Google, Meta and U.S. government to invest $1.8 billion in a "virtual cell"

Biohub, the U.S. government and major technology companies are expanding the Virtual Biology Initiative, which aims to create AI models capable of predicting the behavior of human cells and other biological systems. The total resources announced by initiative participants have reached $1.8 billion.

This is stated in a Biohub release.

Google DeepMind, Isomorphic Labs and Meta have joined the project, as well as the U.S. Department of Energy and the National Institutes of Health (NIH) on the part of the American government.

Google DeepMind, Isomorphic Labs and Meta will jointly invest $300 million in creating technologies and multimodal datasets needed to train models that can predict the functioning of living cells.

The U.S. Department of Energy will allocate more than $500 million to the project over five years. The funds will go to laboratory research, biological data collection, modeling and computing power. The work is planned to use American supercomputers, electron microscopy and automated laboratories.

Biohub itself previously pledged to invest another $500 million. Of this amount, $400 million is intended for the development of new technologies for measuring, visualizing and modifying biological systems, and $100 million for funding research outside the organization.

The U.S. National Institutes of Health will also connect biomedical data sets and research infrastructure created with more than $500 million of previous federal investments to the initiative. Therefore, the declared $1.8 billion includes not only new monetary investments, but also already funded data, computing resources and technologies.

The main goal of the project is to form a large-scale open database of biological data on which researchers can train AI cell models. In the future, such models should allow conducting some experiments not in the laboratory but in a digital environment.

For example, scientists will be able to model how different cell types respond to a drug, genetic modification or other impact, and then select the most promising candidates for real laboratory and clinical studies.

The NIH expects that this approach will make it possible to find potential drug targets faster and shorten the path from fundamental discovery to the development of new therapies.

The Virtual Biology Initiative was launched by Biohub in April 2026. The organization initially allocated $500 million for it and announced its intention to create an open data foundation for the development of so-called predictive models of the human cell.

The collaboration also includes the Allen Institute, Broad Institute, Human Cell Atlas, Human Protein Atlas, Wellcome Sanger Institute and other scientific organizations. Nvidia will provide the project with computing infrastructure, specialized software and technical expertise.

The data obtained under the program is planned to be made available to the scientific community. The organizers expect that common standards will allow combining research results from different laboratories and using them to train more complex biological AI models.