The Genesis Mission Just Committed $5 Billion and Picked 278 Projects. Here Is How the AI-for-Science Machine Actually Works — and How to Get In.

On July 22, 2026, the Department of Energy did something it had never done before: it announced 278 selected projects in a single funding round, spanning all 50 states, drawn from a pool of roughly 5,000 applications — the largest response to any DOE funding call in the agency's history. The next day, White House Office of Science and Technology Policy Director Michael Kratsios put a number on the effort behind it: more than $5 billion in federal commitments across 15-plus federal agencies, all pointed at a single idea — using artificial intelligence to accelerate scientific discovery itself.

This is the Genesis Mission, launched by Executive Order in November 2025 and now, eight months later, moving real money. It is the most ambitious reorganization of federal science funding in decades, and it does not behave like a normal grant program. If you are a researcher, a national lab, a university, or a company that touches AI, materials, biology, energy, or computing, understanding how this machine is built matters more than any single deadline — because the machine is going to keep issuing calls.

Here is how it actually works, and where the openings are.

The architecture: one platform, fifteen agencies, one two-phase funnel

The Genesis Mission is not a grant program in the conventional sense. It is a whole-of-government initiative stitched together on top of shared infrastructure — the DOE-built American Science and Security Platform, which connects researchers to data, compute, and AI tools across 17 national laboratories. The pitch, in Kratsios's framing, is explicitly historical: "America's greatest scientific achievements have been born of national mobilizations paired with the construction of new institutions." The reference points are the Manhattan Project and Apollo. The mechanism is AI.

The money flows through a set of National Science and Technology Challenges — problem areas each agency owns:

• Health (HHS/NIH): chronic-disease root causes, pediatric cancer, and drug discovery through an NIH-led "Bio Genesis Mission."
• Energy and infrastructure (DOE/DOT): grid planning and operations using AI and digital twins.
• Security (Department of War/NNSA): weapons-design timelines and, across six agencies, biological threat detection and nuclear-materials attribution.
• Semiconductors, quantum, and autonomous laboratories as cross-cutting technology bets.
• Space (NASA): mission-data analysis over 150-plus petabytes.

The DOE flagship call — the one that produced the 278 awards — uses a clean two-phase funnel that every prospective applicant should memorize, because it is becoming the template:

• Phase I: $500,000 to $750,000 over 9 months. This is the proof-of-concept tier.
• Phase II: $6 million to $15 million over 3 years. Successful Phase I teams become eligible to compete for Phase II in subsequent cycles, and in FY26 teams could apply directly to either phase.

That is a genuinely unusual funding curve. A 9-month, three-quarter-million-dollar grant is small enough that a mid-sized lab can execute it without reorganizing, but a Phase II award an order of magnitude larger sits directly on top of it. The design rewards teams that can show fast, concrete AI-enabled results and then scale. The governing solicitation is NOFO DE-FOA-0003612; the associated $293 million DOE tranche funds work across advanced manufacturing, biotechnology, critical materials, nuclear energy, and quantum information science.

Who actually won — and what the breakdown tells you

The composition of the first 278-project cohort is the single most useful strategic signal available, because it reveals who the reviewers actually funded when 5,000 teams competed:

• 168 projects led by universities
• 87 led by DOE and NNSA national laboratories
• 19 led by companies
• 4 led by nonprofit organizations

Read that carefully. Universities took roughly 60 percent of the awards, national labs another 31 percent, and companies just 7 percent. For a program marketed around AI and industrial competitiveness, the private-sector share is strikingly thin — which tells you the review process rewarded deep scientific credibility and access to specialized instrumentation over commercial polish. If you are a company chasing Genesis dollars, the lesson is to partner into a university or national-lab-led team rather than lead alone. If you are a university PI, the door is demonstrably open.

The flagship makes the ceiling vivid. The first — and so far only — Phase II project, Prometheus, will receive roughly $60 million and unites 32 partners: five national labs, four universities, and more than 20 corporate partners. That is what the top of the funnel looks like. It is a consortium, not a lab. The trajectory the program is signaling runs from a single-PI $500K Phase I concept to a 32-partner nine-figure-adjacent Phase II coalition. Teams that want to end up at the Prometheus tier should be building relationships now, not when the Phase II call posts.

The money behind the money: where the $5 billion lives

The $5 billion is not one appropriation — it is a stack of agency commitments, and each is its own front door:

• NSF put $380 million into 20 AI-enabled "cloud laboratory" nodes (the Physical-AI / PCL initiative), each funded for four years, with the Astera Institute adding over $20 million in philanthropic match. NSF also committed $80 million-plus for data infrastructure and up to $100 million for AI-ready datasets. These automated, AI-programmable labs — spanning biology, chemistry, soft and 2D materials, metals, and electronics — are shared instruments; even researchers not funded directly may gain access.
• DOE anchored the effort with the $293 million tranche and a separate $60 million, three-year nuclear-energy investment.
• Department of War committed $200 million-plus in FY26 and $1.3 billion-plus in FY27 — the largest forward commitment in the stack, and a signal that defense AI-for-science funding is about to surge.
• IBM added $50 million in quantum-computing access over five years.

This distributed structure is the strategic point. There is no single "Genesis application." There are fifteen-plus agency programs, each contributing awards, datasets, and facilities under a shared banner. A team shut out of the DOE call may be a natural fit for an NSF PCL node or a Department of War topic. The winning move is to map which challenge your work serves and then find the agency that owns it.

The strategic read: opportunity, concentration, and risk

Three things are worth saying plainly.

First, this is a rare expansionary signal in an otherwise defensive year. As we have documented, federal grant windows have collapsed from months to weeks and the OMB grant-rule rewrite has injected political-review uncertainty into the whole system. Against that backdrop, Genesis is money moving toward researchers at scale. It pairs naturally with the NSF State and Regional AI Infrastructure Hubs and the DOE Genesis Mission SBIR/STTR Phase I — the small-business on-ramp to the same challenge areas, with pitch applications due September 10, 2026.

Second, the concentration is real, and it cuts both ways. Compute, data, and the flagship dollars cluster around the 17 national labs and the American Science and Security Platform. For a team plugged into that ecosystem, the leverage is enormous. For a team outside it, the practical path runs through partnership — get onto a lab-led or university-led proposal, or apply for access to the shared PCL instruments rather than trying to stand up your own.

Third, position now for the next window, not the last one. FY26's direct-to-either-phase flexibility will not necessarily repeat; the program's own language points to future phases involving industry, philanthropy, and international partners. The teams that win the next Genesis call will be the ones that spent this quarter building the consortium, aligning their science to a named challenge, and generating the fast, AI-enabled proof points that a 9-month Phase I is designed to reward.

The Genesis Mission is betting $5 billion that AI can compress the timeline of discovery. Whether or not that bet pays off scientifically, it has already changed the funding map. The researchers who treat it as a standing machine — one that will keep issuing calls across fifteen agencies — rather than a one-time announcement will be the ones standing in front of the next window when it opens.

Tracking a specific Genesis challenge area or agency call? Granted maps live federal opportunities against your research profile and flags the on-ramps — SBIR, cooperative agreement, or shared-facility access — that fit your team.