Google DeepMind announced on June 9, 2026, the first cohort of its Accelerator: Robotics program - 15 early-stage companies drawn from Norway to Greece, building everything from brain-navigating microrobots to electronic skin for industrial arms. They gathered in London this week to begin a three-month program that gives them access to Google's AI stack, Gemini robotics models, and direct technical mentorship from DeepMind researchers.

The program is equity-free. That detail matters: startups receive mentorship, infrastructure access, and product guidance without surrendering any ownership stake. According to the program documentation published by Google DeepMind, selected companies may also be eligible for up to $350,000 in credits from the Google for Startups Cloud program, subject to applicable terms.

How the program works

The structure is a hybrid of in-person and online sessions. A five-day kickoff event in London this week opens the program. Digital training and one-to-one mentoring then run online from June through September 2026. The program closes with a three-day graduation event in London in September, culminating in a Demo Day where the cohort presents to a network of mentors, partners, and Google teams.

According to Google DeepMind, the cohort will "work hand-in-hand with Google experts on their biggest tech challenges, engaging with Google AI and Gemini Robotics Models to help move them from prompt to production." That phrase - from prompt to production - captures the accelerator's stated aim: bridging the distance between AI research and deployable hardware.

The technical core of the offer is access to the Gemini Robotics model family. Gemini Robotics is a vision-language-action model built on the Gemini 2.0 architecture, launched by Google DeepMind in March 2025 alongside a companion model called Gemini Robotics-ER, which focuses on spatial reasoning. A third variant, Gemini Robotics On-Device, released in June 2025, is designed to run locally on robotic hardware without requiring a cloud connection - significant for deployment in environments where latency or connectivity is a constraint. Access to these models has so far been restricted to a small set of trusted testers including Boston Dynamics, Agility Robotics, and Enchanted Tools.

The program's 12 to 15 intensive weeks are structured around what the program documentation describes as a "hybrid combination of digital training, 1:1 mentorship, and hands-on workshops." Each startup's support is tailored to its individual technical and business challenges, with exclusive invitations to tech bootcamps hosted by Google.

Who applied - and who got in

Applications opened on February 24, 2026, and closed on March 25, 2026. The program sought companies at pre-seed stage or above, headquartered in Europe, with verified venture capital backing and technical teams of at least five members - including founders who must be available to join all program elements. The target verticals were logistics, manufacturing, health and life sciences, human-robot interaction, education and learning, and advanced navigation. The program description notably invited companies outside those verticals to apply if they were "helping to drive the next generation of physical AI."

The 15 selected companies span nine European countries. Here is each cohort member, with the focus area Google DeepMind attributed to them:

3D-Components AS (Norway) is developing RobTrack, a platform that automates parameter selection and quality control for robotic welding and metal 3D-printing. According to Google DeepMind, the platform performs those tasks 280 times faster than current practices.

Acumino (Greece) develops hardware-agnostic physical AI that enables robots to perform complex industrial tasks in a scalable, cost-efficient manner.

Adapta Robotics (Romania) deploys physical AI that replicates human touch to test devices and software for healthcare, automotive, and consumer electronics - what the program documentation describes as enabling automated quality assurance and supporting the circular economy.

AUAR (Automated Architecture) (United Kingdom) makes homebuilding more affordable by deploying robotic MicroFactories directly to construction sites.

Bubble Robotics (France) is building what it describes as an autonomous ocean workforce - a vessel-free constellation of self-docking surface and subsea robots. The company raised a $5 million pre-seed round in April 2026, led by Episode 1 Ventures and Asterion Ventures. Founded in 2025 by Jean Crosetti and Patricia Apostol, both former robotics engineers from NASA and ETH Zurich, the company's core product is a docking station called BubbleDock that can remain at sea for up to six months, deploying subsea robots to inspect offshore wind farms, pipelines, and ports. The company reports letters of intent worth over $4 million and has offices in France, Zurich, and San Francisco.

Danu Robotics (United Kingdom) uses embodied AI robotic systems to automate complex waste sorting, focusing on recovering valuable materials at scale.

Deltia GmbH (Germany) digitizes production-line work, transforming workflows into process graphs that help teams identify where manual processes can be optimised or automated.

Embodied AI (Switzerland) deploys teleoperated humanoids that collect data during customer service operations to continuously train and improve their manipulation skills.

Extend Robotics (United Kingdom) provides teleoperation software and data pipelines for training and fine-tuning foundation models for real-world robotics.

Forgis (Switzerland) develops AI agents that understand machines "like experienced engineers," according to the program documentation, predicting failures and optimizing operations.

Generative Bionics (Italy) amplifies human potential by developing humanoid robots based on physical AI, described as developed in Europe but designed to scale globally.

Qualia (Denmark) is building infrastructure that allows companies to convert robotic foundation models into working deployments, automating time-consuming manual labor.

