$1.8 Billion Virtual Biology Initiative Targets AI Models That Could Predict How Cells Behave

The U.S. Department of Energy (DOE), National Institutes of Health (NIH), Biohub and technology partners have expanded a nearly $1.8 billion Virtual Biology Initiative to build the biological data, computing infrastructure and measurement technologies needed for predictive AI models of living systems. The effort aims to move beyond analyzing biological data toward models that can predict how cells respond to drugs, disease and other interventions. The initiative combines new commitments with existing federal datasets and resources rather than representing a single $1.8 billion cash grant.

Why Scientists Want a “Virtual Cell”

Modern biology can measure genes, proteins and cellular structures in extraordinary detail, but understanding how all those components interact remains one of science’s hardest problems.

A cell is not simply a collection of independent biological parts. Thousands of molecular processes interact continuously, changing with cell type, environment, disease state and external interventions. Researchers can observe pieces of these processes experimentally, but testing every possible biological condition in the laboratory is slow and expensive.

The Virtual Biology Initiative is designed around a different possibility: building AI models that can learn from sufficiently large and diverse biological datasets and eventually predict how cells respond to specific changes.

In practical terms, a future virtual-cell model could allow researchers to pose biological questions digitally before deciding which experiments are worth performing in the laboratory.

The initiative is not yet delivering such a universal virtual cell. Its immediate objective is to generate and organize the experimental data required to make predictive models possible.

A Nearly $1.8 Billion Scientific Effort

The expanded initiative brings together Biohub, DOE, NIH, Google DeepMind, Isomorphic Labs, Meta and major scientific institutions.

Biohub’s original commitment is $500 million over five years. About $400 million is dedicated to developing technologies capable of measuring biology at much larger scales, while another $100 million supports research beyond Biohub itself.

The DOE plans to invest more than $500 million over five years through its Genesis Mission. That work will combine laboratory measurement, biological research, AI analytics, modeling and high-performance computing across the U.S. national laboratory system.

NIH will coordinate existing biomedical datasets, repositories and knowledge resources supported by more than $500 million in prior federal investment. Rather than simply creating another isolated database, NIH and Biohub intend to standardize relevant resources so they can be used more effectively for AI model development.

Google DeepMind, Isomorphic Labs and Meta are collectively investing $300 million in the Virtual Biology Initiative to develop technologies and multimodal datasets needed for predictive biological models.

Taken together, these commitments and resources form the nearly $1.8 billion initiative announced on October 7, 2026.

Building the Data Behind Virtual Biology

The central challenge is not simply computing power. AI models of biology need experimental data that captures what happens when living systems are changed.

The initiative therefore aims to expand measurements across more cell types, biological conditions and interventions than have previously been studied at comparable scale.

Among the technologies being developed or expanded are cryo-electron tomography, which can reveal cellular structures at near-atomic detail, advanced microscopy capable of examining very large numbers of cells, and molecular and cellular engineering tools that allow researchers to perturb biological systems and observe their responses.

The DOE’s national laboratories add another major component: access to exascale computing, advanced imaging and scattering facilities, cryo-electron microscopy and tomography, and autonomous laboratory capabilities.

This combination is important because biological AI requires a continuous loop between measurement, computation and experimentation.

A model can generate a prediction, researchers can test it experimentally, and the resulting measurements can then be used to improve the model.

From Biological Databases to Predictive Models

Existing projects already provide some of the infrastructure needed for this approach.

Biohub has developed or supported resources including CELLxGENE, the CryoET Data Portal, Tabula Sapiens, OpenCell and other large biological datasets.

The new initiative seeks to connect such resources through common standards, identifiers and interoperable data infrastructure.

That could address one of the less visible problems in AI-driven biology: valuable biological datasets often differ in format, experimental conditions and terminology, making them difficult to combine.

Creating standardized, AI-ready datasets could allow models to learn across experiments rather than treating every dataset as an isolated source of information.

