Google DeepMind has found another way into the physical world — this time through a Munich-based robotics company with 20,000 machines already in the field.
Agile Robots SE and Google DeepMind announced a strategic research partnership Tuesday that will embed Gemini Robotics foundation models into Agile's industrial robot fleet. The companies will jointly train, deploy, and test robots across electronics manufacturing, automotive, data centers, and logistics — and use the operational data from those deployments to improve the underlying AI models, creating what both sides describe as a scalable flywheel.
"Agile Robots has already installed over 20,000 robotics solutions worldwide, proving intelligent automation at scale," said Zhaopeng Chen, Agile Robots' co-founder and CEO, in the joint press release. "The huge opportunity ahead lies in autonomous, intelligent production systems that can transform entire industries."
The partnership is the second confirmed Google DeepMind robotics collaboration in 2026, following a January announcement that the lab would embed Gemini models into Boston Dynamics' next-generation Atlas humanoid robot. Google DeepMind confirmed Carolina Parada, Senior Director and Head of Robotics, as the lead on the Agile Robots engagement.
Agile Robots was founded in Munich in 2018 by researchers from the German Aerospace Center and has raised more than $270 million from SoftBank Vision Fund, Xiaomi, and Midas Group, among others. The company employs more than 2,500 people globally and operates a portfolio that includes audio AI (audEERING), collaborative robotics (Franka Robotics), and logistics automation (idealworks) alongside its consumer and industrial hardware.
The data question is the structural story. Gemini Robotics models are designed to allow robots to perceive, reason, use tools, and interact with humans — but foundation models are only as good as the data they train on. A lab that can pull operational telemetry from 20,000 deployed machines across multiple industries has a meaningful advantage over competitors relying on simulation data or small pilot datasets. That is the trade Google is making: hardware reach for data depth.
For Agile Robots, the calculus is different. The company has the hardware and the industrial customer relationships; Google DeepMind has the models. Neither alone can easily build what they are building together. As the partnership description puts it: data from real operations improves the models, and improved models expand robotic capabilities — unlocking broader deployment.
The collaboration will focus first on high-value industrial tasks where reliability and scale are non-negotiable — the kinds of environments where a failed automation deployment is expensive in both money and safety terms. Neither company disclosed financial terms or timeline beyond calling the partnership long-term.
The broader context: physical AI has become the next frontier for the major AI labs. Nvidia CEO Jensen Huang has called it the next frontier for the AI market. Google DeepMind is methodically building a roster of hardware partners — Boston Dynamics for humanoids, Agile Robots for industrial automation — each of which provides both a deployment channel and a data source. The question is not whether the models work in the lab. The question is who owns the deployment footprint when the models are ready to ship.
