Tahoe Therapeutics Raises $30 Million to Expand Single-Cell Data Production for AI Models
Tahoe Therapeutics has raised $30 million in Series A financing, increasing total funding to $42 million and valuing the Palo Alto-based company at $120 million.
The round was led by Amplify Partners, with Databricks Ventures, Wing Venture Capital, General Catalyst, AIX Ventures, Mubadala Ventures, Civilization Ventures, and Conviction. Capital will be used to scale the company’s Mosaic platform, designed to generate large-scale single-cell datasets for training AI models in oncology drug discovery.
Tahoe Therapeutics, based in South San Francisco, formerly Vevo Therapeutics, was founded in 2022 by Nima Alidoust, Hani Goodarzi, Johnny Yu, and Kevin Shokat to develop large-scale single-cell datasets and virtual cell models for AI-driven drug discovery. Tahoe raised a $12 million seed round in its founding year (2022) and rebranded from Vevo to Tahoe in April 2025 following a legal challenge over its original name, choosing to invest resources in data generation and model development rather than litigation.
Earlier this year, Tahoe released Tahoe-100M, a dataset containing more than 100 million single-cell transcriptomic profiles that chart how 1,100 small-molecule perturbations affect 50 cancer cell lines. Produced using Mosaic’s pooled "cell village" method—where diverse cell types from multiple patient sources are tested together—Tahoe-100M captured responses across 60,000 experimental conditions at single-cell resolution. This approach enables profiling at a scale reportedly 50 times larger than all previously available drug-perturbed single-cell datasets combined, while reducing batch effects common in conventional studies.
Tahoe-100M was incorporated into the Arc Virtual Cell Atlas alongside 200 million other single-cell profiles from Arc Institute’s scBaseCamp, and was used to train an open-source virtual cell model that Arc reported had twice the predictive accuracy of comparable AI systems. While Tahoe-100M was open-sourced, the company plans to keep future datasets proprietary.
Now, Tahoe plans to expand to a 1 billion-datapoint single-cell perturbation dataset. The company intends to use this as the backbone for its proprietary virtual cell models and a single strategic collaboration—granting one pharmaceutical or AI partner controlled access to co-develop medicines that leverage the dataset—aiming to extend biological foundation models, lower clinical-trial failure rates, and accelerate precision-medicine development, while advancing its own therapeutic programs toward the clinic.
Topics: Startups & Deals
