Apptronik expands Robot Park, deploys Apollo 2 humanoid robots with Google DeepMind

Apptronik's strategy centers on creating a closed-loop system between hardware deployment and AI training. By operating Apollo 2 robots across distributed locations—its own Robot Park facility and customer sites—the company generates a continuous stream of real-world data. This data is fed into Google DeepMind's Gemini Robotics models, which are then refined and deployed back to the robots. The CEO's comment that the company is focused on "what they can do every day on the job," rather than demos, signals an emphasis on practical, production-ready deployment over laboratory proofs of concept.

The modular design of Apollo 2—offered in both bipedal and wheeled forms—appears to address a customer demand constraint: existing industrial facilities have safety standards and workflows built around mobile robots on wheels. By offering the same core technology in a wheeled variant that complies with those standards, Apptronik can expand data collection across more operational environments today while refining the bipedal walking capability in parallel. This approach enables the company to gather diverse task experience (logistics, manufacturing, retail, and others) needed to train embodied AI models that can generalize across environments.

The announcement of Apollo 3 as the next commercial product suggests that the year-plus of Apollo 2 deployment is explicitly feeding into the development of a commercial fleet. The body indicates that Apollo 3 is expected to arrive with "unprecedented out-of-the-box embodied intelligence," implying that the massive data streams from Apollo 2 operations will materially improve the AI models that ship in the next generation.