Biohub and US agencies commit $1.8bn to AI biology data

Biohub, the U.S. Department of Energy and the National Institutes of Health are committing $1.8bn to data and tools for AI models that predict cell behaviour. The initiative also includes $300M from Meta, Google DeepMind and Isomorphic Labs combined. The announcement describes a research and data programme; the source does not identify a legal or regulatory proceeding.

The investment is intended to address a constraint in computational biology: models need much larger and more consistent datasets to predict how cells behave. Biohub plans to make the resulting datasets a public resource, with commercial funders receiving an initial exclusive-access period. The government-funded work will not carry that restriction, according to the source.

Federal funding follows separate routes

The U.S. Department of Energy plans to spend more than $500M over five years on measurement in laboratories, modelling and computing. Its contribution will run through the Genesis Mission science initiative. The department expects the work to use exascale supercomputers, X-ray and neutron scattering, cryo-electron microscopy and self-running laboratories.

The National Institutes of Health will contribute datasets and repositories developed with more than $500M in earlier federal funding. Biohub will standardise those materials for AI training. The announcement does not describe a new grant award or set out a formal regulatory status for this contribution.

Existing commitments shape the programme

Biohub is adding its own $500M commitment, pledged in April as part of the Virtual Biology Initiative. The source allocates $400M of that amount to new tools for measuring cells, while $100M is designated for research outside Biohub. Those allocations sit alongside the wider commitment described in the announcement.

The participating research organisations also include the Allen Institute, the Broad Institute, the Gladstone Institutes and the UK Wellcome Sanger Institute. The Human Cell Atlas and Human Protein Atlas consortia are also involved. Nvidia is to provide computing and software, while Renaissance Philanthropy is helping to raise additional funds.

Data access terms remain central

Biohub says the datasets will ultimately be public, while commercial funders will receive one year of exclusive access first. Alex Rives, Biohub’s head of science, told Axios that the access period is intended to give commercial participants an incentive to contribute. The source does not specify how access will be administered or what happens if a dataset is delayed.

The government-funded work is expected to have no comparable restriction, Rives told Reuters. Biohub plans to approach drug companies and philanthropies for further participation. No response from a company or government agency to criticism is included in the supplied material, and the source reports no such criticism.

Current cell datasets contain hundreds of millions of cells, Rives told Reuters; he said an accurate model would require billions and later trillions. The partners aim to produce a first dataset in about a year and accurate predictive models within five years. The timetable is an aim attributed to the partners, rather than a confirmed delivery record.

The source leaves open how the public resource will be governed and how the exclusive-access period will work in practice. It also does not specify the full allocation of the combined $300M commercial contribution. The next concrete step is the planned first dataset, which partners target in about a year. Progress against the five-year model aim will show whether the programme delivers the scale and predictive accuracy it seeks.