Apptronik Builds a 90,000-Square-Foot Data Factory to Train Apollo 2 Humanoids
Apptronik's 90,000 sq ft Robot Park and Apollo 2 form a live data pipeline feeding Google DeepMind's Gemini Robotics foundation model, with Apollo 3 in the crosshairs.

- Apptronik opened a nearly 90,000 sq ft Robot Park in Austin where Apollo 2 humanoids continuously generate real-world training data for AI models.
- Data from Apollo 2 feeds directly into Google DeepMind's Gemini Robotics foundation models, advancing the AI that will power future commercial humanoid fleets.
- Apollo 2 ships in both bipedal and wheeled configurations; the wheeled version meets existing industrial safety standards for faster enterprise deployment.
- Apptronik has raised $935M at a $5.3B valuation, backed by Google, Mercedes-Benz, B Capital, John Deere, and the Qatar Investment Authority.
- All data collected now is feeding development of Apollo 3, the first fully commercial product, expected in production by 2027.
- Google DeepMind is simultaneously partnering with Boston Dynamics and Agile Robots, positioning Gemini Robotics as a shared AI layer across multiple hardware competitors.
The hardest problem in humanoid robotics has never been building the robot. It has been teaching it to work. Apptronik just made its most concrete bet yet on how to solve that: a massive, purpose-built data factory in Austin, Texas, where fleets of its Apollo 2 humanoids clock in every day, not to ship products, but to generate the training data that will make future robots actually useful.
A Factory That Makes Data, Not Widgets
Apptronik announced the opening of the newly expanded Robot Park, its flagship data collection and training facility for humanoid robots in Austin, Texas, anchoring a growing global network of Robot Parks at customer and partner sites around the world, with plans to open new locations in more cities soon. Inside the nearly 90,000-square-foot facility, Apollo 2 robots in bipedal and wheeled configurations learn across an extensive array of customer use cases, performing tasks in logistics, manufacturing, and retail.
The facility operates seven days a week, with fleets of humanoid robots moving boxes, sorting items onto conveyor belts, and practicing logistics routines under the guidance of remote operators. Through a combination of teleoperation and autonomous execution, Apollo 2 robots continuously generate significant quantities of high-quality training data, which is used to train and refine the Gemini Robotics AI models that will prepare Apptronik's commercial fleet for real-world deployment.
The Google DeepMind Connection
As part of Apptronik's research partnership with Google DeepMind, the high-quality data collected by Apollo 2 helps to advance Gemini Robotics, Google DeepMind's foundational AI models for robotics. Gemini Robotics is a vision-language-action (VLA) model, meaning it takes in camera feeds and language instructions and outputs physical robot movements, rather than relying on hand-coded task scripts. Gemini Robotics can be specialized for complex embodiments such as the humanoid Apollo robot developed by Apptronik, with the goal of completing real-world tasks.
DeepMind has adapted Gemini Robotics On-Device, a lower-latency, locally-executed variant, to Apollo, enabling on-robot inference without cloud round-trips and requiring as few as 50 to 100 demonstrations to adapt to new tasks. That last number matters: the ability to teach a robot a new skill in under 100 demonstrations, rather than thousands, is what makes real-world deployment economically viable.
DeepMind is both investor and collaborator, and the two have been partners for nearly two years. But the CEO of Apptronik, Jeff Cardenas, has been candid about the tension baked into that arrangement. "Google has been honest with us from the very beginning about their ambitions to build the Android for robotics," he said. "That means their model will go to many different partners." His bet is on the depth of the relationship rather than exclusivity.
Apollo 2: The Workhorse Behind the Curtain
Apptronik also unveiled Apollo 2, the current version of its humanoid robot platform, in both bipedal and wheeled-base configurations. The robot has actually been operating quietly for over a year. Apollo 2 has existed since February of last year without a big announcement, appearing suddenly on Apptronik's website with little fanfare. The public reveal now is a signal that the data collection operation has reached a scale worth talking about.
The two-body strategy is deliberate:
- The wheeled-base configuration is designed to conform with existing safety standards for industrial mobile robots, allowing it to fit easily into existing customer operations.
