US Government, Google and Meta Back $1.8B Biohub AI Project

The US government and major technology companies are joining forces with nonprofit Biohub on a $1.8 billion effort to build large open datasets for Artificial Intelligence research in biology. The initiative aims to generate and organize biological data at a scale that could help AI models better understand how cells behave and potentially accelerate drug discovery.

Meta, Google DeepMind and drug-discovery company Isomorphic Labs are jointly investing $300 million in the project. The US Department of Energy will contribute more than $500 million over five years, while the National Institutes of Health will coordinate datasets and repositories created through more than $500 million in earlier federal funding.

Biohub’s Virtual Biology Initiative

The investment will support Biohub’s Virtual Biology Initiative, a major effort designed to create standardized datasets for training AI systems used in biological research.

Biohub is a philanthropic venture backed by Meta CEO Mark Zuckerberg and his wife, Dr. Priscilla Chan. The organization previously committed $500 million to the project in April.

With the latest commitments, total funding associated with the effort reaches about $1.8 billion. The goal is to create a large-scale biological data resource that can be used by researchers and AI developers to build predictive models of biological systems.

The initiative focuses particularly on understanding how cells respond to different conditions. Scientists can use information about cellular responses to train models that predict what might happen when biological systems are exposed to new environments, treatments or other changes.

Why AI Biology Data Is Important

Artificial intelligence has become increasingly important in scientific research, but the effectiveness of AI models depends heavily on the quality and quantity of the data used to train them.

In biology, researchers have generated enormous amounts of information about genes, proteins and cells. However, much of that information has been produced through separate experiments using different methods and conditions.

That fragmentation makes it difficult to create models capable of accurately predicting biological behavior across many situations.

Biohub’s initiative is intended to address that problem by generating measurements in a more coordinated way and standardizing existing datasets for AI training.

Alex Rives, Biohub’s head of science, said current cell datasets contain information on hundreds of millions of cells, while accurate predictive models could eventually require data from billions or even trillions of cells.

$300 Million Investment From Google and Meta

Meta, Google DeepMind and Isomorphic Labs are jointly investing $300 million in the initiative.

The participation of major technology companies highlights the growing interest in applying AI to biological research. Google DeepMind has already developed AI systems for scientific applications, while Isomorphic Labs focuses on using AI for drug discovery.

Meta’s involvement also gives the initiative a connection to one of the world’s largest technology companies and its extensive AI research operations.

The companies’ investment is expected to help fund the generation of new biological measurements and the development of datasets that can eventually be used by researchers outside the participating organizations.

US Government Commits More Than $500 Million

The US Department of Energy will invest more than $500 million over five years in laboratory measurement, modeling and computational work related to the initiative.

The National Institutes of Health will also play a major role by coordinating datasets and repositories created through more than $500 million in earlier federal funding.

Biohub plans to standardize these datasets so that they can be more effectively used for AI training.

The government contribution is important because large-scale biological research can require expensive laboratory equipment, computing resources and coordinated scientific infrastructure. Public funding can help create data that would be difficult for a single company or research group to generate independently.

First Dataset Expected in About One Year

Biohub’s head of science said the first dataset from the initiative is expected to be ready in about a year.

The broader objective is to compress work that would normally take decades into approximately five years. Biohub expects that the resulting data could eventually support accurate predictive models of biological systems.

The ambitious timeline reflects the rapid progress in AI research and the increasing availability of technologies capable of measuring biological activity at much larger scales.

However, creating large datasets is only one part of the challenge. Researchers must also ensure that the measurements are reliable, standardized and sufficiently diverse for AI models to make useful predictions.

How Scientists Will Generate the Data

The initiative will use several experimental techniques to collect information about biological systems.

One important technology is spatial transcriptomics, which allows scientists to map molecular activity within intact tissue. This can provide information about where different biological processes are taking place rather than simply measuring activity across an entire sample.

The project will also use screening techniques that measure how cells respond to changes in their environment.

Combining information from these approaches could give AI models a more detailed picture of cellular behavior. Instead of learning only from isolated observations, models could potentially identify relationships between different biological conditions and cellular responses.

Biohub Wants to Capture the “Language” of Cells

Biohub scientists describe the project as an effort to capture what they consider the language of biology and the cell.

The idea is that sufficiently large and detailed datasets could allow AI systems to learn patterns that are difficult for humans to identify manually. Researchers could then use those models to predict how biological systems might behave under conditions that have not yet been experimentally tested.

This could potentially reduce the number of experiments needed during some stages of research and help scientists identify promising directions more efficiently.

However, AI predictions would still need to be tested experimentally. Biological systems are highly complex, and a prediction generated by a model would not automatically establish that a treatment or intervention works in the real world.

Private Companies Will Get Early Access

Although Biohub plans to eventually release the datasets publicly, companies that help fund the project will receive an initial period of access before the information becomes broadly available.

Rives said Biohub uses embargo periods for commercial funders, allowing participating companies to work with the data before it becomes a public scientific resource.

This arrangement is designed to attract private investment while preserving the project’s longer-term open-science objective.

