Pomegra
Basel-based Atlas Bio has come out of stealth with $12.5 million to run AI-simulated Phase II trials, backed by Creandum, First Momentum Ventures and LEA Partners.
Key Takeaways
- Atlas Bio disclosed $12.5M in total funding on Oct. 27, 2026: a $10M seed led by Creandum and a $2.5M pre-seed.
- Its model, trained on about 2 million human RNA readouts, simulates drug effects in virtual patient cohorts in roughly 12 weeks.
- The company works with one top-10 pharma firm. Valuation is undisclosed.
Lead
Atlas Bio, a Basel-based techbio startup, emerged from stealth on October 27, 2026 with $12.5 million in combined funding. The money covers a $10 million seed round led by Creandum and a previously undisclosed $2.5 million pre-seed round led by First Momentum Ventures and LEA Partners. The company builds a foundation model that predicts how a drug candidate affects individual patients in a simulated cohort, with the aim of replacing parts of Phase II clinical trials and early animal testing.
What Does Atlas Bio Actually Build?
Atlas Bio builds an AI foundation model for in silico clinical trials, trained on roughly 2 million human RNA readouts taken from human tissue. The software compares a drug against patient profiles, then identifies which patients are likely to respond and which diseases are worth pursuing. A simulation project takes about 12 weeks.
The pitch differs from much of the sector in one respect. The model is designed to test a compound across multiple diseases at once, without requiring prior efficacy data on that compound. That matters for drugmakers deciding where to spend a Phase II budget.
The company is incorporated in Delaware, runs a Swiss subsidiary and is headquartered in Basel, the center of Switzerland's pharmaceutical industry.
Why Target Phase II Trials?
Phase II is where most drug programs die. About 70% of candidates fail at that stage, and a single trial can cost more than $100 million and take two to three years, according to the company.
Atlas Bio's argument is that a model able to screen indications and patient subgroups beforehand cuts the number of expensive failures. The figures are the company's own, and no independent validation has been published alongside the announcement. A simulation that reads RNA data is also a proxy: it predicts a transcriptional response, not a clinical outcome such as survival or symptom relief.
Who Is Behind the Company?
CEO Simon Romanski studied electrical engineering at Karlsruhe Institute of Technology, EPFL Lausanne and Stanford, and worked on autonomous vehicles at Google starting in 2014. He has Type 1 diabetes, which he cites as his motivation for moving into drug development.
CTO Sebastian Bayer built perception models for autonomous driving in data-limited settings, then led Google research on language models for Search, YouTube and Google Play. Dominik Hartl, the chief scientific officer, was a doctoral fellow at the Friedrich Miescher Institute for Biomedical Research and has more than 10 years in drug development.
The mix is typical of the current wave of techbio founders: machine learning engineers from large technology companies paired with a scientist who knows how trials fail. Romanski framed the ambition bluntly, saying Atlas Bio is "the end of Phase II as we know it."
How Does the Round Compare With Competitors?
The round is modest next to its direct rivals. Unlearn, which builds digital twins for trials, has raised more than $130 million in total, including a $50 million Series C. QuantHealth raised a $45 million Series B and says it has simulated more than 600 trials.
Atlas Bio's $10 million seed is therefore small for a company targeting drug development, where validation is slow and expensive. It does have a commercial foothold. The company says it is working with an unnamed top-10 global pharmaceutical company and is in talks with others. A Basel biotech has also skipped animal studies for a program, using Atlas Bio's simulations before moving to human trials.
Both customers are unnamed, and the company has not said whether they pay at commercial rates or are running pilots. That gap is what the next round will test. Creandum partner Katharina Wilhelm has joined the board.
What Will the Money Fund?
The company plans to spend the funds on scientific and commercial hiring, platform validation and signing additional enterprise customers. Headcount was not disclosed. Valuation was also not disclosed.
Market-sizing estimates put the in silico clinical trials market at about $4.16 billion in 2026, rising to $5.93 billion by 2031. Those projections come from the sector's own research and should be read as directional.
What Comes Next for AI Trial Simulation?
The next milestone is evidence. Drugmakers will adopt simulation for go or no-go decisions only if predictions match what later happens in real patients, and that requires prospective comparisons that take years to produce. Regulators, including the FDA and EMA, have signaled openness to reducing animal testing, which helps the early-stage use case more than it helps replacing human Phase II data.
A realistic near-term scenario is that Atlas Bio becomes a screening layer. It would narrow which indications and patient groups go into trials rather than substituting for them. A harder scenario for the company is that large pharma builds comparable models in house, given its access to proprietary patient data.
Outlook
Atlas Bio enters a field where rivals have raised several times more capital, with a technically credible team and one major pharma relationship. The $12.5 million buys time to validate the model, convert pilots into contracts and show that RNA-based simulation predicts clinical results. Until published validation arrives, the claim of ending Phase II remains a hypothesis.
