ORNL Selected for 6 Phase II Genesis Mission Awards

Oct. 8, 2026 — The Department of Energy’s (DOE) Oak Ridge National Laboratory (ORNL) has been selected for six Phase II awards under the DOE’s Genesis Mission. The six multi-year projects will advance Artificial Intelligence (AI)-enabled research in materials science, microelectronics, quantum computing, accelerator operations, HPC software development and fusion energy, building on work initiated during Genesis Mission Phase I.

“These selections align with ORNL priorities in [Artificial] Intelligence, quantum, fusion energy, critical materials and user facilities, underscoring ORNL’s focus on transforming discovery into national impact,” said ORNL Director Stephen Streiffer. “The goal is straightforward — faster, more reliable science that delivers practical results where they matter most.”

The Genesis Mission is a national initiative to mobilize government, industry, academia, nonprofits, and international partners to advance AI for science and technology. Using the DOE-built American Science and Security Platform, partners work together on shared infrastructure that connects researchers with data, computing, and AI tools to advance scientific discovery. Partners focus on the most critical challenges facing our nation in energy, national security, scientific discovery, health, space and more.

ORNL is one of five DOE national laboratories anchoring the computing side of the Genesis Mission.

The goal of the Phase II Request for Application (RFA) awards is to scale up and expand the impact of projects that have already shown potential for AI advantage and demonstrated a trajectory toward a transformative scientific capability. The selected projects are multi-year and use interdisciplinary teams to address national science and technology challenges across key DOE mission areas.

AI‑Empowered Design of Functional Quantum Magnets

This project, led by ORNL’s Yongqiang Cheng, integrates theoretical modeling, high-fidelity computation, and closed-loop experimentation to develop a physics-informed AI framework. Building on this framework, generative AI will identify candidate materials and operating conditions designed to realize targeted quantum phases of matter. Agent-assisted synthesis and AI-enabled laboratories will fabricate and characterize these candidates, with experimental feedback continuously refining the underlying models and guiding subsequent design cycles. The goal is to shift quantum-materials discovery toward predictive, function-driven design and to enable fundamental advances that could transform sensing, energy-efficient computing and emerging quantum technologies.

AI‑Assisted Scientific Software Development (AI4HPC)

AI4HPC, led at ORNL by Jeffrey Vetter in partnership with Argonne National Laboratory, envisions Artificial Intelligence becoming an active partner in scientific computing — not simply generating code, but helping scientists understand, transform, validate, optimize and continuously evolve the complex software that underpins discovery. By combining AI with the compilers, performance tools, numerical validation methods and leadership-class computing environments used in real scientific workflows, the project aims to dramatically reduce the time and specialized effort required to move new ideas into production. Ultimately, AI4HPC could change the economics and speed to deployment of computational science, allowing researchers to explore new algorithms, models and computing architectures that today are often impractical because of the enormous software engineering effort required to realize them.

AI‑eXtended Interfacial Separations (AXIS)

Rare‑earth elements are difficult to separate because they have nearly identical chemical properties. AXIS, led at ORNL by Jim Browning in partnership with the University of Illinois Urbana-Champaign, will create a closed-loop discovery process that tightly integrates AI, machine learning, and experiment. Chemical language models will propose redox‑active ligands — molecules that bind target selected rare-earth ions and release them when switched electrically — while machine‑learned molecular simulations predict how promising candidates will behave. Laboratory experiments and operando neutron measurements will then test those predictions by revealing how the ligands bind ions, organize and transport materials across liquid interfaces. The team will then feed experimental results into the AI and simulation models to improve subsequent designs. AI-assisted thermodynamic and process modeling will then connect molecular and interfacial behavior to continuous separation performance, accelerating the development of selective, reversible and more resource-efficient rare-earth recovery processes.

Multi‑Office Accelerator Team Core (MOAT-Core)

Modern particle accelerators depend on complex software, expert knowledge, and years of operating experience. MOAT‑Core, led at ORNL by Alexander Zhukov in partnership with Lawrence Berkeley National Laboratory, will grow a cross‑facility AI platform that organizes control‑system data, logbooks, simulations, manuals and procedures into SA‑ready formats. Using agentic workflows, the platform will help staff diagnose issues faster, plan machine setups and improve beam quality and uptime. The shared approach lets gains at one facility benefit many others.

Accelerating eXtreme Environment Specs‑to‑Silicon (AXESS)

Designing electronics for extreme environments, such as cryogenic temperatures or intense radiation, can require years of specialized work across disconnected layers of materials, devices, circuits and architectures. AXESS, led at ORNL by Jeffrey Vetter in partnership with Fermilab, will create an AI-driven, specs-to-silicon capability that links those layers, using experimental materials data, physics-based models and intelligent design agents to rapidly explore and validate complete hardware solutions. That capability could enable faster and more capable readout and control systems for quantum computers, fusion experiments, particle detectors, accelerators and other scientific instruments, shrinking the timeline from a new scientific idea to the specialized electronics needed to make an experiment possible.

AI‑enabled Digital Twin Platform for SPARC

Fusion experiments run in pulses, and operators need to plan the next pulse quickly. This project, led by Commonwealth Fusion Systems (CFS) and conducted at ORNL by Jeremy Lore, will build an orchestration layer between the Commonwealth Fusion Systems ecosystem and DOE computing, generate large datasets using trusted community codes, and train surrogate models that can run in control‑room timescales. Agentic AI will prepare and execute code runs, adapt to new data between pulses and help operators design scenarios — demonstrated on SPARC and informed by the U.S. tokamak fleet.