ROBEAUTE (France) is building microrobots that navigate through brain tissue to diagnose, treat, and monitor neuropathology. Paris-based and founded in 2017, the company raised $28 million in January 2025 from a round led by Plural, Cherry Ventures, and Kindred Ventures, with strategic investment from Brainlab. The microrobots are roughly the size of a grain of rice - approximately 3mm in length - and can move along curved trajectories through brain tissue, unlike conventional surgical instruments that follow rigid, linear paths. The company has filed over 50 patents in eight years. Human trials were planned to begin in 2026 at the time of the funding announcement, and the company was also establishing US operations ahead of seeking FDA approval.

Staer (Sweden) uses computer vision on existing cameras and sensors to build 3D spatial models of facilities, giving robots a shared environment to navigate while giving operators real-time visibility into their physical operations.

Touchlab (United Kingdom) uses advanced nano inks to create an e-skin - an electronic skin layer - that gives robots high-resolution tactile sensing across flexible surfaces. Based at the National Robotarium in Edinburgh, the company was founded in 2018 by Dr. Zaki Hussein and raised $4.8 million in seed funding from Octopus Ventures in 2022. The e-skin is thinner than human skin and can withstand extreme environments including acid exposure, high and low temperatures, and radioactive conditions.

Physical AI and the wider context

The program sits inside a broader push by Google DeepMind into embodied AI - systems where the model does not merely generate text or images but perceives the physical world and acts within it. Google I/O 2026 in May included a dedicated robotics session featuring Kanishka Rao and Boston Dynamics' Alberto Rodriguez, placing robotics alongside quantum computing and AI agents as a central theme for the year. The same event featured Gemini 3.5 as the new default model across Google's product stack - an upgrade that also underpins the technical foundation available to the accelerator cohort.

Carolina Parada, VP of Robotics at Google DeepMind, announced the cohort in a post on the Google Keyword blog dated June 9, 2026. Parada wrote that "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."

The Gemini Robotics model family - the technical backbone of the accelerator - was itself introduced by Parada's team in March 2025. That launch demonstrated robots folding origami, preparing packed lunches, tying shoelaces, and making a basketball dunk, running on the ALOHA 2 dual-arm platform but designed to generalise across robot hardware types. The June 2025 on-device variant extended that capability to local inference, removing the dependency on remote servers.

Google DeepMind CEO Demis Hassabis has publicly articulated an ambition to build what he described as an AI operating system for robots - a general-purpose brain that could run on any physical platform in a manner analogous to how Android runs on billions of smartphones. In December 2025, the company announced its first automated research laboratory in the United Kingdom, a facility that will combine Gemini with robotics to synthesise and test hundreds of materials each day, targeting superconductor and semiconductor applications.

The accelerator represents the startup-facing side of that same infrastructure investment. Rather than building hardware directly, Google DeepMind is positioning itself as the software and model layer for a generation of European companies that will deploy physical AI in specific industrial, medical, and environmental contexts.

The sectors covered and what they reveal

The 15 cohort companies collectively map a set of industries where physical AI is expected to create distinct value. The range is wide. Robeaute operates at the boundary of neurosurgery and robotics, targeting the roughly 350,000 people diagnosed with primary brain cancer each year globally - a population whose treatment options are constrained by the limits of current surgical instruments. Bubble Robotics addresses the offshore inspection market, where vessel-based operations can cost up to $100,000 per day and the energy sector faces a structural shortage of 600,000 professionals by 2030.

Construction appears through AUAR, which targets housing affordability by deploying on-site robotic factories. Waste management appears through Danu Robotics. Welding and industrial fabrication appear through 3D-Components AS, which claims its RobTrack system delivers a 280-times speed improvement over current parameter-selection practices. Manufacturing process optimisation appears through Deltia. Ocean monitoring through Bubble. Humanoid development through Generative Bionics and Embodied AI. Quality assurance through Adapta. Predictive maintenance through Forgis.

Touch sensing is perhaps the most enabling category across the cohort. Touchlab's e-skin technology, if it matures into reliable deployment, solves a problem shared by nearly every other company in the group: robots that cannot detect pressure, direction, or slip cannot safely handle the varied materials they encounter in real-world operation. The company's nano ink approach produces a sensing layer that is retrofittable to existing robot arms and grippers without reducing their range of motion.

The equity-free structure and what startups receive

The equity-free model is worth examining in detail because it distinguishes this program from traditional accelerators. According to the program documentation, the support package includes tailored training on technical and business challenges, dedicated mentoring from Google teams and AI experts, technical project partnerships with Google engineers on specific challenges, exclusive invitations to tech bootcamps, access to Google's AI knowledge and technical infrastructure, and potential Cloud credits of up to $350,000 via the Google for Startups program.

After graduation, companies join the Google Accelerator Alumni Network and receive ongoing support. The program targets a cohort size of 10 to 15 companies - this first cohort landed at 15.

