Google DeepMind launches robotics accelerator for 15 European startups, offering Gemini Robotics model access

Google DeepMind kicked off its first dedicated robotics accelerator on June 9, 2026, gathering 15 selected European startups in London to begin a three-month program built around access to the company's Gemini Robotics models and AI engineering stack.

What's new

The program, formally named the Google DeepMind Accelerator: Robotics, selects early-stage companies and gives them direct access to Google's AI infrastructure: "They'll have access to our AI stack, technical expertise and Gemini robotics models," the company wrote in the announcement. That access is paired with hands-on mentorship from Google DeepMind researchers, product guidance, and connections to a partner network.

The cohort of 15 spans a range of physical AI applications. According to the announcement, "The cohort joining us in London this week reflects the breadth of opportunity in embodied AI — from logistics and manufacturing to healthcare, climate, and advanced navigation." Specific startup categories include welding automation, waste sorting, humanoid robots, brain surgery microrobots, and precision navigation systems.

The five-day in-person launch in London is followed by online sessions running through September 2026. Participants are also eligible to apply for up to $350,000 in Google Cloud credits through the Google for Startups Cloud program, providing significant compute runway for training and inference at scale.

Context

This is Google DeepMind's first cohort focused exclusively on robotics as a domain. The company had previously run broader AI accelerators for startups in Europe and Asia Pacific, but the Gemini Robotics branding — and the direct technical access to those models — makes this program structurally different from a general startup mentorship initiative.

Gemini Robotics itself is a family of models designed for embodied AI tasks: spatial reasoning, manipulation, and physical-world planning. Giving early-stage startups direct access to these models at the prototyping stage is an unusual commitment. Most robotics AI programs provide cloud credits and business mentorship, not hands-on model access from the lab building the frontier systems.

The timing also follows a period of rapid progress in physical AI. Google DeepMind's Gemini Robotics ER-1.6 was released in April 2026, followed by the Project Genie simulator expansion in May. The accelerator appears designed to convert that internal R&D progress into a commercial ecosystem in Europe before competing labs establish comparable programs.

Why it matters

The selection of 15 companies rather than a larger cohort suggests an emphasis on depth of engagement over breadth. For the startups involved, direct access to Gemini Robotics models while they are still in active development creates an alignment between the lab's research roadmap and the companies' product timelines — a structural advantage that money alone cannot replicate.

For Google DeepMind, running the program in Europe is a deliberate move. The EU has significant robotics hardware manufacturing capacity — Germany, Italy, and Scandinavia all have established industrial automation sectors — and European healthcare and infrastructure regulation shapes the kinds of physical AI that can be deployed at scale. Building a startup ecosystem there from an early stage helps ensure that Gemini Robotics becomes the default AI foundation for European physical AI companies as the field matures.