US Government, Meta and Google Join Biohub's $1.8B Open Dataset Effort for Biology AI

What Happened

The numbers first: Biohub announced this on Wednesday (October 7 ET), first reported by Reuters. The $1.8 billion breaks down as follows: Meta, Google DeepMind and Isomorphic Labs jointly invest $300 million; the Department of Energy (DOE) puts more than $500 million over five years into laboratory measurement, modeling and computation; the National Institutes of Health (NIH) hands over datasets and repositories built with more than $500 million in earlier federal funding, for Biohub to standardize as AI training data; plus the $500 million Biohub itself put in back in April. The $300M corporate investment and the DOE's $500M+ are new commitments; roughly another $1 billion is pre-existing. The $1.8B total combines old and new money.

The money funds the Virtual Biology Initiative: systematically measuring how cells respond to far more combinations of conditions than scientists have studied, then training predictive models that could compress drug-development timelines (currently measured in years). Biohub head of science Alex Rives says today’s cell datasets cover hundreds of millions of cells, while accurate predictive models need billions and eventually trillions — "We need to capture the language of biology, we need to capture the language of the cell. And that doesn’t exist today." The toolkit includes spatial transcriptomics (mapping molecular activity inside intact tissue) and large-scale screens recording how cells respond to environmental changes — much of it "never generated in a coordinated way" before. Rives’ timeline: work that would normally take decades, compressed into five years; first datasets ready in about a year, accurate predictive models within five years.

Key Facts

1. Of the $1.8B, about $800M is new commitments: The $300M from Meta / DeepMind / Isomorphic plus the DOE's $500M+ over five years is new; Biohub's own $500M from April and the NIH datasets built with $500M+ in earlier federal funding are pre-existing. The $1.8B is the sum of all four.
2. Funding companies get an embargo period: The data will eventually be released publicly, but the companies funding it get an embargo period — they can work on the data before it becomes a public research resource. Rives says this is how Biohub brings private money into "open science." The government-funded parallel work carries no such condition.
3. First data in a year, predictive models within five: Rives says this kind of work "would normally take decades." The data has to be produced with new techniques like spatial transcriptomics, "never generated in a coordinated way" before, scaling from hundreds of millions of cells to billions and trillions. Biohub still plans to bring in pharma companies and philanthropies.
4. Other AI companies' biology efforts: Anthropic recently expanded into biology (building a wet lab), and the OpenAI Foundation launched a $125M+ grant program specifically for biological and medical AI datasets.

Context

Biohub is the philanthropic organization of Mark Zuckerberg and Priscilla Chan (under the Chan Zuckerberg Initiative). The Virtual Biology Initiative launched on April 29 of this year, when Biohub committed $500 million itself: $100 million for external research institutions doing global data collection, $400 million for internal technology (next-generation imaging, molecular engineering, data infrastructure); the Broad Institute (MIT / Harvard), the Allen Institute, the Arc Institute, the Wellcome Sanger Institute, the Human Cell Atlas, the Human Protein Atlas and NVIDIA were already on board. Today's news expands that April $500M — the US government plus Meta and Google DeepMind join, taking the total to $1.8 billion.