Google DeepMind has selected 15 early-stage robotics startups from across Europe for a new three-month accelerator program, giving founders direct access to its Gemini robotics models, its broader AI stack, and hands-on mentorship from Google DeepMind and Google engineers. The cohort kicks off this week in London and spans 10 countries, including the United Kingdom, France, Germany, Switzerland, Italy, Norway, Sweden, Denmark, Greece, and Romania. The program, called the Google DeepMind Accelerator: Robotics, is the company's most concrete push yet to seed a European physical-AI ecosystem.
Carolina Parada, VP of Robotics at Google DeepMind, is leading the effort. The bet is that the next leg of AI progress runs through robots that can perceive, plan, and act in the real world, and that Europe — with its concentration of robotics research labs and industrial customers — is a natural place to compound that work. Selected startups will get technical mentorship, product guidance, and introductions to a partner network across the three months.
The 15 companies were chosen from what Google DeepMind described as a strong applicant pool. They cluster into a handful of recognizable themes: industrial automation, healthcare, construction, ocean and subsea autonomy, humanoid platforms, and the developer tooling layer that makes robotic foundation models deployable.
“Robotics is one of the most exciting frontiers of AI, where advances in language, vision and action models can help create intelligent machines that interact with the real world in safer, more helpful and more adaptive ways.”— Carolina Parada, VP Robotics, Google DeepMind
Key facts
- 01Google DeepMind selected 15 early-stage robotics startups from across Europe for a three-month accelerator program kicking off this week in London.
- 02Cohort spans 10 countries including the UK, France, Germany, Switzerland, Italy, Norway, Sweden, Denmark, Greece, and Romania.
- 03Participants gain access to Google DeepMind's AI stack, technical experts, and Gemini robotics models.
- 04Norway's 3D-Components AS claims its AI welding platform delivers results 280× faster than traditional trial-and-error methods.
- 05Verticals represented include healthcare microrobots, ocean autonomy, humanoid robots, construction MicroFactories, and waste sorting.
On the industrial side, Norway's 3D-Components AS automates welding selection and inspection and claims results 280× faster than traditional trial-and-error methods. Greece's Acumino is building hardware-agnostic physical AI for complex industrial tasks. Romania's Adapta Robotics offers a software ecosystem for scalable device testing across automotive and healthcare. Germany's Deltia GmbH digitizes production-line work into process graphs that teams can optimize and partially automate.
Construction and recycling are represented by AUAR (Automated Architecture) in the UK, which deploys robotic MicroFactories directly to construction sites to bring down home-building costs, and Danu Robotics, also UK-based, which uses embodied AI to automate waste sorting and recover materials for the circular economy.
Healthcare and human-augmentation entries include France's ROBEAUTE, which is building microrobots that navigate through brain tissue to diagnose, treat, and monitor neuropathology, and Italy's Generative Bionics, which is developing humanoid robots based on physical AI. Switzerland's Embodied AI deploys teleoperated humanoids that collect data during customer service interactions to keep training their manipulation skills.
Ocean autonomy gets its own entry in France's Bubble Robotics, which is building a vessel-free constellation of self-docking surface and subsea robots that feed a live underwater world model. Sweden's Staer uses computer vision on existing cameras and sensors to build 3D spatial models of facilities, giving robots a shared environment to navigate and operators real-time visibility.
The cohort also includes the infrastructure layer that makes any of this work. UK-based Extend Robotics provides teleoperation software and data pipelines to fine-tune foundation models for real-world robotics. Switzerland's Forgis is building AI agents that understand machines like experienced engineers, predicting failures and optimizing operations. Denmark's Qualia is building infrastructure to turn robotic foundation models into working deployments. UK-based Touchlab uses advanced nano inks to create an e-skin that gives robots high-resolution touch across flexible surfaces.
Three months is a short runway, and none of the headline numbers — funding amounts, equity terms, commercial commitments — were disclosed. Most of the cohort is pre-revenue or early-revenue, and access to Gemini robotics models matters most if those models meaningfully shorten the path from demo to deployment. That is the bet Google DeepMind is making and the one the cohort is making back. Whether the partnerships outlast the program will be the real test.
The strategic logic is straightforward. Foundation-model providers competing in physical AI need a flywheel of real-world deployment data, varied hardware, and varied task domains — exactly what 15 startups working across welding, brain surgery, waste sorting, and subsea inspection generate. For Google DeepMind, the accelerator is a low-cost way to put Gemini robotics models in front of operators solving concrete problems, and to anchor a European robotics ecosystem before US and Chinese competitors crowd in. For the founders, the value of an AI stack and a partner network is real if it converts into customers; otherwise it's an expensive press release. The next 12 months will show which it is.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
