Extend Robotics at Google DeepMind Robotics Accelerator Demo Day
Updated: Sep 21
8 September 2026 | Google DeepMind Robotics Accelerator
What does it take to move robotics from an impressive demonstration to a solution that delivers genuine value for customers? For Extend Robotics, the answer is to start with the problem, not the robot.
On 8 September 2026, Extend Robotics joined fellow cohort companies, Google DeepMind, Google stakeholders and investors at the Google DeepMind Robotics Accelerator Demo Day. Extend Robotics was one of the top 12 companies to be chosen as part of the three-month programme, focused on advancing robotics innovation and deployment.
The day brought together cohorts for a full day of shared learning, demonstrations, company showcases and conversations about the progress made throughout the accelerator. For Extend Robotics, it was an opportunity to present the work developed during the programme, share the lessons learned and demonstrate how its Action-as-a-Service model is helping move robotics towards real-world deployment.
Teams from across the accelerator in an environment designed to encourage collaboration and the exchange of ideas. It combined informal conversations with a structured debrief, giving companies the opportunity to connect with their own Technical Success Managers (TSMs), other cohort teams, the DeepMind team and wider Google stakeholders.
Alongside the conversations and presentations, companies showcased their work and demonstrated the progress made throughout the programme.
For robotics companies working to solve complex real-world problems, these opportunities to exchange experiences are particularly valuable. Building and deploying robotic systems involves more than developing technical capability. It requires understanding the environments in which robots operate, the people using them and the outcomes customers need to achieve.
Extend Robotics takes to the stage
Extend Robotics presented on stage, sharing both the output of its accelerator work and the wider lessons taken from the programme.
A key focus was how Google Cloud now powers the deployment of Extend Robotics’ Action-as-a-Service platform, alongside the company's approach to solving real-world robotics challenges.
But the most important takeaway was not simply about technology. It was about the customer. Customers buy outcomes, not robots.
Robotics is developing at an extraordinary pace. Advances in humanoid systems, AI models, teleoperation and autonomous capabilities are opening up new possibilities across industries. Yet the technology itself is not what customers ultimately buy.
A customer does not necessarily want a robot, a model or a sophisticated architecture. They want a result. They want a quality-control task completed more efficiently. A repetitive process improved. A difficult operational challenge solved. A measurable improvement in productivity, consistency or cost.
The robot and the technology behind it are the tools that deliver that outcome.
For Extend Robotics, one of the most important lessons from the accelerator was the value of reversing the traditional approach to robotics deployment. Rather than starting with a technology and looking for a problem it can solve, start by understanding the customer's problem and the outcome they need. Then select the technology that can deliver it.
This customer-first approach can make the path from concept to deployment more focused and repeatable. It begins with understanding the customer's environment, the task they need to complete and the result they want to achieve.
From there, the appropriate combination of robotics, AI, teleoperation, data collection and automation can be selected to meet the requirements.
The technology remains critically important. However, it is the means to an end, rather than the end itself. The strongest technical capability only earns its place when it delivers meaningful value in the real world.
For robotics companies, this means staying anchored to the customer's use case throughout the deployment journey, and bringing the customer on the journey with them in understanding how the technology can deliver the result.
Presenting Action-as-a-Service to investors and industry leaders
Dr Chang Liu joined the other cohort companies in presenting Extend Robotics to a room of investors, DeepMind and Google executives, and wider stakeholders.
His session covered the Action-as-a-Service model, the problem it is designed to solve and the challenges involved in bringing robotic deployments into real-world environments.
The presentation provided an opportunity to explain how Extend Robotics is approaching the gap between robotics capability and practical deployment.
Action-as-a-Service is built around the idea that robotic systems should be able to carry out useful actions in the real world, supported by the software, AI and human-in-the-loop capabilities needed to make those actions possible. For customers, the value lies in the task being completed.
For the robotics industry, the challenge is making those tasks repeatable, scalable and increasingly autonomous, from the stage to a live production line.
The day concluded at the Extend Robotics booth, where the team continued conversations with attendees and demonstrated a live stream of a robotic deployment. The stream showed Gilbert, a humanoid system deployed on an active production line, carrying out quality control on apples. This was not a staged demonstration in a controlled environment, it showed a real deployment, on a live production line, carrying out the actual job.
For visitors, it provided a tangible example of what robotics looks like when it moves beyond the laboratory and into an operational environment with Google Cloud and Gemini powering the deployment.
Operational data is collected continuously through the deployment and used to improve classification and detection performance over time, creating a pathway towards progressively greater automation.
The system can carry out the task, collect information about what is happening and use that data to improve how the task is performed. That combination of robotics, AI, cloud infrastructure and continuous learning is an important part of the journey towards more capable autonomous systems. It also reinforces the importance of deploying technology in real environments.
The difference between a robot that can perform a task in a demonstration and a system that can deliver value on a production line is significant. Real-world deployment introduces changing conditions, operational requirements and the need for consistent performance. It is through these deployments that robotics systems can be tested, improved and developed to meet the needs of customers.
What the accelerator means for the future of robotics
The Google DeepMind Robotics Accelerator Demo Day provided a valuable opportunity for Extend Robotics to share its progress, connect with the wider robotics ecosystem and reflect on the lessons learned.
For the company, the experience reinforced an important principle: The future of robotics will be shaped not only by what robots can do, but by how effectively they can deliver the outcomes customers need. The industry is moving towards a world where robots, AI models, cloud platforms and human expertise work together to solve increasingly complex tasks. But the route to adoption starts with a clear understanding of the problem.
What does the customer need to achieve? What does success look like? What technology is required to deliver it? By keeping those questions at the centre of deployment, robotics companies can focus on building solutions that are useful, measurable and capable of delivering real-world value.
As Extend Robotics continues to develop its Action-as-a-Service approach, the focus remains on extending human capabilities beyond physical presence and helping move robotics towards practical deployment.
The live apple quality-control deployment demonstrated the potential of combining humanoid robotics, AI, cloud technology and continuous operational data to solve real-world tasks.
The accelerator provided an opportunity to share that progress and learn from others working towards the same goal: making robotics more useful, more accessible and more capable of delivering value in the real world. Robotics is not just about building machines that can act. It is about enabling those actions to deliver something meaningful, and that is where the next phase of robotics innovation will be measured.
Extend Robotics would like to thank Google DeepMind, Google Cloud and the wider Robotics Accelerator community for the opportunity to participate in the programme and share in the learning, collaboration and innovation of the cohort.
For more information contact us.
