Women led 60.2% of terminated doctoral grants and 59.8% of canceled assistant professor projects in the 2025 NIH cuts. That number, from a PNAS study published last week, is the kind of failure signal that demands root-cause analysis, not just outrage.
My position: the data is accurate, the disproportion is real, and it reflects a pre-existing structural tilt in how NIH distributed funding, not a deliberate targeting of women or early-career scientists. The December 2025 Unified Funding Strategy is the correct engineering response. Whether NIH executes it well is a separate, important question.
Diagnose the Failure Mode
When a system breaks under stress, the break reveals where the system was weakest. The 2025 cuts terminated 2,291 grants worth $2.45 billion. Women held smaller grants with more unspent funds at the time of cancellation: 57.9% of terminated funding versus 48.2% for men. At Harvard, the median canceled grant for a woman investigator was $362,000. For men, $689,000.
Diego Oliveira, the study's lead author at the University of North Dakota, said it plainly: "Funding shocks do not affect all researchers equally. They interact with existing structural features of the system." That word, "existing," matters. The cuts did not create the gap. They exposed it.
Women were concentrated in training awards, F31 doctoral fellowships, T34 grants. They held 58% of predoctoral fellowships and 66% of training grants. These are the smallest, earliest-stage awards in the NIH portfolio. When you cut across the board, the people holding the smallest, most vulnerable grants lose the most proportionally. This is not conspiracy. It is math.
Donna Ginther at the University of Kansas called the paper "great" and noted that training grant terminations risk "derailing scientific careers just as they are getting started." She is right about the risk. But the risk existed before February 2025. The pipeline was already leaking.
Fix the Intake Valve, Not Just the Leak
NIH's spokesperson insisted the agency "allocated its full budget" and that "any other suggestion is false." That response is inadequate. Allocating a full budget through a system that funnels women into smaller, more precarious grants is not meritocracy. It is inertia wearing a lab coat.
The Unified Funding Strategy announced in December 2025 does something specific: it factors career stage into funding decisions and reduces score cutoffs for early-career applicants. If you have spent any time watching iterative design processes, you recognize this. Identify the variable that produces the unwanted outcome. Adjust the variable. Test again.
I will grant the critics a fair point: no policy adjustment reverses the damage already done to the 405 doctoral fellows whose F31 awards were terminated, many of them diversity awards. Those careers may never recover their trajectory. That loss is real and permanent.
But the argument that post-hoc correction is worthless because the damage is done leads nowhere useful. Every engineering fix is post-hoc. You do not redesign the O-ring before the failure teaches you the O-ring was the problem. The question is whether the redesign addresses the actual failure mode.
The Unified Funding Strategy does. It targets exactly the structural feature Oliveira identified: early-career researchers holding small, vulnerable grants get evaluated with career-stage context instead of competing head-to-head against senior investigators with 3 active R01s. That is a system-level correction.
The $6.3 billion in estimated lost economic output from these cuts is staggering. The 60% female share of early-career terminations is a clear signal. But signals are only useful if you read them correctly. Reading this signal as "the cuts targeted women" misidentifies the failure mode. The failure mode was a funding architecture that made women and junior scientists structurally fragile before anyone cut a single grant.
NIH now has a policy that, on paper, addresses that fragility. The engineering question is execution. Will review panels actually weight career stage? Will the score adjustments be large enough to matter? Those are testable, measurable outcomes. I want to see the data from the first 2026 funding cycle. The diagnosis is done. The fix is designed. Now build it and fly it.