The structure also reflects the stage of the companies involved. Pre-seed startups with verified VC backing are unlikely to give up meaningful equity in exchange for three months of technical support. The equity-free model lowers the barrier to entry and positions Google DeepMind as an infrastructure partner rather than an investor, avoiding the conflicts of interest that can arise when a strategic partner holds an ownership stake.

Why this matters for the marketing and ad tech community

The connection to advertising and media may not be immediately visible in a robotics accelerator announcement, but there are structural links. PPC Land has tracked Google's broader AI infrastructure expansion through I/O 2026 and Google Marketing Live, noting that the Gemini model family now underpins not only search but also campaign tools, creative generation, and agentic workflows across Google's advertising products.

The accelerator extends that same Gemini infrastructure into physical AI. Companies like Deltia, which digitizes production-line workflows, or Staer, which builds spatial facility models, are creating the kind of structured operational data that feeds into AI systems at every layer - including the advertising and commerce infrastructure that Google is simultaneously rebuilding around AI Mode and Performance Max.

More broadly, the physical AI sector is attracting the same investment dynamics that have reshaped digital advertising over the past decade. The aggregation of data, the emergence of platform dependencies, and the concentration of infrastructure in a small number of companies - all of which PPC Land has documented extensively in the programmatic and search advertising contexts - are beginning to appear in robotics. A startup that builds its product on Gemini Robotics models, receives credits through Google for Startups, and graduates into the Google Alumni Network is connected to Google's infrastructure in ways that will matter commercially long after the three-month program ends.

Google's commitment to the EU's General Purpose AI Code of Practice, announced in July 2025, is also relevant context for European startups entering this program. The AI Act's full enforcement powers apply to new models from August 2026 and to existing models from August 2027, meaning the accelerator cohort will be building with Gemini Robotics models during the period when regulatory expectations for general-purpose AI in Europe are solidifying.

Timeline

  • 2010: Google DeepMind (then DeepMind) founded in London by Demis Hassabis.
  • 2014: Google acquires DeepMind.
  • 2017: Robeaute founded in Paris by Bertrand Duplat and Joana Cartocci, beginning eight years of microrobot development for neurosurgery.
  • 2018: Touchlab founded in Edinburgh by Dr. Zaki Hussein, developing e-skin technology for robots.
  • 2022: Touchlab raises $4.8 million in seed funding led by Octopus Ventures for its e-skin robot technology.
  • March 12, 2025: Google DeepMind launches Gemini Robotics and Gemini Robotics-ER, vision-language-action models built on Gemini 2.0, as covered in the context of Google DeepMind's expanding AI infrastructure.
  • January 14, 2025: Robeaute raises $28 million, led by Plural, Cherry Ventures, and Kindred Ventures, for neurosurgical microrobots the size of a grain of rice.
  • June 24, 2025: Google DeepMind releases Gemini Robotics On-Device, enabling local inference on robotic hardware without cloud connectivity.
  • July 30, 2025: Google commits to signing the EU General Purpose AI Code of Practice, ahead of August 2025 enforcement deadlines from the EU AI Office.
  • December 11, 2025: Google DeepMind announces plans for its first automated research laboratory in the United Kingdom, combining Gemini with robotics to test materials.
  • February 24, 2026: Applications open for the Google DeepMind Accelerator: Robotics.
  • March 25, 2026: Applications close.
  • April 21-22, 2026: Bubble Robotics emerges from stealth with a $5 million pre-seed round led by Episode 1 Ventures, reporting over $4 million in signed letters of intent.
  • May 19, 2026: Google I/O 2026 includes a dedicated robotics session alongside the Gemini 3.5 announcement.
  • June 9, 2026: Google DeepMind announces the 15-company first cohort of the Accelerator: Robotics, with founders gathering in London to begin the program.

Summary

Who: Google DeepMind, led by VP of Robotics Carolina Parada, selected 15 early-stage robotics startups from nine European countries as the first cohort of its Accelerator: Robotics program.

What: A three-month, equity-free accelerator providing hands-on mentorship, access to Google's AI stack, Gemini Robotics model access, potential Cloud credits of up to $350,000, and a technical partnership with Google DeepMind engineers. The program culminates in a Demo Day graduation in London in September 2026.

When: Announced June 9, 2026. The cohort gathered in London during the week of June 9 to begin a five-day in-person kickoff. Online training and mentoring continues through September 2026, with a three-day in-person graduation event closing the program.

Where: The in-person elements are hosted in London. The online training and mentoring phases run digitally, covering the cohort members spread across Norway, Greece, Romania, the United Kingdom, France, Germany, Switzerland, Italy, Denmark, and Sweden.

Why: Google DeepMind is extending its Gemini model infrastructure into the physical AI and robotics sector. The accelerator positions Google as the platform layer for European companies building embodied AI systems - at a moment when the Gemini Robotics model family has reached the stage where startups can integrate it into production products, and when European regulatory pressure under the EU AI Act is increasing the importance of trusted infrastructure relationships for AI-first startups.