Agile Robots and Google DeepMind Partner to Bring Gemini to the Factory Floor

A Marriage of Industrial Scale and Frontier AI

MUNICH — March 24, 2026 — In a move that signals a major consolidation of "Physical AI" efforts in Europe and North America, Agile Robots SE and Google DeepMind have announced a strategic research partnership. The collaboration aims to integrate DeepMind’s Gemini Robotics foundation models with Agile Robots’ industrial hardware to develop adaptable, reasoning-capable machines for manufacturing and logistics.

The partnership seeks to create a "scalable AI flywheel," where real-world deployment data from industrial operations is used to train and iterate on robotic models. This expanded intelligence, in turn, allows for broader and more complex deployments in sectors with acute demand for reliable automation.

Closing the Data Gap

For Google DeepMind, the partnership provides a critical pipeline of high-quality industrial data. This addresses the "data bottleneck" previously identified by Kanishka Rao, DeepMind’s Director of Robotics, who noted that unlike large language models, robotics lacks internet-scale physical interaction data.

By partnering with Agile Robots—which has already installed over 20,000 robotics solutions worldwide—DeepMind gains access to a massive, diverse dataset of real-world interactions. This data is essential for training the "Infinite Training Loop," a strategy championed by CEO Demis Hassabis to move AI from passive information summarizers to agentic systems that understand cause-and-effect in the physical universe.

Leveraging the Agile ONE Platform

While the partnership covers Agile Robots’ entire scalable industrial platform, the recently unveiled Agile ONE humanoid is a primary candidate for these advanced "brains".

• Hardware Dexterity: Standing 174 cm tall, the Agile ONE features what the company claims are the "world's most dexterous robotic hands," equipped with torque sensors in every joint.
• Industrial Utility: Unlike general-purpose prototypes, Agile ONE was designed specifically for "co-worker" tasks on the factory floor, matching human walking speeds of 2.0 m/s.

The integration of Gemini Robotics models into this hardware follows the "cross-embodiment" strategy detailed by DeepMind’s Head of Robotics, Carolina Parada. Previous tests showed that skills trained on one robot could be performed by different platforms, such as the Apptronik Apollo humanoid, without specific fine-tuning.

The 'Android of Robotics' Strategy

The deal reinforces DeepMind’s ambition to build a universal "operating system" for robots—an "Android for the physical world". To achieve this, the lab has aggressively recruited mechanical talent, including former Boston Dynamics CTO Aaron Saunders and Cartwheel Robotics founder Scott LaValley.

By deploying the Gemini 1.5 Robotics framework onto Agile’s hardware, the robots will utilize a "two-part brain":

1. Embodied Reasoning (ER): A strategic planner that can interpret vague goals and even use Google Search to learn context, such as local recycling rules.
2. Vision-Language-Action (VLA): A motor cortex that translates plans into physical commands while generating an "internal monologue" to ensure semantic understanding of the task.

Industrial Implications

Agile Robots' CEO and Founder, Zhaopeng Chen, emphasized that the "huge opportunity" lies in autonomous production systems that transform entire industries. This aligns with broader trends in the German robotics sector, where competitors like Neura Robotics are negotiating massive funding rounds and players like Schaeffler are committing to thousands of humanoids.

The partnership is slated to begin full production-level integration at Agile Robots’ new facility in Bavaria, where the company maintains a vertically integrated manufacturing strategy. As 2026 progresses, the industry will watch to see if this "AI flywheel" can finally bridge the gap between simulation and the messy reality of the factory floor.