Published: October 7, 2026 · Last updated: October 7, 2026
TL;DR: Chan Zuckerberg Biohub’s Virtual Biology Initiative just pulled in roughly $800 million in new commitments, $300 million from Meta, Google DeepMind, and Isomorphic Labs combined, and more than $500 million from the US government through the Department of Energy and the National Institutes of Health. Added to the $500 million Zuckerberg already put in, the initiative’s total backing is now about $1.8 billion. The goal is open biological data at a scale big enough to train AI models that can simulate how an actual human cell behaves, aimed at predicting and treating disease rather than shipping another chatbot.
Mark Zuckerberg’s science outfit just talked Google, Meta’s own research arm, and two federal agencies into writing checks for a project that has nothing to do with social media. The Chan Zuckerberg Biohub announced this week that its Virtual Biology Initiative picked up roughly $800 million in new funding on top of the $500 million Zuckerberg already committed, pushing the project’s total backing to about $1.8 billion.
The new money splits two ways. Meta, Google DeepMind, and Isomorphic Labs, the Alphabet-owned drug discovery company that grew out of DeepMind’s AlphaFold research, are putting in a combined $300 million. The US government is contributing more than $500 million through the Department of Energy and the National Institutes of Health. That federal piece is the detail worth noticing: this is real taxpayer science money attaching itself to a Silicon Valley-led AI project, not another round of venture capital chasing a model release.
What all of that money actually buys is data, and a lot of it. Biohub wants to build what researchers are calling a universal virtual cell, an AI model trained on enough gene expression, protein structure, and live-cell imaging data that it can simulate how a real human cell behaves under different conditions. Point it at a disease mechanism or a drug candidate, and in theory it predicts the outcome computationally instead of waiting months on a lab experiment and a grant cycle. Isomorphic Labs already does something adjacent to this on the drug discovery side, which makes its involvement here less of a stretch than it sounds.
Biology has a different problem than language models ever did. ChatGPT and its competitors took off because the internet already handed the industry trillions of words to train on. Cell-level biological data doesn’t exist anywhere near that scale, and what does exist sits locked inside pharmaceutical companies, individual university labs, and hospital systems with no particular reason to share it. Biohub’s pitch is that if enough well-funded players commit to generating and publishing data in the open, AI labs finally get their dataset problem solved the way text got solved a decade ago, except this time for the inside of a cell.
The open part is the detail worth watching closest. Meta and Google DeepMind don’t typically fund things without an eye on what they get back, and a company that size sitting inside a consortium that shapes how biological training data gets structured and released is not a neutral position to hold. If the datasets genuinely stay open to any lab or startup that wants them, this starts to look like a Human Genome Project for AI, shared infrastructure the whole field builds on. If access quietly tilts toward whoever funded it first, the open-science framing ends up being a better story than the outcome.
Related: This is the latest example of AI’s biggest names treating fundamental infrastructure, not just bigger models, as the real competitive edge, a pattern also showing up in Field AI’s $700 million round to build foundation models for robot bodies and in Instinct’s $1 billion raise at a $10 billion valuation.
Bottom Line: A $1.8 billion initiative to simulate a human cell sounds like science fiction until you remember these are the same three companies already betting hundreds of billions on data centers for the exact same underlying idea, that more and better data wins. The difference is this particular data doesn’t exist yet and has to be built from scratch, which puts any payoff on a timeline no quarterly earnings call can measure. Watch this less for a near-term breakthrough and more for whether “open” survives contact with three of the most commercially aggressive companies in tech.
Follow Teck Hustlers: Facebook · X · Instagram · LinkedIn · TikTok · Pinterest
