Biohub expands virtual biology initiative with $1.8 billion in combined resources

Biohub has announced an expansion of its Virtual Biology Initiative involving the US Department of Energy, the National Institutes of Health and several AI companies. The combined commitment totals $1.8 billion across funding, data, computation and new measurement technology. That wording matters: it is not a claim that the initiative has received $1.8 billion in newly transferred cash.

The Biohub announcement describes a coordinated effort to build models that can predict how cells respond to interventions across different cell types and conditions. It connects measurement, shared biological data and computing resources. These are research goals and infrastructure commitments, not evidence that a universal predictive cell model already exists.

The commitments cover different kinds of resources

DOE will invest more than $500 million over five years in lab measurement, modeling and computation. Biohub says that work will include AI-ready open data and access to capabilities across the national laboratories. The measurement and computing components are intended to help researchers produce the information and models needed for the initiative.

NIH's contribution has a different basis. It will coordinate relevant datasets, repositories and knowledge bases developed through more than $500 million in prior federal investment. The figure describes the investment that created those resources, rather than an additional $500 million grant announced here. Biohub plans to help standardize that material for AI training.

Google DeepMind, Isomorphic Labs, and Meta are collectively investing $300 million in the Virtual Biology Initiative. Biohub's founding $500 million commitment anchors the effort. Its announcement allocates $400 million toward new measurement technology and $100 million toward external research. Keeping those categories separate helps explain what the headline total represents.

Better measurements are central to the plan

The initiative, first announced in April 2026, seeks to combine biological observations at several scales. Biohub describes cryo-electron tomography for near-atomic views inside cells, microscopy spanning very large numbers of living cells and engineering methods that perturb biology. The goal is to observe both structure and response, rather than train models only on descriptions of biology.

That makes measurement a substantial part of the AI effort. A prediction about how a cell changes after an intervention needs suitable observations against which researchers can evaluate it. Generating larger models cannot, by itself, establish whether a prediction holds across another cell type or experimental condition. The announcement lays out resources for addressing that research problem; it does not report a completed validation across all of those settings.

DOE's role includes the Genesis Mission, national-laboratory computing and facilities such as X-ray and neutron scattering, electron microscopy and tomography. NIH's BioGenesis work brings existing data infrastructure, including NLM and NCBI resources and NIH Common Fund atlases. Those contributions connect the proposed modeling work to established measurement and data programmes.

Shared data needs shared identifiers and standards

Biohub also describes plans for common identifiers, shared standards and a single access point for the initiative's resources. These are practical requirements for combining datasets that were produced for different research questions. A model developer needs to know what an observation represents and how it relates to other measurements before treating the records as interchangeable training examples.

The announced partners include the Allen Institute, Broad Institute, Gladstone Institutes, Human Cell Atlas, Human Protein Atlas and Wellcome Sanger Institute. NVIDIA is providing computing, software and technical support, while Renaissance Philanthropy is helping expand funding. Biohub points to earlier resources including Tabula Sapiens, OpenCell, Zebrahub, CELLxGENE and the CryoET Data Portal as part of its experience building biological datasets.

The announcement is therefore best understood as an expansion of a research programme and its supporting infrastructure. The next meaningful results will be measurements, usable shared datasets and models evaluated against biological experiments, not the size of the funding headline alone.