Swiss AI start-up Atlas Bio has raised $12.5 million in funding led by Swedish venture capital firm Creandum, an early backer of Spotify and Trade Republic, to run drug trials on virtual patients instead of recruiting people for the most expensive stage of testing. The company said on Tuesday, October 27, 2026, that it aims to replace Phase II trials, which it says cost more than $100 million, take two to three years, and see about 70 percent of drugs fail at that stage, with 12-week computer simulations, and that results from a first trial are expected within six weeks.
Why Phase II Is The Bottleneck
Drug development fails most expensively in the middle. Early laboratory work can show that a molecule does something interesting, but only a trial in patients can show whether it helps, harms, or does nothing at all. Phase II is where that question is first asked at scale, and it is where most candidates fall. Atlas Bio argues that the cost, time, and failure rate of this stage are not just a commercial problem: they decide which diseases get new treatments, because only the best-funded programmes can afford to fail slowly.
The figures the company cites frame its pitch in stark terms. A single Phase II trial, on its account, can cost more than $100 million and run for two to three years, with roughly seven in ten drugs failing at that stage. Those are company figures, reported by Reuters, not independently audited trial statistics, and they will vary widely by disease, trial design, and country. Even allowing for that variation, the direction is not in dispute: human trials are slow, costly, and uncertain, which is why any credible shortcut attracts attention from investors and drugmakers alike.
How A Virtual Trial Would Work
Atlas Bio says its answer is an AI foundation model trained on about two million human RNA readouts, the molecular signals that show which genes are active in a cell. Instead of giving a drug to a group of patients and waiting to see what happens, the model simulates how virtual patients would respond, and it can do so across several diseases at once, something the company says is not feasible in a conventional trial. Animal testing, on its account, would also be replaced by the same simulations.
The claim is ambitious, and it should be read as a claim, not a result. A model trained on RNA data can find patterns that humans miss, but a simulation is only as good as the biology it has learned, and regulators, doctors, and drugmakers will want to see its predictions tested against real outcomes before they trust it with decisions about real patients. Atlas Bio says results from a first trial are expected within six weeks, which, if it means a first real-world comparison, would be the first concrete test of whether the approach works outside its own laboratory.
Who Is Behind The Company
Atlas Bio was founded by former Google employees alongside AI and pharmaceutical specialists, a combination that reflects the two disciplines its pitch depends on: large-scale AI modelling and drug development. Creandum, which led the round, is best known as an early backer of Spotify and Trade Republic, and its involvement signals that the bet is being judged as a technology investment as much as a life-sciences one. The size of the round, $12.5 million, is modest by drug-development standards, where a single trial can cost many times that amount, and it suggests the company is funded to prove its simulation approach rather than to carry a drug through to market itself.
Why It Matters
If virtual trials worked, the effect would reach beyond one company. Cheaper, faster Phase II testing could let smaller drugmakers test ideas that today die for lack of funding, and it could let larger ones test more candidates, and more diseases, in parallel. It could also change who bears the risk of a failed trial, shifting some of it from patients who volunteer for long studies to computer models that can be run and re-run. The same AI-in-biology logic is driving other work in the field, from Claude’s move into drug discovery work to Pathos AI’s $237 million raise to guide cancer drugs with AI, where the pitch is the same in outline: use AI to raise the odds that a drug succeeds before it reaches patients.
The caution is equally clear. No virtual patient has yet replaced a real one in a regulatory decision, and the history of drug development is full of models that looked convincing in the laboratory and failed in the clinic. Atlas Bio’s six-week timeline for first trial results will therefore matter more than its funding announcement: it is the point at which a bold claim meets a measurable outcome.
Background
Phase II trials test whether a drug works in patients with the disease it targets, after earlier safety testing, and before the larger Phase III trials that regulators rely on for approval. RNA readouts measure gene activity in cells, giving researchers a molecular picture of how a disease behaves and how a drug might change it. Creandum is a Swedish venture capital firm whose early investments include Spotify and Trade Republic.
Source: Reuters report
0 Comments