- The bipedal configuration provides maximum adaptability for complex environments, allowing Apptronik to continuously refine the safety and reliability of its walking platform in real-world scenarios.
- The robot stands roughly 6 feet tall, can lift 55 pounds with both hands, and runs for about four hours on an upgraded battery pack with improved motors and sensors.
- Apollo 2 incorporates proprietary high-efficiency actuators, swappable batteries for extended operation, and multiple charging options to maximize uptime.
The company captures high-quality data through a state-of-the-art teleoperation stack and high-fidelity physics simulations to accelerate both hardware design and algorithmic development. The combination of real-world and simulated data is the same playbook used in large language model training: more diverse data sources, better generalization.
The Network Effect Nobody Is Talking About
Robot Park is not just one building. Similar data-collection workflows have been deployed across a growing network of Robot Parks including at research partner Google DeepMind and at customers like Mercedes-Benz and GXO, a global leader in high-tech, high-volume logistics. Each site adds a different flavor of real-world experience: automotive assembly lines, high-throughput logistics, and pure AI research environments all generate meaningfully different data.
The logic is simple: more real-world task data leads to better AI models, which enable more capable robots, which can then collect even more complex data. This is the data flywheel that every serious robotics company is chasing right now, and Apptronik is one of the few that has the hardware fleet, the facility footprint, and the foundation model partner to actually run it.
The Money and the Roadmap
Apptronik has raised a total of $935 million, and TechCrunch separately learned that its post-money valuation is now about $5.3 billion. The latest $520 million tranche came from earlier investors Google, Mercedes-Benz, and B Capital, alongside some new investors. New participants include AT&T Ventures, John Deere, and the Qatar Investment Authority.
The money has a clear destination. While most humanoids have historically cost hundreds of thousands or millions of dollars, Apollo has a bill-of-materials targeting under $50,000, making it economically viable for regular industrial use. Apptronik's business is built on a Robotics as a Service model, which includes the robot hardware, software updates, service and support.
Apollo 2 is explicitly a stepping stone. Everything Apptronik is proving and learning through the Apollo 2 platform is directly powering the development of its next-generation commercial product, Apollo 3. By leveraging the massive data streams generated by Apollo 2 today in partnership with Google DeepMind, the upcoming commercial fleet will debut with unprecedented out-of-the-box embodied intelligence. A commercial-grade Apollo 3 is already in development, with real production versions expected in 2027 and beyond.
The Competitive Stakes
Apptronik is not operating in a vacuum. Google DeepMind is simultaneously running a separate Gemini Robotics partnership with Boston Dynamics and its Atlas robot. At CES 2026, Boston Dynamics announced a partnership with Google DeepMind to integrate Gemini Robotics AI, giving Atlas foundational intelligence for perception, reasoning, and human interaction. DeepMind is also working with Agile Robots on its Agile ONE humanoid. The Gemini Robotics platform is shaping up to be the Android of robotics, with multiple hardware partners competing on top of a shared AI layer.
The broader competitive landscape is fierce:
- Tesla Optimus: Tesla announced it would convert its Fremont assembly line into the first-generation Optimus production line, designed for capacity of up to one million units per year, with external commercial sales not expected until late 2026 at the earliest.
- Figure AI: Figure AI achieved a $39 billion valuation after raising $1 billion in a Series C round, positioning it as a leader in the AI-powered humanoid robot market.
- Chinese competition: Average humanoid robot prices fell from approximately $85,000 in 2023 to $25,000 in 2025, driven largely by Chinese manufacturers, even as gross margins improved.
The announcement reflects a broader shift across the robotics industry. Companies are increasingly competing not just on robot hardware, but on who can gather the most valuable real-world data to improve AI systems. Apptronik's Robot Park is a direct play for that advantage, and the partnership with Google DeepMind's Gemini Robotics gives it a foundation model partner with the scale to turn that data into something deployable.
The open question, as TechCrunch noted, is whether the intelligence Apptronik helps train eventually flows to its rivals too. Cardenas is betting that being the deepest partner, not the exclusive one, is enough to win.