Alibaba Group’s research division, Damo Academy, has developed an artificial intelligence (AI) agent that has successfully identified four previously unknown superconducting materials, which were later validated through laboratory testing. This development marks a significant step in the application of AI to materials science, particularly in the search for superconductors.

Superconductors are materials that can conduct electricity without resistance and expel magnetic fields when cooled to extremely low temperatures. Their unique properties hold promise for advancing technologies in power transmission, quantum computing, and high-speed transportation systems. However, discovering new superconductors has traditionally involved lengthy trial-and-error experimentation, as a comprehensive theoretical framework for predicting superconductivity remains elusive. To date, roughly 2,000 superconducting compounds have been documented.

The AI system, named Elements Claw, was designed to accelerate the discovery process by analyzing scientific publications and screening extensive databases of crystal structures. Developed in collaboration with Renmin University of China and the University of Chinese Academy of Sciences, the system uses a specialized foundation model with one billion parameters trained on 125 million molecular and crystal structures.

In a computationally efficient run using 28 GPU hours, Elements Claw analyzed 2.4 million stable crystal structures, identifying approximately 68,000 candidates with potential superconducting properties. This list was further narrowed to select the most promising materials for physical laboratory testing, resulting in the verification of four novel superconductors.

Rong Yu, head of scientific intelligence at Damo Academy, indicated that these four compounds represent the first superconducting materials discovered using an AI agent and subsequently confirmed in the lab, with many more candidates awaiting exploration.

Alibaba’s initiative follows a broader industry trend where technology companies apply AI beyond traditional applications such as chatbots and coding assistants toward scientific research. AI models are increasingly being used to process vast datasets, analyze scientific literature, and generate testable hypotheses. For example, Google DeepMind’s AlphaFold has revolutionized protein structure prediction, and Microsoft has developed AI tools targeted at the discovery of new materials.

Materials discovery is viewed as a particularly promising area for AI, given the vast chemical space researchers must navigate to find viable candidates. AI-based screening can significantly reduce the time and resources spent identifying compounds worth experimental exploration.

Huang Wenbing, associate professor at Renmin University’s Gaoling School of Artificial Intelligence, noted that the framework employed in Elements Claw could extend to discovering materials for applications such as solid-state batteries, catalysts, and thermoelectric devices.

Alibaba also intends to share its AI-generated predictions with the wider research community, aiming to accelerate discoveries beyond the capabilities of any single laboratory, a strategy similar to that used for the AlphaFold protein database.