Google's robotics strategy: software-first, replicating the Android approach to develop embodied intelligence.

Google is approaching the robotics field with a different mindset than companies like Tesla: instead of rushing to build flagship hardware, it prioritizes developing the models and software capabilities behind the robots and expanding their applications with the help of partners.

According to a recent report by The Information, Google DeepMind CEO Koray Kavukcuoglu spoke publicly for the first time last week in relatively detail about the company's robotics strategy. He stated that Google's core strength lies in its models, not the robot chassis themselves. At the same time, Google launched Gemini Robotics—a version of the Gemini model optimized for physical control scenarios—and has already begun collaborating with robotics companies such as Boston Dynamics.

This model is quite similar to Google's early promotion of Android: Google provides the core software, and then achieves large-scale expansion through hardware from different manufacturers, rather than relying on its own devices from the beginning. This also clearly distinguishes Google from Tesla and Figure, which is supported by OpenAI—both of which have made building flagship humanoid robots a key focus. For Google, robots are more like new application scenarios for extending Gemini's capabilities into the real world.

Software first, replicate Android expansion path

Kavukcuoglu stated at The Information's AI Agenda Live event that Google hopes to "collaborate with other robotics companies to enable them to use this system, thereby injecting intelligence into all robots in a safe and effective way."

This strategy mirrors Google's approach to promoting Android. The report points out that between 2007 and 2008, Google co-founder Sergey Brin responded to questions about whether Google would develop its own mobile phone business with a similar line of thinking. At that time, Google was partnering with T-Mobile and HTC to launch Android devices before gradually developing its Pixel hardware business.

Kavukcuoglu has not completely ruled out the possibility of Google creating humanoid robots in the future, but emphasizes that the current focus is still on model capabilities and the collaborative ecosystem.

Gemini enters the physical world, intensifying the competition for its models.

Google's robotics deployment is also an extension of its AGI strategy. Kavukcuoglu views Gemini Robotics as a natural extension of the Gemini multimodal architecture into the field of physical control: the same underlying model capable of handling text, vision, and audio is being further used for robot control, rather than developing dedicated models for different tasks.

He sees Waymo's autonomous driving as a key validation case for this approach, calling it a "very, very safe way" to demonstrate the large-scale deployment of physical AI. If the Gemini architecture can support Waymo's commercial operation, its feasibility for migration to other physical scenarios will be further validated.

Meanwhile, Google announced this week the initial release of Gemini 4 Argon to select enterprise customers, primarily for cybersecurity defense, with plans to expand to a wider user base soon. As the new model enters the application phase, Google's ability to further narrow the capability gap with Anthropic and OpenAI will be a crucial window for the market to observe its AI competitiveness.

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.