML, through collaborations. It also supports partnerships to build foundation models for new applications in science and energy using DOE’s computing capabilities and AI testbeds.
• Advanced Computing Research: Includes efforts to develop emerging computing technologies and supports workforce development initiatives crucial for sustaining leadership in AI and computational science.

The High Performance Computing and Network Facilities subprogram supports the construction and operation of advanced research computing, networking, and data user facilities to meet critical mission needs. It manages several key facilities and activities, some of which have relevance to AI:

• National Energy Research Scientific Computing Center (NERSC): Provides HPC resources and large-scale storage to SC researchers. The NERSC-9 system (Perlmutter) is an AI-enabled GPU-CPU system.
• Leadership Computing Facilities (LCFs): Includes facilities at Oak Ridge and Argonne National Laboratories, providing diverse leading-edge HPC capabilities. Both LCFs support AI-related initiatives:
  • Oak Ridge LCF: Operates the Frontier exascale system and expands the Quantum Computing User Program.
  • Argonne LCF: Operates the Aurora exascale system and maintains an AI testbed program.
• ESnet: A high-performance network user facility connecting DOE laboratories and sites globally, supporting data transport needs for all SC user facilities.
• Integrated Research Infrastructure (IRI): An initiative to integrate DOE’s data, user facilities, and computing resources to accelerate discovery and innovation, including AI-driven research.