White House blueprint rewires $200 billion in US federal research funding

The biggest overhaul of American science in 80 years just dropped, and AI is at the center of it.

The blueprint

On July 21, 2026, the White House released a blueprint called “Science: A New Golden Age,” rewiring how the US distributes roughly $200 billion a year in federal research funding.

The direction is consistent throughout: less money flowing through big universities, more going directly to individual scientists, to agile ARPA-style organizations built for high-risk projects, and to AI-driven research. Federal agencies have 90 days to submit their implementation plans.

The stated goal is speed: accelerate discovery, beat China, and cut the bureaucracy that sits between a scientist and a result.

The critics’ picture

Critics see a different outcome from the same document. Universities stand to lose the funding that underwrites labs, PhD programs and decades of basic research, the slow institutional substrate that produced much of modern science. Meanwhile, “AI-driven science,” the blueprint’s favored beneficiary, is still more promise than product.

Nature calls the plan the most extensive restructuring of US research since the 1940s, which is exactly the scale of both the opportunity and the risk. The postwar university-centered funding system being rewired here is the same one that built the modern research world.

A generational bet

Strip away the rhetoric on both sides and the blueprint amounts to a wager: that AI plus independent scientists will out-discover the university system. If AI-assisted research delivers on its promise, direct funding to agile teams could genuinely compress discovery timelines. If it does not, the US will have defunded proven institutions to pay for an unproven model.

The 90-day implementation clock is now running, and every research university and federal agency in the country is drafting its answer. Genius move or a $200 billion gamble: the next few years of American science will be the experiment that decides.