Google DeepMind Launches European Robotics Accelerator: 15 Startups Selected to Build Physical AI on Gemini Robotics

Google DeepMind announced on June 12 the official launch of its Robotics Accelerator program, selecting 15 early-stage robotics startups from Europe for the first cohort. The three-month program opens up DeepMind's complete AI technology stack and the Gemini Robotics model, with DeepMind teams providing hands-on mentorship. It represents Google's most systematic ecosystem play in embodied intelligence to date.

Background: The Embodied AI Landscape

Over the past two years, the embodied AI race has taken shape as "US-China duopoly plus a fragmented Europe":

• US: Star companies like Figure AI (valued at $40 billion), Tesla Optimus, Physical Intelligence, and Skild AI have seen valuations soar
• China: Unitree, Zhiyuan (AgiBot), Galbot, and others are accelerating mass production, with the national humanoid robot innovation center leading the supply chain
• Europe: Strong industrial robotics foundations (ABB, KUKA) and top research institutions, but lacking embodied-AI-native unicorns

By launching the accelerator now, DeepMind is effectively pushing Gemini Robotics from the lab into industry — selecting 15 seed-stage teams for a three-month deep support program to cultivate a European embodied AI ecosystem, and to avoid the next Physical Intelligence emerging in Boston rather than London.

Core Program Terms

• Program: Google DeepMind Accelerator: Robotics
• Duration: 3 months
• First cohort: 15 startups (kicked off in London)
• Company origins: 10 European countries including the UK, France, Germany, Switzerland, Italy, Norway, Greece, Romania, Denmark, and Sweden
• Criteria: AI must be core to the product or technical approach; early-stage companies
• Resources offered: DeepMind's full AI stack, Gemini Robotics model, technical mentorship, product guidance, and partner network
• Key boundary: No disclosed investment, acquisition, equity terms, compute credits, or commercial return commitments

This is a critical point — DeepMind explicitly stated that no investment, acquisition, or equity terms were disclosed, making this a pure technology and ecosystem investment rather than a financial incubator. This "heavy tech binding, light capital binding" approach reflects DeepMind's strategic intent to claim the power to define embodied AI standards.

Segments Covered by the 15 Startups

The portfolio reveals the full scope of DeepMind's bets:

• Manufacturing: factory automation, assembly line robots
• Logistics: warehouse handling, last-mile delivery
• Healthcare: surgical assistance, rehabilitation robotics
• Construction: on-site building, 3D-printed construction
• Marine operations: underwater inspection, offshore wind maintenance
• Waste recycling: sorting, circular reuse
• Tactile skin: flexible sensors, e-skin
• Teleoperation: remote control, low-latency human-robot collaboration
• Humanoid robots: general-purpose humanoids, bipedal platforms

The logic is a wide net with a heavy bet on humanoids — covering every segment that could produce a unicorn while treating humanoid robotics as the ultimate form factor.

Strategic Significance of Gemini Robotics

DeepMind's confidence in running this large-scale ecosystem incubation rests on the maturity of the Gemini Robotics foundation model:

• Extends multimodal large model capabilities into the physical world (vision + language + action)
• Lets startups skip training foundation models from scratch and fine-tune Gemini Robotics for specific tasks
• Dramatically lowers the entry cost for embodied AI startups

This "foundation model + vertical applications" division of labor mirrors how GPT spawned waves of AI application companies. Expect a surge of Gemini Robotics-based vertical application startups within the next 12 months.

The China Comparison

Placing this program in the US-China competitive frame:

• Lead players: US — top AI labs (DeepMind); China — state-owned enterprises, national innovation centers, internet giants
• Funding: US — parent company technology investment (no equity ties); China — state capital, industrial funds, corporate strategic investment
• Tech stacks: US — Gemini Robotics + own stack; China — Huawei CloudRobo and various in-house models
• Deployment speed: US — lab incubation first; China — mass production and delivery already underway (e.g., Unitree H1)
• Strengths: US — top-tier fundamental research plus ecosystem cohesion; China — supply chain, policy support, manufacturing scale

The key difference: DeepMind's path is "standards first, then ecosystem" — opening its tech stack to grow an ecosystem that feeds back into model iteration. China's path is "hardware first, then models" — leveraging manufacturing advantages to iterate robot hardware quickly, then moving up into models. Which wins is undecided, but the next 18 months are the critical observation window: whether this cohort produces unicorns will determine if the foundation-model camp can come from behind.

My Take: Three Signals for Practitioners

1. Embodied AI has entered the "model + ecosystem" stage — companies doing only hardware or only applications will increasingly depend on foundation model providers. Whoever controls the base model controls the ecosystem.

2. Europe's embodied AI window — among these 15 companies, 1–2 are likely to reach $1B+ valuations within 18 months, worth close tracking.

3. Gemini Robotics open-source strategy remains to be seen — DeepMind has not committed to open-sourcing the model, but to truly win the ecosystem, a Llama-style open-source play will eventually be put on the table.

Reference: https://x.com/GoogleDeepMind/status/2065388989146628563