The Prize Fund

The problem

The federal government's default way to fund science is the grant: propose a plan, get reviewed, get paid if approved, whether or not the work ever produces anything. In fiscal year 2025, 18.5% of early-career researchers who applied for an R01-equivalent grant, the National Institutes of Health's most common research award, received one. Across NIH generally, success rates run 15% to 30% depending on institute. That means, in the typical case, four out of five carefully written, formally reviewed proposals produce nothing but the hours spent writing and reviewing them.

Of the money that is awarded, a meaningful share never reaches the bench at all. NIH's own historical data put the average indirect cost rate, the share of a grant that goes to institutional overhead rather than the research itself, at 27% to 28% over time; a review of $9.85 billion across 5,143 projects found a mean indirect rate of 47.9%. A 2025 policy attempt to cap the rate at 15% triggered immediate litigation from research universities, which tells you how load-bearing that overhead revenue has become to institutions that are not, on paper, in the overhead business.

None of this means grants don't work; most basic science genuinely requires funding an open-ended attempt with no guaranteed result, and that is a defensible thing for the government to pay for. But grants are close to the only instrument the federal government uses, for goals that are not open-ended at all: goals with a known, checkable finish line, drive a car across a desert without a person in it, land a rover, sequence a genome faster and cheaper than last year. For goals like that, there is a second instrument, used sparingly, that pays only for the result: the prize.

The proposal

For federal research goals with a clear, verifiable finish line, expand the use of inducement prizes, cash paid only to whoever actually achieves the goal, funded in part by redirecting a small share of the grant-overhead spending that currently pays for attempts regardless of outcome.

How it would work

1. Prizes for finish-line goals, grants for everything else. Reserve prizes for problems with an objectively checkable outcome, verifiable and unambiguous, not general research direction. "Build a vehicle that autonomously completes this route" is a prize. "Investigate the mechanism of this disease" is a grant. Confusing the two is why past federal prize use has stayed a minor sideline rather than a core tool.

1. Purses that scale to the problem, and are announced years ahead. Modern federal inducement prizes typically run $250,000 to $2 million, occasionally up to $10 million for larger technical leaps. The purse should be announced with enough lead time for competing teams to actually build something, since the entire mechanism depends on competitors spending their own money first.

1. No overhead on a prize, because the government pays after the fact. A grant's overhead is charged against a plan; a prize is paid against a delivered result, so there is nothing for an indirect-cost rate to attach to before the goal is met. Fund a portion of new prize purses by redirecting a small, fixed share, on the order of low single-digit percent, of current federal grant overhead spending into prize purses at the same agencies.

1. Every agency gets standing authority, not a one-off statute. Congress first authorized modern federal inducement prizes in 2003, but authority has stayed narrow and agency-specific ever since, which is why prizes remain a minor tool two decades later. Give every research agency standing, general authority to run inducement prizes for verifiable goals within its own mission, without needing new legislation each time.

1. Open to any team on earth. A grant has to go to an eligible American institution; a prize only has to go to whoever crosses the line. Let a lab in Nairobi or Warsaw compete for an American purse, with the result, the data and the licence coming home to the agency that posted it. Buying the world's best answer to a stated American problem is cheaper than funding every attempt, and it is the kind of leadership other countries can see.

1. Publish the ledger. Every prize purse, every entrant's disclosed spending where teams choose to share it, and every dollar of resulting private follow-on investment, tracked publicly and compared over time to grant-funded programs pursuing comparable goals. If prizes consistently produce more result per federal dollar on finish-line problems, expand their share; if they don't, this catalogue's answer is to say so and stop.

The numbers

What the current default costs in unfunded effort. An 18.5% success rate for early-career R01-equivalent grants means, in a typical funding cycle, roughly four-fifths of formally reviewed proposals are unfunded, after already consuming real researcher time and federal review capacity, producing nothing.

What overhead consumes before research starts. A historical average indirect cost rate of 27% to 28%, and a measured 47.9% mean across a $9.85 billion, 5,143-project sample, is money committed before a single experiment runs. A prize purse carries none of this, because it is paid on delivery.

What a small prize has bought before. DARPA's first Grand Challenge, in 2004, offered $1 million for an autonomous vehicle to complete a desert route; no team finished, and the prize went unpaid. A year later, at $2 million, Stanford's team won by finishing in under seven hours. The Ansari X Prize, a $10 million purse for a reusable private spacecraft, drew 26 teams that collectively spent more than $100 million pursuing it, a roughly ten-to-one leverage ratio of private spending to prize money, and the companies formed to compete for it went on to attract more than $1 billion in follow-on private investment. A purse a fraction of the size of a single major grant program helped start an industry.

The honest objections

"Most science doesn't have a finish line you can specify in advance. This only works for a narrow slice of problems." True, and it is the whole reason this proposal does not argue for replacing grants generally. Basic research, where the point is discovering what the right question even is, cannot be prized, because there is nothing to verify in advance. This proposal is deliberately narrow: it expands a tool for the minority of federal research goals that already have a checkable finish line, and changes nothing about how the rest of science is funded.

"Prizes only reward whoever was already closest to the answer, or richest enough to gamble on getting there first, so you're subsidizing winners who didn't need it." Also largely true, and it is the mechanism's real cost, not a myth about it. Prizes systematically favor well-capitalized entrants who can absorb the risk of losing everything they spent, which grants, paid up front, do not require. The honest answer is that this is the tradeoff for a ten-to-one private leverage ratio: the government captures far more total effort per dollar of its own money, but it does so by shifting risk onto competitors who might get nothing, which is a legitimate reason to keep prizes a supplement to grants rather than a replacement.

"Redirecting grant overhead into prizes just reopens the fight universities are already suing over." Fair, and this proposal is not the indirect-cost-rate fight, on purpose. It asks for a small, fixed redirection at the point new prize funding is created, not a general cut to institutional overhead rates, which is a live, separate, and much larger legal and political dispute this proposal takes no position on.

"Two decades of narrow prize authority suggests agencies don't actually want this tool, not that they lack permission." Partly true. Grants are administratively familiar and prizes are not, and standing authority alone will not change agency habits overnight. But the DARPA Grand Challenge and the Ansari X Prize both happened under exactly the narrow, case-by-case authority this proposal wants to make standing, which suggests the appetite exists when the legal friction is removed, not that it's absent.

Sources

• National Institutes of Health, Office of Extramural Research, R01-equivalent success rate data, FY2025, report.nih.gov
• NIH indirect cost rate historical averages and project-level analysis; Congressional Research Service, NIH Indirect Costs Policy for Research Grants (IN12516), congress.gov
• Congressional Research Service, Federally Funded Innovation Inducement Prizes (R40677), everycrsreport.com
• DARPA, Grand Challenge program history, darpa.mil
• XPRIZE Foundation and McKinsey & Company, Using Prizes to Spur Innovation, on Ansari X Prize team spending and follow-on investment
• Proposals № 025 (The Automation Ledger); № 029 (The Model Commons); № 030 (The Compute Reserve)