Vivodyne HIVE Targets AI Drug Discovery’s Human Data Gap
Vivodyne says HIVE robotic labs grow 20 human tissues and generate causal biological data to improve AI drug discovery and reduce clinical failure rates.
Summary
On August 19, 2026, University of Pennsylvania spinout Vivodyne said its modular HIVE robotic labs can grow 20 human tissue types, autonomously dose and monitor them, and generate causal data unavailable from animal tests or isolated cells and proteins. CEO and cofounder Andrei Georgescu says current AI cannot capture human biology’s complexity: 90% of drugs effective in animals that enter clinical trials fail to gain human approval, while only a handful of AI designed drugs have reached trials and one has reached Phase III. Anthropic CEO Dario Amodei recently questioned familiar AI cancer claims previously advanced by himself, OpenAI’s Sam Altman and Google DeepMind’s Demis Hassabis. Nobel Prize winning AlphaFold has produced no new drug. In February, Isomorphic Labs delayed trials originally planned for 2025 until the end of 2026, saying discovery needs accurate models spanning many biochemical properties and interactions.
Founded in 2021 after Georgescu earned his bioengineering PhD, Vivodyne reports 94% predictive accuracy for liver toxicity, 96% behavioral matching for airway tissue, and 100% concordance for bone marrow tests involving 20 chemotherapy drugs. Having raised just under $80 million in two Khosla Ventures led rounds, it opened what it calls the world’s largest human data center outside San Francisco last week. Georgescu says it delivers twice the throughput of all United States animal trials. Multiple unnamed major pharmaceutical partners are using it to identify likely winners before clinical trials that typically cost tens of millions of dollars, seeking the confidence automakers have before NHTSA tests but drugmakers lack before FDA decisions.
A Nature Methods study published last month found no clear scaling laws for generative models trained on existing cellular data. HIVE instead tracks hundreds of thousands of experiments linking stimuli to changes in diseased tissue, data Georgescu expects could support reinforcement learning, causal biological models and the search for combination therapies targeting multiple pathways.
Positives
- HIVE grows, doses and monitors 20 human tissue types autonomously, producing causal biological data rather than static cellular snapshots.
- Vivodyne reports 94% liver toxicity accuracy, 96% airway tissue matching and 100% bone marrow concordance across 20 chemotherapy drugs.
- Just under $80 million from two Khosla Ventures led rounds funded Vivodyne’s expansion since its 2021 University of Pennsylvania spinout.
- Twice the throughput of all United States animal trials is already being achieved at Vivodyne’s new facility, according to Andrei Georgescu.
- Multiple major pharmaceutical companies are working with Vivodyne to screen candidates before clinical trials costing tens of millions of dollars.
Risks & concerns
- 90% of drugs effective in animal testing that enter clinical trials fail to receive regulatory approval for humans.
- Only a handful of AI designed drugs have entered human trials, with just one reaching Phase III.
- Nobel Prize winning AlphaFold has not yet produced a new drug despite advancing understanding of biological building blocks.
- Isomorphic Labs moved trials originally planned for 2025 to the end of 2026.
- Nature Methods found no clear data scaling laws for generative AI models trained on existing cellular datasets.