A national-lab AI assistant aims to invent better batteries and clean-energy materials in years, not decades

Lawrence Berkeley National Laboratory is leading a new effort to build an artificial-intelligence assistant for discovering the materials behind future clean-energy technologies, from better batteries to advanced semiconductors. Announced on February 3, 2026, the project, called FORUM-AI, is backed by 10 million dollars over four years from the U.S. Department of Energy and aims to shrink research timelines that have traditionally stretched across decades.

The system is built to assist scientists at every stage rather than replace them. FORUM-AI combines three kinds of AI: generative models that produce text and images, reasoning models that work through problems step by step, and agentic systems that can take autonomous actions, such as launching large-scale simulations on supercomputers and running robotic experiments. Crucially, these steps can happen in parallel rather than one after another, letting the system generate hypotheses, test them in silico and feed results back into the next round far faster than a traditional lab cycle.

“Announced on February 3, 2026, the project, called FORUM-AI, is backed by 10 million dollars over four years from the U.”

The collaboration is broad and built in the open. Berkeley Lab is working with Oak Ridge and Argonne national laboratories, MIT and Ohio State, and is designing FORUM-AI as an open-source, general-purpose platform that integrates with the publicly accessible Materials Project database. "It will help scientists at every step of energy materials research, from hypothesis generation and computer simulations to laboratory experiments and analysis," said principal investigator Anubhav Jain.

The team is candid about the risks. Large AI models can hallucinate plausible-sounding but false answers, so FORUM-AI is designed to ground its work in verified databases, expose transparent reasoning traces that researchers can inspect, and rely on physics-based simulation tools using community-standard methods. It is an assistant, and human scientists make the calls and run the validating experiments. Even so, an open platform that helps overstretched labs explore promising materials faster, aimed squarely at clean energy, is the kind of tool that could meaningfully accelerate the transition to a lower-carbon world.