Google DeepMind Expands Robotics Push with Agile Robots Partnership

Google DeepMind now has another robotics hardware partner. Agile Robots — a Munich-based company with over 20,000 deployed robotic systems worldwide — announced a strategic research partnership with DeepMind on March 24, embedding Gemini Robotics foundation models directly into its product line.

The arrangement is two-directional. Agile Robots gains access to DeepMind’s Gemini Robotics models; DeepMind gets real-world operational data to sharpen those same models. Target sectors include electronics manufacturing, automotive, data centers, and logistics — all high-value industrial verticals where the economic case for autonomous robots is already clear.

Why this matters beyond the press release

DeepMind’s robotics program has been one of the worst-kept secrets in AI for years. The lab published a steady stream of research on dexterous manipulation and generalist policies, but translating that into fielded hardware required industrial partners willing to share data at scale. Agile Robots — backed by SoftBank Vision Fund, Xiaomi, and Midas Group, with more than $270 million raised since its 2018 founding — has that scale.

Zhaopeng Chen, co-founder and CEO, framed the deal in terms of market timing rather than technical curiosity: “The huge opportunity ahead lies in autonomous, intelligent production systems that can transform entire industries.” That’s not boilerplate. Companies integrating AI into physical production workflows are seeing cycle time and yield improvements that justify seven-figure robotics capex — and the limiting factor is no longer actuator cost, it’s the intelligence layer.

A pattern taking shape

Agile Robots is not the first company to sign this kind of agreement with DeepMind. The partnership follows a string of similar deals, with DeepMind positioning itself as the foundation-model layer for the broader robotics hardware ecosystem rather than trying to build its own proprietary robots from scratch.

That strategy mirrors what happened in cloud computing: the platform that becomes the default AI substrate for hardware makers wins disproportionately, regardless of which physical robot ends up shipping. Gemini Robotics becoming the operating model of choice in industrial deployments would be a meaningful revenue and data-moat story for Alphabet.

The specifics of the Agile Robots deal — duration, pricing, data-sharing terms — remain undisclosed. A spokesperson confirmed it’s a long-term arrangement, which in robotics typically means multi-year with milestone-based renewals tied to model performance benchmarks.

The competitive backdrop

DeepMind is not operating in a vacuum. OpenAI has been building physical AI capabilities since its acquisition of assets from 1X and its own robotics research. Figure AI, Boston Dynamics, and Apptronik are all running similar foundation-model integration plays. The question is whether DeepMind’s Gemini Robotics stack — trained on diverse robotic hardware data — outperforms task-specific fine-tuned models at the edge.

Industrial customers don’t need robots that can do everything; they need robots that can do their thing reliably, at speed, with minimal downtime. If Gemini Robotics delivers on that narrower brief inside Agile Robots’ already-deployed systems, the case for wider adoption becomes easy to make.