Google DeepMind has teamed up with Boston Dynamics, which is now hiring big time to build smart humanoid robots
Google DeepMind and Boston Dynamics have partnered to blend advanced AI with humanoid hardware, suggesting a serious push to move the Atlas robot from demos to actual industrial operations.
Humanoid robots have been a topic of discussion for years, but Google DeepMind and Boston Dynamics now want to take the conversation further with an engaging demo. The two companies have announced a new artificial intelligence partnership that aims to turn humanoid robots into something businesses can actually use at scale, not just admire in videos.
At the center of the collaboration is Boston Dynamics’ Atlas humanoid, paired with Google DeepMind’s Gemini Robotics AI model. The goal is simple on paper but ambitious in implementation: to make Atlas smarter, more adaptable, and capable of handling real industrial tasks rather than following tightly scripted routines.
Boston Dynamics says the partnership will focus on visual-language-action models, a key area where current robots often struggle. These models are meant to help Atlas understand spoken or written instructions, read complex environments, and quickly decide how to act. In practical terms, this means a robot that can adjust to unpredictable conditions on the factory floor, rather than being frozen when something changes.
The timing of the announcement is not coincidental. At CES 2026, Boston Dynamics unveiled the latest all-electric version of the Atlas and confirmed that production is starting immediately. The company also revealed that all Atlas deployments planned for 2026 are already locked in. The initial fleet will be sent to Hyundai Motor Group and Google DeepMind, where the robots will begin training for industrial roles, starting with automotive manufacturing.
For Boston Dynamics, this is a clear turning point. The company has spent years improving movement, balance and control, but it has only formally committed to building a commercial humanoid robot in 2024. According to the company, recent advances in AI foundation models have changed what is technically and economically possible, opening the door for humanoids to move out of controlled laboratories and into real workplaces.
“We are thrilled to partner with the Google DeepMind team,” Alberto Rodriguez, director of robot behavior for Atlas at Boston Dynamics, said in a statement. “We are building the world’s most capable humanoids, and we knew we needed a partner that could help us establish new types of visual-language-action models for these complex robots. No one in the world is better suited than DeepMind to build reliable, scalable models that can be deployed safely and efficiently across a wide variety of tasks and industries.”
From DeepMind’s perspective, robotics is an important next step in bringing AI into the physical world. Gemini Robotics is built on the company’s multimodal Gemini system and is designed to combine perception, reasoning, and action in a single framework. This allows robots to work in a variety of environments and tasks rather than being trained only for a limited use case.
“We developed our Gemini robotics models to bring AI into the physical world,” said Carolina Parada, senior director of robotics at Google DeepMind. “We are excited to begin working with the Boston Dynamics team to explore what is possible with their new Atlas robots as we develop new models to expand the impact of robotics and scale robots safely and efficiently.”
The partnership also suggests a major talent boost. In a post on X, the DeepMind CEO said that the company is making rapid progress in robotics and confirmed that Google DeepMind is aggressively hiring in this area. The post also highlighted the recent appointment of Aaron Saunders, former CTO of Boston Dynamics, who has joined DeepMind as vice president of hardware engineering.