Trump Unveils $5B+ Genesis Mission, DOE Picks 278 AI Projects
TL;DR
- The 5,000-plus applications for 278 funded slots is the largest response to any single DOE funding call in the agency's history.
- Meta's SAM 3 and DINOv3 are already running on 300 A100 GPUs across Lawrence Berkeley, Argonne, Brookhaven, and Oak Ridge, compressing some analyses from a month to 15 minutes.
- Open-weight models dominate initial deployments because secure government infrastructure requires on-premise fine-tuning, making external cloud access a non-starter for classified and sensitive workloads.
The Trump administration used July 22 to line up a big number and a concrete project list behind the same idea: pull the federal science apparatus, DOE-led, into one AI-for-science push. According to the White House release, more than $5 billion in federal commitments now sit behind the Genesis Mission, an initiative Trump launched by executive order in November 2025. The umbrella spans more than 15 agencies, from DOE and HHS to NASA and NSF, and Michael Kratsios, director of the White House Office of Science and Technology Policy, framed the day in institution-building terms: "America's greatest scientific achievements have been born of national mobilizations paired with the construction of new institutions."
The concrete piece under that umbrella came from Energy Secretary Chris Wright, who announced 278 projects selected for award negotiations, drawing on $250 million in DOE funding. The DOE breakdown is 87 awards led by DOE and NNSA national laboratories, 168 by universities, 19 by companies, and 4 by nonprofits, with 342 participating institutions in total, including 16 national labs, 142 universities and 157 companies. The single largest award is a three-year, $60 million nuclear energy project intended to use AI to deliver nuclear facilities faster and cheaper. Research areas span nuclear energy, critical minerals, intelligent chip design, and commercial fusion energy.
The industry hook worth watching is the model layer. Arcee AI is building Genesis-Science-1, described as an American open-weight AI model with a governed research harness that keeps a reproducible record of scientific workflows. If the reproducibility claim holds up in practice, that harness is the interesting part; it is an implicit answer to the biggest complaint about using LLMs in science, which is that you cannot audit what the model actually did.
Read carefully, "selected for award negotiations" is not the same as funded and running, and neither release spells out per-agency dollar splits, timelines, or which projects are net-new spend versus rebadged R&D. What is clear is the direction: national labs, universities, and a specific list of vendors are being lined up as an American AI-science stack, and the near-term winners are the labs and universities holding the biggest allocations, plus the model vendors whose weights get plugged into the harness.
What others are reporting
-
U.S. Department of Energy Read →
Official DOE source confirming 278 selections, all-50-states coverage, and the largest-ever DOE funding-call response; ground truth for project counts and institutional breakdown.
"The 278 projects selected today represent the very best of our nation's scientific enterprise." — Secretary Chris Wright
-
Meta AI Read →
First-party account of SAM 3 and DINOv3 on 300 A100s across four national labs; documents the open-weight rationale for secure government compute environments.
"This compresses discovery time from days to moments and establishes a continuous, self-improving model of science."
-
The Next Platform Read →
Technical depth on named infrastructure: HPE Lux supercomputer at Oak Ridge, Microsoft's $60M breakdown, 340 consortium members; includes skeptical analyst view on scope vs. demonstrable wins.
"We can only fully harness AI and its associated productivity uplift by boldly reforming our scientific institutions" — Michael Kratsios, OSTP
-
The Next Web Read →
Frames Genesis Mission inside the broader $200B OSTP R&D redirection and US-China competition; cites Stanford finding that the AI performance gap has narrowed to under 3% despite a 23x US private-investment lead.
"The largest marshalling of federal scientific resources since the Apollo programme"
Originally reported by whitehouse.gov
Read the original article →Original headline: Trump Admin Unveils $5B+ Genesis Mission for AI-Driven Science, DOE Awards 278 Projects Totaling $293M