What a Virtual Cell Could Eventually Do

If the underlying models become sufficiently accurate, researchers could potentially use them to simulate biological responses before conducting selected physical experiments.

For example, a predictive model could eventually be asked how a particular cell type might respond to a drug candidate, genetic intervention or environmental change.

That could help scientists prioritize experiments, identify unexpected biological responses and explore disease mechanisms that would otherwise require extensive laboratory testing.

Drug discovery is one potential application. Instead of testing large numbers of possibilities sequentially, researchers could use computational predictions to narrow the experimental search space.

But this remains a research objective, not an established replacement for laboratory or clinical experimentation.

Biological systems are exceptionally complex, and a model that performs well on one cell type or experimental condition may not necessarily generalize to another.

NIH and DOE Bring Different Strengths

The partnership is significant partly because the organizations contribute different pieces of the problem.

NIH brings decades of biomedical research, health datasets, repositories and biological knowledge infrastructure.

DOE contributes national laboratories, high-performance computing, advanced scientific instruments and large-scale experimental capabilities.

Biohub contributes biological measurement technologies, AI research and an existing ecosystem of datasets.

Technology companies and AI research organizations add expertise in large-scale machine learning, computing and biological modeling.

The objective is to combine these capabilities into an open scientific foundation rather than build isolated proprietary systems.

The Bigger Shift: Biology as a Computational Science

The initiative reflects a broader change in how scientists are approaching biology.

For decades, computational tools have helped researchers analyze biological measurements. The emerging goal is more ambitious: use AI and high-performance computing to construct models that can predict biological behavior.

That distinction matters.

A system that identifies patterns in an existing dataset can help researchers understand what has already been observed. A predictive biological model aims to estimate what could happen under conditions that have not yet been directly tested.

Reaching that level of reliability would require not only larger AI models, but better experimental measurements, carefully standardized datasets and rigorous validation.

Biohub describes the creation of a virtual cell as a major collective scientific challenge because no single institution can generate the necessary biological data at the required scale.

What Happens Next

The first stage of the initiative is therefore focused on infrastructure: generating new measurements, integrating existing datasets, improving computational resources and establishing standards that researchers can use across institutions.

The scientific test will come later.

Researchers will need to determine whether models trained on these increasingly large datasets can accurately predict cellular behavior in experiments that were not part of their training data.

If they can, virtual biology could become a powerful complement to laboratory science—allowing researchers to explore far more biological possibilities computationally before committing resources to physical experiments.

If they cannot, the resulting datasets and measurement technologies would still represent a major resource for conventional biological research.

For now, the $1.8 billion effort is best understood not as the arrival of a functioning digital human cell, but as a large-scale attempt to build the data and scientific infrastructure that could make predictive virtual cells possible.

Most Searched 5 FAQs

1. What is the Virtual Biology Initiative?

The Virtual Biology Initiative is a multi-institution effort led by Biohub with the U.S. Department of Energy, NIH and technology partners to generate large-scale, standardized biological data for developing predictive AI models of living systems.

2. How much money is being invested in the Virtual Biology Initiative?

The initiative represents a nearly $1.8 billion commitment involving funding, existing federal research resources, computing capabilities and biological measurement technologies. It is not a single $1.8 billion cash grant.

3. What is a virtual cell?

A virtual cell is the proposed concept of an AI-based predictive model capable of representing complex cellular processes and estimating how a cell could respond to drugs, genetic changes or other interventions. A fully reliable universal virtual cell does not yet exist.

4. How could virtual cells help drug discovery?

If sufficiently accurate models can be developed, researchers could use them to predict biological responses and prioritize laboratory experiments, potentially reducing the number of physical experiments required to investigate promising drug candidates.

5. Who is funding the virtual cell project?

Biohub committed $500 million, DOE plans more than $500 million over five years, NIH is coordinating biomedical resources built through more than $500 million in prior federal investment, and Google DeepMind, Isomorphic Labs and Meta are collectively investing $300 million.