A $1.8bn collaboration involving Biohub, Google DeepMind, Meta and US agencies aims to build the biological data needed to train AI models that can predict how cells behave. The first step is a broad map of cell biology, with the longer-term hope of testing some experiments virtually before researchers spend time and money in the lab.
Watch Desk analysis
What happened
Axios reports that the effort brings together Biohub, the Department of Energy, the National Institutes of Health, Google DeepMind, Isomorphic Labs, Meta and scientific organisations to create and standardise data for what Biohub calls a “universal virtual cell”. The project combines new funding, existing datasets, computing and biological measurement technology.
The Department of Energy plans to invest more than $500 million over five years. The NIH will contribute datasets and other resources from more than $500 million in previous federal investment. Google DeepMind, Isomorphic Labs and Meta are investing a combined $300 million, while Biohub has already committed $500 million.
Why it matters
AI models can only predict what their evidence lets them learn. Biology poses a particular challenge: much of the data needed to connect digital predictions with what happens in living cells still has to be measured in the physical world. The project is an attempt to build that missing foundation, not a finished virtual cell ready to run experiments on its own.
Commercial partners will get one year of exclusive access to the data they help develop before it is shared publicly. That arrangement is meant to encourage companies to contribute while making the resulting resource available to wider scientific research later.
Our read
The eye-catching prize is virtual experimentation; the less glamorous work of collecting and standardising data is what must make it plausible. Axios reports that Biohub’s head of science, Alex Rives, expects researchers to train models and assess their capabilities within a year of the first large-scale dataset. That will be an early test of the project’s premise, not proof that a model can reliably predict the behaviour of an entire living cell.
What to watch
- When the first large-scale dataset is ready and what it covers.
- Whether models trained on it make predictions that hold up against biological evidence.
- How the one-year data embargo works in practice, and when the first data becomes broadly available.
- Whether researchers can identify which additional measurements improve model performance.
Discussion spark: Should commercial partners get a year’s head start on data built through a major public-private research effort, or should publicly funded biology data be shared from the outset?
Sources and evidence
- "The beginning of a new scientific paradigm": Zuckerberg's Biohub, U.S. and Google build virtual cell (7 October 2026, 13:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.