The government-funded work being conducted in parallel will not have the same restrictions, according to Biohub.

Biohub also plans to approach pharmaceutical companies and philanthropic organizations for additional support as the initiative develops.

Why Open Biological Data Could Accelerate Drug Discovery

Drug development is a lengthy and expensive process that involves identifying promising biological targets, testing potential compounds and conducting extensive laboratory and clinical studies.

AI could potentially shorten some early research stages by helping scientists identify patterns in biological data and predict which compounds or biological targets are more likely to produce useful results.

Large datasets could make those models more capable because AI systems generally benefit from having access to broad and diverse training information.

Biohub’s goal is therefore not simply to create another biological database. It is seeking to build a much larger and more systematically organized resource specifically suited to training predictive AI models.

Anthropic and OpenAI Are Also Investing in Biology

Biohub is not the only organization seeking to build AI capabilities around biological data.

Anthropic has expanded its biology efforts and established a wet laboratory as part of its broader research strategy.

The OpenAI Foundation has also launched a grant program worth more than $125 million to support biological and medical datasets for AI research.

The growing number of initiatives suggests that biological data is becoming an important area of competition among AI companies and research organizations.

As AI models become more capable, access to high-quality scientific data could become an increasingly important factor in determining how useful they are for specialized research.

The Challenge of Scaling Biological Data

One of the biggest challenges facing the initiative is scale. Existing datasets may contain information from hundreds of millions of cells, but Biohub believes that billions and eventually trillions of observations could be needed for highly accurate predictive models.

Generating that volume of data will require large investments in laboratory automation, measurement technologies, computing infrastructure and scientific coordination.

Data quality will also be critical. More data does not automatically produce better AI models if measurements are inconsistent, biased or poorly documented.

Standardization is therefore a central part of the Biohub initiative. Making datasets compatible with one another could allow researchers to combine information from different experiments and use it more effectively.

Potential Impact on Scientific Research

If the initiative succeeds, its datasets could provide researchers with new ways to study biological systems and test scientific hypotheses.

• Drug discovery: AI models could help identify promising targets and potential treatments more efficiently.
• Cell research: Large datasets could improve understanding of how cells respond to different conditions.
• Predictive modeling: AI could be used to forecast biological responses before conducting additional experiments.
• Open science: Publicly released datasets could give researchers access to resources that would otherwise be expensive to create.
• Research collaboration: Standardized data could make it easier for laboratories and organizations to combine scientific findings.

The actual impact will depend on how accurately AI systems can learn from the data and how effectively scientists can validate their predictions.

Commercial and Scientific Interests Must Be Balanced

The project’s funding model highlights a challenge facing modern scientific research: how to attract private investment while maintaining broad public access to important discoveries and data.

Biohub’s approach allows commercial funders to receive an initial period of access to certain datasets while ultimately making the information available as a public scientific resource.

That structure could encourage technology and pharmaceutical companies to contribute funding without permanently restricting access to the resulting data.

The approach may also become a model for future collaborations between governments, nonprofits and technology companies seeking to build large scientific datasets.

What the $1.8 Billion AI Biology Push Means

The combined commitment represents a major investment in the intersection of artificial intelligence and biological research. It brings together government agencies, technology companies, a nonprofit research organization and potentially future pharmaceutical partners.

The immediate goal is to generate and standardize biological data at a scale that existing research programs have not achieved. The longer-term ambition is to use that data to develop predictive models that can help scientists understand biological systems and accelerate parts of the drug-development process.

The first dataset is expected in about a year, while Biohub is targeting a five-year period for the broader effort.

If the project can overcome the technical and scientific challenges involved in producing reliable data at massive scale, it could provide an important foundation for the next generation of AI-powered biological research.

Frequently Asked Questions

What is the Biohub Virtual Biology Initiative?

The Virtual Biology Initiative is a Biohub-led effort to generate and standardize large biological datasets for training AI models and supporting biological research.

How much money is being committed to the AI biology project?

The total investment associated with the initiative is about $1.8 billion, including commitments from Biohub, the US government, Meta, Google DeepMind and Isomorphic Labs.

How much are Meta, Google DeepMind and Isomorphic Labs investing?

The three companies are jointly investing $300 million in the initiative.

How much is the US government investing?

The Department of Energy will invest more than $500 million over five years, while the NIH will coordinate datasets and repositories created through more than $500 million in earlier federal funding.

When will the first Biohub dataset be available?

Biohub’s head of science expects the first dataset to be ready in about a year.

How could AI biology data help drug discovery?

Large biological datasets could help AI models identify patterns and make predictions about cellular behavior, potentially making some stages of drug research faster and more efficient.

Will Biohub’s datasets be publicly available?

Yes. Biohub plans to eventually release the datasets publicly, although commercial funders may receive an initial period of access before the data becomes a public scientific resource.

Which other AI companies are working on biology data?

Anthropic and the OpenAI Foundation are also pursuing biology-related initiatives, including laboratory research and funding programs focused on biological and medical datasets for AI research.