Biohub, the United States Department of Energy and the National Institutes of Health have announced a coordinated 1.8-billion-dollar commitment of funding, data, computing and measurement technology to build open, AI-ready biological datasets — described by the partners as the largest coordinated commitment of its kind — toward predictive models that can simulate how living cells respond to interventions.
The parts matter more than the headline. The Energy Department will invest more than 500 million dollars over five years, through its Genesis Mission and national laboratories, in cell research spanning data collection, laboratory measurement and imaging, modelling and computation. The health institutes will coordinate datasets, repositories and knowledge bases built with more than 500 million dollars of prior federal investment and work with Biohub to standardise them for model training. Biohub’s own 500-million-dollar Virtual Biology Initiative anchors the philanthropic side.
Industry is inside the tent rather than watching it. Google DeepMind, Isomorphic Labs and Meta are collectively investing 300 million dollars in the Virtual Biology Initiative’s technologies and multimodal datasets, NVIDIA is contributing computing infrastructure and expertise, and research partners include the Allen Institute, Broad Institute, Gladstone Institutes, the Human Cell Atlas, Human Protein Atlas and Wellcome Sanger Institute. Existing single-cell datasets cover on the order of a billion cells; the stated ambition is an order of magnitude beyond that, standardised enough for models to learn from across laboratories and countries.
The scientific claim should be read carefully. This is a data and infrastructure announcement, not a model release and not a treatment. Its premise — that biology lacks the large, clean, standardised datasets that made AI work in language and images, and that generating them deliberately is now the fastest route to predictive ‘virtual cells’ — is plausible and widely shared; whether the resulting models predict real cellular behaviour well enough to shorten drug discovery is the unproven step the money is buying the chance to test.
For morning readers, the significance is the coalition’s shape: federal laboratories, public health data, philanthropy and the three largest AI-in-biology companies agreeing to build a shared, substantially open foundation instead of competing private hoards. If the openness holds, including the reported detail that government-funded work carries no embargo, the dataset could become the field’s common ground. If it fragments, 1.8 billion dollars will have bought excellent silos.
