NIH's Bridge2AI Network for AI Health Science: A $7M Forecast Academic PIs Should Read Now

Academic PIs and NIH-funded researchers have a four-day-old signal worth acting on: on July 8, 2026, NIH posted a forecast for RFA-RM-28-010, the Bridge2AI Network for AI Health Science, an estimated $7 million initiative whose grants.gov listing points to a March 1, 2027 application target.

Forecasts are the quietest part of the federal funding calendar, and they are also where the best-prepared applicants win. A forecast is not a solicitation. It is NIH telling you, on the record, what it intends to fund and roughly when, precisely so that "potential applicants" have "sufficient time to develop meaningful collaborations and responsive projects" — the agency's own words on this listing. For a program built entirely around multidisciplinary teams, that lead time is not a courtesy. It is the competition.

What the RFA-RM-28-010 forecast actually commits to

The primary document is the grants.gov forecast at grants.gov/search-results-detail/363107, posted by the NIH Common Fund (CFDA 93.310) on July 8, 2026. Strip away the boilerplate and the numbers are specific. NIH anticipates making up to five awards to establish three to five Centers, with an estimated $7 million in funding, individual awards floored at $1 million and ceilinged at $7 million, under the U24 cooperative-agreement activity code. The estimated application due date is March 1, 2027, with an estimated award and project-start date of August 1, 2027, placing the money in fiscal year 2028. The award is administered out of the NIH Office of the Director rather than a single institute, which is the Common Fund's signature: programs it judges too cross-cutting for any one institute to own.

Two words in that paragraph carry weight. "Forecast" means the Notice of Funding Opportunity has not been published yet — the U24 NOFO is still coming, and terms can shift when it lands. "Cooperative agreement" means these are not hands-off R01s; U24 awards come with substantial NIH programmatic involvement, milestone governance, and cross-Center coordination obligations. If you have only ever run investigator-initiated grants, the U24 mechanism is a different operating model, and pretending otherwise is how strong science loses to better-organized consortia. The forecast is also explicit that this is one of two Stage 2 initiatives; the other will be run through a research contract rather than a grant, so RFA-RM-28-010 is the entry point for teams that want in through the traditional application route.

From AI-ready datasets to "the science of AI science"

Bridge2AI is a NIH Common Fund program, and this RFA is the back half of a deliberate two-stage arc. Stage 1, launched in 2022, spent roughly $130 million building something the biomedical field did not have: large-scale, ethically sourced, well-documented, AI-ready datasets, released as open resources along with the best practices for producing them. On January 29, 2026, the NIH Council of Councils approved the program's move into Stage 2, which splits into two initiatives. One will push those datasets toward products that attack biomedical grand challenges. The other — the one RFA-RM-28-010 funds — is stranger and, for the right team, more interesting.

The Network for AI Health Science is not primarily about building AI models. It is about building the discipline that judges them. The forecast describes a network of multidisciplinary Centers created "to advance the science of AI science" by developing "a framework for trustworthy, reproducible and explainable AI-enabled biomedical and behavioral research." Concretely, Centers are expected to evaluate AI models and products built elsewhere in the Bridge2AI program and beyond; convene expert panels and roundtables; support pilot research; run multidisciplinary cross-training; and disseminate what they learn. The listed deliverables are telling: best practices for AI methods and trustworthy model-building, frameworks for assessing reproducibility, replicability and transparency, and — in a phrase that signals where NIH thinks the field is heading — "guiding principles for agentic autonomous labs of the future."

That framing should reset how prospective PIs scope a proposal. NIH is not shopping for another benchmark leaderboard or a single clever architecture. It is funding the connective tissue of a nascent field: evaluation standards, reproducibility infrastructure, and workforce training that a fragmented research community cannot produce on its own. A competitive Center will look less like a lab and more like a standing institution — one whose output is other people's ability to trust AI results.

Who should be assembling a team right now

Eligibility on the forecast is broad in the usual NIH way — the applicant list runs from tribal governments and U.S. territories to faith- and community-based organizations, regional organizations, and, notably, non-domestic (foreign) institutions. In practice, the realistic applicant is a research university or academic medical center with genuine depth across three areas that rarely sit under one roof: applied machine learning, a biomedical or behavioral science domain, and the harder-to-staff disciplines of AI evaluation, ethics, metrology, and reproducibility. The Centers are explicitly meant to strengthen "the American biomedical and behavioral research AI workforce," so a serious training and cross-disciplinary mentoring component is not decoration — it is scored substance.

Because this is a network, NIH is buying interoperability. A Center that can only evaluate its own models is worth less than one that can define metrics other groups adopt. If you are a PI weighing whether to compete, the honest test is whether you can credibly convene the field, not just contribute to it. That is precisely the capability the forecast's lead time is designed to let you build before the NOFO drops.

Why the long runway is the real story

The gap between a July 2026 forecast and a March 2027 application target is roughly eight months, and that is not slack. Consortium-scale U24 proposals live or die on partnerships that take a full grant cycle to negotiate — data-use agreements, subaward budgets, letters of support from the domain and methods sides, and governance plans for a multi-institution Center. Teams that wait for the formal NOFO to start recruiting collaborators will be assembling a coalition on the same timeline they should be writing specific aims.

The prudent move now is threefold. First, treat March 1, 2027 as a planning anchor while watching for the actual U24 NOFO, since a forecast's terms — award counts, ceilings, even the mechanism — can change on publication. Second, get on NIH's radar through the listed program contact, Bridge2AI@od.nih.gov, and read Stage 1's public outputs closely; a Network Center that cannot speak fluently about the datasets and best practices it will help evaluate will not be credible. Third, map your own institution honestly against the "science of AI science" mandate and decide now whether you are a lead applicant, a subaward partner, or an expert-panel contributor. All three roles are real, and knowing which one you are shapes every conversation you have between now and next spring.

For broader context on how AI-and-health funding is holding up across NIH despite budget pressure elsewhere, Granted's evergreen guide to AI healthcare research grants in 2026 is a useful map of where the dollars are concentrating and which mechanisms are still open.

The next step for a prospective Center PI

If Bridge2AI's Network for AI Health Science fits your portfolio, do not wait on the NOFO to start the parts you already control. Track the live NIH AI-and-biomedical opportunities alongside RFA-RM-28-010 and start building your target list today: search active AI health science solicitations on Granted. Use the eight-month runway the way NIH intends — to walk into the March 2027 deadline with a coalition already built, not a partner search just beginning.