An Open Biocenter is a data center for biology.
GaudiLabs (Urs Gaudenz, Miranda Moss) and Radical Design Studio (JB Labrune) have partnered inside the Open Science Foundation at Etherlaken to build a physical laboratory operable end to end through human-AI collaboration, from a browser or a phone. The first prototype, the OpenDrop BioServer, was installed on Tuesday, September 1, 2026, in the premises that prefigure the foundation site in Interlaken, Switzerland.
Premise
Biology still runs by hand
Most experiments happen in a local lab, performed by a person standing at a bench. Compute met the same limit and answered it with data centers: shared machines, remote access, capacity added one unit at a time. Biology has no equivalent.
Canty et al., Science acceleration and accessibility with self-driving labs, Nature Communications, 2025.
Vision
Open Biocenters
An Open Biocenter is a building that holds machines instead of benches. Each machine automates one class of experiment and occupies one slot in a rack. Capacity grows by adding slots.
Remote experimentation
How remote experimentation works
A team anywhere on the planet opens a web page or a phone. Commands travel to a machine in Interlaken, which moves a real sample. Onboard sensors and a camera report what happened.
On the way back, deep machine vision reads the images and world models turn them into a state the team can act on.
Design study, placeholder values
First prototype
OpenDrop BioServer
The unit is a digital microfluidics device in a glass enclosure. Two OpenDrop boards sit on the upper shelf. Each carries an electrode array that moves droplets across a surface by switching voltage cell by cell, with magnets and heaters acting on the droplet as it travels. Samples and reagents sit on the shelf below.
A console reports the state of the machine. The unit answers to a web page over the network.
Principles
What follows from publishing the machine
- Open hardware
- The design files for every machine in the rack are public.
- Open source
- The control software is published and runs on hardware you own.
- Auditable by reading
- A claim about what a machine did can be checked against its design.
- Forkable
- Anyone can build a second unit and change it.
- No vendor lock-in
- No part of the stack is held by a single supplier.
- Reproducible elsewhere
- A result obtained here can be repeated on a machine someone else owns.
- Remote by default
- Every operation available at the bench is available over the network.
- Observation first
- The state of the sample returns before the next command is sent.
Partners
Who is building it
- GaudiLabs
- Urs Gaudenz, Miranda Moss
- Radical Design Studio
- JB Labrune
- Open Science Foundation
- Host of the partnership, at Etherlaken
- Etherlaken
- The site, Interlaken, Switzerland