About half the money in the terminated NIH grants had already been spent when they were cancelled

The clearest number in the 2025 National Institutes of Health grant terminations is not the headline dollar figure. It is the share of the money that had already left the building. In a brief report published in the Proceedings of the National Academy of Sciences on 23 March 2026, Diego F. M. Oliveira, Qian Huang, Teresa K. Woodruff and Brian Uzzi analysed 2,291 active grants terminated between 28 February and 22 August 2025. Those grants carried $5.08 billion in commitments. Of that, $2.45 billion was withdrawn, and about 52 per cent of the committed total had already been spent at the moment of cancellation.

That ratio changes what the exercise was. Cancelling a grant halfway through does not return the first half to the Treasury. It converts money already paid to salaries, animal colonies, sequencing runs, patient recruitment and equipment into an incomplete result. The saving is the unspent remainder. The cost includes the spent portion, which now buys a partial study rather than a finished one. Higher Ed Dive, reporting the paper on 26 March 2026, added that a further 1,534 grants were frozen rather than terminated.

The second underreported structure in the paper is who held the unspent halves. Women held 46.1 per cent of the terminated grants, close to parity. But at the point of cancellation women had 57.9 per cent of their award still unspent, against 48.2 per cent for men, meaning the disruption fell earlier in the life of women's projects. Median awards were smaller too, $940,000 for women against $1.4 million for men, which the authors attribute to a concentration of women's grants in training and early career mechanisms. Among terminated grants held by assistant professors, 59.8 per cent were led by women.

The provenance of the figures is itself checkable, which is unusual for a fast moving policy story. The analysis is built on public termination records compiled by Grant Witness, a database listing 2,295 terminated grants, and the authors deposited their analytical file, a single CSV of 444.3 kilobytes, in Zenodo on 30 January 2026 under a Creative Commons Attribution 4.0 licence, with the digital object identifier 10.5281/zenodo.18434242. Readers who dispute the counts can recount them.

The legal position around the money is more tangled than the arithmetic. On 16 June 2025 Judge William G. Young of the United States District Court in Boston held that hundreds of the terminations were arbitrary and capricious, and void and illegal, in litigation brought by the American Public Health Association, labour unions, a group of public health researchers and sixteen states led by Massachusetts. On 22 August 2025 the Supreme Court, by five votes to four, paused the part of that order restoring $783 million, holding that the district court lacked jurisdiction over the money and that the Court of Federal Claims, which handles monetary claims against the federal government, is the proper venue. Chief Justice John Roberts wrote in dissent that a court with jurisdiction to vacate the directives also had jurisdiction to vacate the terminations that followed. Justice Ketanji Brown Jackson wrote that the ruling turns an administrative law statute into "a gauntlet rather than a refuge".

The practical effect of that split is a change of clock. An administrative law finding, if it stands, is fast and applies to a class. A contract claim in the Court of Federal Claims is slow and is litigated grant by grant. Researchers who won on the reasoning were moved to a forum where the reasoning is not the question.

What happened to the people is being measured separately. STAT News, working with the MassINC Polling Group, emailed roughly 41,000 NIH funded scientists between 28 January and 18 February 2026 and received 989 responses from 45 states, Puerto Rico and Washington DC, with a margin of error of 3.3 percentage points. Twenty seven per cent had laid off lab members, 41 per cent had cancelled planned research, 66 per cent had counselled students to pursue careers outside academia and 53 per cent had suggested looking outside the United States. Of those whose grants were cut or delayed, 35 per cent reported full restoration of funding by the end of 2025.

The authors of the PNAS paper are explicit about what they cannot show. Their data cannot establish causal downstream effects, and they say conclusions about long term career impact will require systematic monitoring of affected cohorts over time. Nor is there yet a public accounting of how much of the $2.45 billion has been paid back out, restored by settlement, or is still sitting in litigation. The 52 per cent that was already spent is not coming back in any scenario.