DeepMind Hires Former Boston Dynamics CTO to Build the ‘Android of Robotics’

Google DeepMind is making it clear that its ambitions extend far beyond the digital realm. The AI lab has hired Aaron Saunders, the former Chief Technology Officer of Boston Dynamics, as its new Vice President of Hardware Engineering.

The hire, first reported by Wired, represents a significant injection of mechanical DNA into a company that has historically focused on pure software. Saunders spent over two decades at Boston Dynamics, where he was instrumental in the development of iconic machines ranging from the quadruped BigDog to the commercial Spot robot and the humanoid Atlas.

His arrival comes as DeepMind CEO Demis Hassabis articulates a strategy to make the Gemini AI model a universal "operating system" for robots—an "Android for the physical world" that can function across different machine bodies.

The 'Android' Play

While DeepMind has long used simulation to train agents, the company is increasingly focused on "Physical AI"—the messy, difficult task of deploying foundation models into real-world hardware.

Hassabis envisions a future where Gemini serves as a general-purpose brain capable of controlling diverse robot forms "out-of-the-box." This aligns with recent comments from DeepMind’s robotics lead, Carolina Parada, who detailed the company's "north star" of creating a single, general-purpose AI smart enough to embody any robot.

However, bridging the "sim-to-real" gap—where AI models trained in virtual environments fail when facing the friction and unpredictability of the real world—remains a massive hurdle. By bringing in Saunders, DeepMind gains a leader who has spent 23 years solving exactly those kinds of hardware-software integration problems.

A Full Circle Moment

There is a degree of irony in the appointment. Google’s parent company, Alphabet, acquired Boston Dynamics in 2013 before selling it to SoftBank in 2017 (it is now majority-owned by Hyundai Motor Group).

During Google's ownership, culture clashes between the "moonshot" robotics engineers and Google's software-focused management were widely reported. Saunders’ return to the Google fold—this time under the DeepMind banner—suggests a convergence of these philosophies. The industry has reached a point where hardware has become capable enough to support advanced AI, and AI has become capable enough to warrant sophisticated hardware.

From BigDog to Gemini

Saunders left Boston Dynamics earlier this summer after a tenure that saw the company transition from a research-heavy R&D lab to a commercial entity. His experience covers the full spectrum of robotics development, from the "Phase One" hardware validation challenges to the complexities of scaling fleets.

At DeepMind, Saunders will likely oversee the hardware platforms necessary to test and validate Gemini’s agentic capabilities. While DeepMind is not expected to pivot to manufacturing robots for mass sale, having bespoke, high-performance hardware is critical for closing the feedback loop for their foundation models.

The Race for Physical Intelligence

The hiring comes at a time of intense competition. Tesla is aiming to produce millions of Optimus bots, Figure AI is deploying robots into BMW plants, and Chinese competitors like Unitree are rapidly driving down the cost of hardware.

Hassabis predicted in his interview with Wired that AI-powered robotics will have its "breakthrough moment" in the next couple of years. With the brain of Gemini and the engineering expertise behind Atlas now under one roof, DeepMind is positioning itself to be the architect of that breakthrough.