White House redirects research funding from universities to AI

The Trump administration's Office of Science and Technology Policy, in its July 21, 2026 report 'Science: A New Golden Age,' calls for reallocating roughly $200 billion in annual federal R&D spending away from universities and traditional peer-review panels and toward individual scientists and AI-driven research ^[1]. The report directs agencies to refocus support on individual scientists rather than legacy institutions and to diversify funding mechanisms beyond consensus peer review ^[2]. Any federal agency with at least $3 billion in R&D authority now has 90 days to submit an implementation plan, and a companion Fiscal Year 2028 R&D Priorities Memorandum co-signed by OMB Director Russ Vought and Kratsios pushes the directive across every federal department ^[3]. The administration argues traditional peer review is good at filtering out bad science but poor at spotting exceptional, high-risk ideas, hence the push toward portable fellowships and long-horizon grants modeled on things like the NIH Director's Pioneer Award ^[4].

The vehicle for this shift already exists. The Genesis Mission is a Department of Energy-led initiative that Trump launched by executive order in November 2025 with the explicit goal of doubling the productivity of U.S. federal research within a decade ^[5]. More than 15 federal agencies have pledged over $5 billion to the Mission ^[6], and at a July 22, 2026 summit the DOE unveiled its first wave of 278 AI-for-science project awards, selected from more than 5,000 applications ^[7]. Twenty-four companies, including Microsoft, Google, AWS, NVIDIA, OpenAI, and Anthropic, signed memorandums of understanding with the DOE covering compute, models, and expertise, with Dr. Dario Gil, the DOE's Under Secretary for Science and the Mission's director, calling the agreements 'only the beginning' of bringing companies, universities, non-profits, and federal agencies into a single effort ^[8]. The scale is meant to be systemic rather than a single grant program: the Mission builds on the pre-existing National AI Research Resource, links 17 DOE national laboratories and roughly 40,000 DOE scientists and engineers into a unified digital environment, and aims to connect supercomputers to robotic laboratories through what officials have described as an 'American Science and Security Platform' - a closed-loop system meant to let AI agents test hypotheses and run live experiments, not just analyze data. Officials have compared the effort, in urgency and scale, to the Manhattan Project. That framing raises the stakes on the follow-through: Lynn Parker, who co-chaired the federal task force that first stood up the National AI Research Resource, has cautioned that sustained, public-interest-aligned investment is what makes this kind of infrastructure pay off, a marker worth watching against the administration's bet that redirecting existing dollars, rather than adding new ones, is enough.

Congressional and academic pushback has focused on mechanics as much as ideology. Because federal grants have historically flowed through universities, which charge overhead rates that can run to roughly 50 percent, meaning a $1 million grant can return about $500,000 to the institution for infrastructure, administrative staff, and cross-subsidies to smaller departments, moving money directly to individual scientists threatens that revenue stream ^[9]. The Association of American Universities says leading research universities already saw a 15 percent drop in doctoral program applicants in fall 2026 versus the prior year, which it links to funding uncertainty ^[9]. Rep. Zoe Lofgren called the report's title 'stunningly ironic,' arguing the administration 'is systematically destroying U.S. science' ^[10]. Vanderbilt astrophysicist Keivan Stassun warned the plan trades fundamental research and peer expertise for faster commercialization - 'you don't get more apple pie by cutting down the apple trees' - while University of Pittsburgh's Jeremy Berg dismissed the report's underlying claims as 'tremendously misinformed' ^[10]. That critique isn't confined to voices with an institutional stake in university funding: Lynn Parker's warning that abandoning long-term investment risks ceding leadership in the technologies that define the economy and national security offers a more measured version of the same concern, from inside AI policy circles rather than a university lobby. Whether the 15 percent applicant decline keeps climbing, or whether 278 new AI-for-science awards convince skeptics the money is finding better targets, is the number to watch as the 90-day agency deadline approaches.