Tahoe Therapeutics, the Arc Institute and the Chan‑Zuckerberg Biohub committed multimillion‑dollar resources to build a perturbation‑heavy single‑cell dataset intended to power virtual cell models. Tahoe will generate over 120 million single‑cell datapoints across ~225,000 drug–patient interactions using its Mosaic platform; partners said the public release will include broad perturbation diversity, cell types and patient contexts while providing an exclusive access period for collaborators. The dataset aims to address a core bottleneck for predictive cell‑state models—data scarcity—and to accelerate AI tools that predict transcriptomic responses to drugs and gene perturbations. For modelers and drug hunters, the resource promises richer training data to improve generalizability of virtual cell predictions and to guide lead selection and biomarker choices.