Atlas Bio raises $10M seed to replace Phase II trials with AI simulations
What's the deal? Swiss tech-bio startup Atlas BioDealroom has a profile for this one. Try Dealroom → has launched from stealth with a $10 million seed round led by Creandum. Alongside a previously unannounced pre-seed led by First Momentum VenturesDealroom has a profile for this one. Try Dealroom → and LEA PartnersDealroom has a profile for this one. Try Dealroom →, its total funding reaches $12.5 million.
What's the endgame? Atlas Bio is building an AI foundation model to run virtual, or in silico, clinical trials. The model simulates patient-level efficacy and safety before Phase II, aiming to cut years of testing to a 12-week simulation.
The model is trained on roughly 2 million RNA readouts from human tissue — the molecular signals showing how genes switch on and off. By learning these patterns, it predicts how a drug behaves across a virtual cohort of patients.
Why now? Phase II trials cost more than $100 million and take two to three years, yet around 70% of drugs fail at this stage. Regulators are also moving away from animal models: the FDA Modernization Act 2.0, passed in 2022, removed the legal requirement for animal testing before human trials.
That shift has continued. In March 2026, the Food and Drug Administration issued draft guidance on how developers can validate computer models, organoids, and organ-on-chip systems for drug development — though acceptance remains evidence-dependent and animal testing has not been broadly abolished.
Who's behind it? Atlas Bio was founded in 2024 by Simon Romanski and Sebastian Bayer, who previously led large language model research at Google. The company is already working with a top-10 global pharma firm and in talks with several others.
"For sixty years, drug development has been built on two flawed assumptions: that mice predict humans, and that the only way to learn whether a drug works is to give it to thousands of patients and wait," said Romanski, chief executive officer and co-founder. "Human biology is finally legible to AI."
What could go wrong? The model's value hinges on proving its predictions reliably match human outcomes. Regulatory acceptance remains fit-for-purpose, meaning each use must be validated — a headline failure-rate claim alone won't persuade sponsors.
The signal: Atlas Bio's $10 million seed sits in the 98th percentile for Swiss seed rounds in big data, a sign of investor appetite for AI tools that de-risk expensive drug development. As regulators open the door to non-animal methods, the race is on to prove virtual trials can inform real decisions.
Read more: Tech.eu
Image credit: Atlas Bio
