A Tulane University team has won a $750,000 federal grant to prove artificial intelligence can accelerate the search for new superconductors, competing for up to $15 million in additional funding under the government's Genesis Mission. The project places the researchers at the center of a global race to find materials that conduct electricity without energy loss at practical temperatures - a breakthrough that could reshape power grids, electronics, and quantum computing.
In July, physics professor Jianwei Sun and his colleagues were among 278 awardees selected from roughly 5,000 applicants for the Genesis Mission, a $5 billion multiagency initiative launched by the Trump administration in November 2025. The Tulane team is partnering with Oak Ridge National Laboratory in Tennessee and has nine months to demonstrate results before the next funding decision.
The high-temperature prize
Superconductors already enable MRI machines, particle accelerators, and some quantum computers. But every known superconductor must be cooled to temperatures colder than outer space or subjected to crushing pressures to function. Those constraints require expensive cooling systems - MRI machines, for example, encase superconducting wires in liquid helium-filled cryostats - and severely limit where the technology can be deployed.
"Finding a high-temperature superconductor would change the world just like the Copper Age or Iron Age did thousands of years ago," Sun said. Daniel Straus, an assistant professor of chemistry on the team, called the search "the holy grail of science."
A room-temperature superconductor could enable superefficient power grids that transmit electricity over long distances with minimal loss, tiny hyperefficient motors and electronics, and magnetically levitated trains. It could also advance quantum computing to the point of making today's data centers obsolete. The appetite for such a discovery was on full display in 2023, when a South Korean team's claim of a room-temperature superconductor triggered a frenzy among scientists and Silicon Valley investors before replication attempts failed.
How the AI search works
The Tulane team's approach does not start from scratch. Instead, physicists and computer scientists will write algorithms to mine massive databases of existing materials - compounds that were synthesized and catalogued but never tested for superconductivity. "Based off of what AI says, we can narrow down the search faster with AI than with human eyes," said Jackson Smith, a chemistry Ph.D. student on the project.
The work requires elaborate quantum-mechanical simulations to predict how atomic and subatomic matter will behave. Access to LONI, Louisiana's state-owned supercomputer, will cut the time needed to run these algorithms from months to days, Straus said. Aron Culotta, a professor of computer science, noted that AI excels at pattern recognition across vast datasets. "There's a lot of room to make the process faster and more effective while humans stay in control," he said.
Once algorithms flag promising candidates, the researchers will synthesize the materials in Tulane's labs by combining chemical compounds and heating them in specialized furnaces. The most promising results will then travel to Oak Ridge for testing on equipment the university does not possess. "It is impossible to just use theory alone, like computations, to predict a new superconductor," Straus said. "We have to make them and test them."
The funding clock
By March 2027, the six-month mark, the team aims to show that AI identified at least a handful of existing materials with superconductive properties. Success would unlock second-phase funding that Culotta said could scale the project toward discovering entirely new materials and automating more of the discovery pipeline.
"The goal is to demonstrate accelerated discovery, compressing what would normally take 10 or 20 years into that time," Straus said. Culotta added: "We've already got plenty of superconductors that require impractically low temperatures to operate. Finding ones that function at higher temperatures is the ultimate goal."
Why this matters for science and research professionals
The Tulane project is a concrete test case for whether AI for Science & Research can meaningfully shorten the timeline from hypothesis to verified result in materials science. For researchers in adjacent fields, the methodology - training algorithms on curated structural databases like the Inorganic Crystal Structure Database, then validating predictions through physical synthesis and testing - offers a template that could transfer to catalyst design, battery chemistry, and drug discovery. The Genesis Mission's structure, with its rapid six-month proof-of-concept window followed by scaled funding, also signals how federal agencies are structuring AI research grants, rewarding teams that can demonstrate speed and reproducibility rather than just computational novelty. Professionals building or leading AI Learning Path for Research Scientists programs should watch how the Tulane-Oak Ridge partnership balances domain expertise across physics, chemistry, and computer science - a multidisciplinary model likely to become standard as AI-assisted discovery moves deeper into experimental labs.
Your membership also unlocks:
