The Genesis Mission, the Department of Energy’s big AI push, unveils first grants

WASHINGTON, D.C.—The U.S. Department of Energy (DOE) has announced the first round of grants supported by the Genesis Mission, the governmentwide push into artificial intelligence (AI) in research that the agency spearheads. Unveiled yesterday at the inaugural Genesis Mission Summit, the 278 awards will draw on the more than $250 million that DOE has pledged to support multi-institutional teams in using AI for everything from designing nuclear reactors to studying soil microbes. Officials say those initial awards are just a taste of what’s to come. Yesterday, the White House declared an expanded commitment of more than $5 billion to the effort—though it did not specify where the funding would come from or the time period for the spending. Several agencies, including the National Science Foundation, NASA, and the National Institutes of Health, also announced their intention to join Genesis. “The 278 projects selected today represent the very best of our nation’s scientific enterprise,” said Secretary of Energy Chris Wright in a statement unveiling the awards. Launched in November 2025, Genesis aims to harness AI to tackle 26 “challenging problems of this century.” This spring, DOE invited researchers to propose projects addressing 21 of those challenges. Scientists had to scramble to apply: DOE gave them just 6 weeks to assemble teams that had to include members from at least two of three types of institutions: universities and research institutes, DOE’s 17 national labs, and private companies. In spite of the rush, DOE received more than 5000 applications. All but one of the winners of the first awards will receive so-called phase 1 funding of $500,000 to $750,000 for 9 months of work. If that work is successful, a team can apply for an additional phase 2 award of $6 million to $15 million. Teams that chose to skip phase 1 and apply directly for phase 2 funding in the original round of proposals can expect to hear back this fall, DOE says. Grant winner Amanda Albright Olsen, an environmental geochemist at the University of Maine, says that access to DOE’s vast troves of data was a major draw. Together with colleagues from the University of Delaware and Argonne National Laboratory, Olsen will study how underground microbes influence water chemistry, mineral formation, and the movement of pollutants through soil. Because microbial communities constantly change their behavior, these processes have long resisted conventional simulation through physics-based models. So, starting with DOE’s large data sets of coastal ecosystems, where fresh- and saltwater mix to create especially dynamic microbial environments, researchers plan to incorporate AI models trained on microbial genomic data into existing soil simulations to build a model that can forecast how subsurface ecosystems evolve. Such findings could inform groundwater management and planning for energy infrastructure, Olsen says. Another winner focuses on the supply of critical minerals. Ahamed Irshad, an electrochemist at SLAC National Accelerator Laboratory, is partnering with researchers at the University of Southern California to use AI to improve the recovery of metals such as cobalt, nickel, and manganese from spent lithium-ion batteries. Because the metals have similar chemical properties, they often separate poorly during recycling. So the researchers plan to deploy AI agents to more efficiently search for chemical ligands that selectively bind individual metals. The agents will review the literature, generate hypotheses, interpret experimental results, and propose new experiments far faster than the researchers could do themselves. “We don’t want to have to do 1000 or 10,000 experiments,” Irshad says. The first and so far only phase 2 project will receive a whopping $60 million. Called Prometheus, it brings together 32 partners from five national labs, four universities, and more than 20 corporate partners. They aim to develop the first autonomously designed, built, and operated nuclear reactor. Designing a new reactor typically takes about 15 years and hundreds of engineers, says Peter Suyderhoud, a systems engineer at Idaho National Laboratory. The consortium hopes AI can cut that timeline roughly in half. Suyderhoud estimates the computing-intensive undertaking will require on the order of trillions of tokens—the small chunks of text that large language models use to process information. The project is also backed by more than $200 million in investment from industry. DOE funded less than 5.6% of applications, which surprised some researchers—especially as DOE has funded Genesis by siphoning money from its other long-standing grants programs. Given the low success rate, Jaki Noronha-Hostler, a theoretical nuclear physicist at the University of Illinois Urbana-Champaign, worries some fields could get left out. “The real question is are you just taking away part of the budget that was for people doing machine learning and redistributing it through another pipeline?” she says. “Or are you taking away research funds from pen-and-paper theorists or people building the experiment that we need to do?” Winners say they’re still waiting for details on how the next 9 months will unfold. Olsen says her team still doesn’t know the official start date, or how DOE will conduct the 6-month evaluations, where teams will be winnowed for phase 2 funding. “It’s a little unnerving,” she says. “Once the clock starts, you really have to be working to get everything done.” Competition to get those limited phase 2 awards will be fierce, suspects Stephen Streiffer, director of Oak Ridge National Laboratory. “There will be a downtick,” he says. “I would expect maybe a tenth [of phase 1 projects] will make it that far.” In fact, given the short time frame and the slim chances of getting a phase 2 grant, one condensed matter physicist whose proposal was not funded says Genesis funding may be a mixed blessing. “I don’t even know if I feel happy or unhappy that I didn’t get the Genesis funding,” says the physicist, who requested anonymity because none of his university’s multiple proposals was funded. “If you hire new graduate students or postdocs, what are they going to do if you don’t advance to phase 2?” The physicist says the Genesis funding model may not be the best for basic research and worries other agencies will adopt it. Such concerns notwithstanding, the sheer number of proposals reflects the interest scientists have in harnessing AI, says Darío Gil, DOE’s undersecretary for science and director of Genesis. “Look, we didn’t design the program to have a 5% selection ratio,” he says. “This is purely a consequence of the level of enthusiasm that the community has had.” More funding is coming, Gil notes, so researchers hoping to get in on the action should “keep going.”