Humanoid wants to make one thing clear. The future of humanoid robotics is no longer a distant lab experiment. It is already being tested, trained, and deployed in environments that look a lot like the real world.
The UK based AI and robotics company is accelerating the pace of humanoid robot development by building deeply into NVIDIA’s robotics ecosystem. By integrating NVIDIA Jetson Thor, NVIDIA Isaac Sim, and NVIDIA Isaac Lab into its core technology stack, Humanoid has moved quickly from concept to capable alpha systems that are already operating outside controlled research settings. Those efforts recently culminated in a successful proof of concept with Schaeffler, signaling that these robots are beginning to cross the line from experimental to operational.
At the heart of Humanoid’s approach is Jetson Thor, NVIDIA’s latest edge computing platform, which powers the HMND 01 Alpha robots across two distinct platforms. One is a wheeled system designed for industrial environments, where mobility, robustness, and reliability matter most. The other is a newly released bipedal robot, aimed primarily at research and development for future service and household applications. In both cases, Jetson Thor functions as a powerful AI native brain at the edge, allowing Humanoid to run large robotic foundation models directly on the robot rather than relying on remote infrastructure.
That local intelligence matters. By simplifying system architecture and reducing wiring complexity, Jetson Thor improves manufacturability, reliability, and serviceability in the field. It also unlocks a level of autonomy and responsiveness that brings the robots closer to human environments, not just factory floors.
To get there faster, Humanoid has adopted a simulation first mindset. Using NVIDIA Isaac Lab and NVIDIA Isaac Sim, the team develops, tests, and refines robot behavior long before hardware touches the ground. Isaac Lab plays a central role in reinforcement learning, helping improve locomotion and manipulation skills through large scale simulated training. Meanwhile, Isaac Sim enables high fidelity digital twins that mirror the real robots down to their interfaces, allowing engineers to move seamlessly between virtual and physical systems.
That continuity is key. Engineers can test middleware, control systems, teleoperation, navigation, and full body control in simulation, then deploy the same software stack on real robots with confidence. Navigation and manipulation policies are validated months ahead of on site deployment, reducing surprises and accelerating iteration once robots arrive in real environments.
Simulation also shapes the hardware itself. By analyzing forces, motion, stability, and torque requirements in virtual environments, Humanoid can make informed decisions about actuator selection, joint strength, mass distribution, and balance before building physical prototypes. For the bipedal robot alone, the team evaluated six different leg configurations in simulation, narrowing down designs and reducing risk well ahead of manufacturing.
The payoff is speed. Humanoid designed and delivered its first alpha prototypes in just seven months, a dramatic compression compared to the typical eighteen to twenty four month hardware development cycle. That pace is further amplified by NVIDIA AI infrastructure, which Humanoid uses to train Vision Language Action models. What once required extensive post training pipelines now takes just a few hours, allowing the team to move from training to deployment at remarkable speed.
In practical terms, this means tasks that traditionally took months and large teams of specialists can now be completed far more quickly. Today, Humanoid can train a policy from scratch and deploy it on a real robot within twenty four hours. That kind of velocity changes how robotics development feels, shifting it closer to modern software iteration than traditional hardware engineering.
NVIDIA’s open libraries and frameworks also play a critical role in surfacing problems early. Using Isaac Sim, the team validated navigation and SLAM performance, stress tested stability and torque limits, and optimized sensor and camera placement to ensure reliable perception in complex real world environments. These insights feed directly back into both hardware and software decisions, tightening the feedback loop across the entire system.
This matters because Humanoid is not building robots to live only in demos or labs. As a commercially focused company, its strategy is to place robots into real operational environments as early as possible, gather feedback, and iterate fast. That approach is already producing results. Humanoid has received 20,500 pre orders, completed six proof of concepts, and currently has three additional pilot programs underway.
According to Jarad Cannon, Chief Technology Officer at Humanoid, NVIDIA’s open robotics development platform is helping the industry move beyond legacy industrial communication standards and toward modern, software defined systems. The company is working closely with NVIDIA and other partners on a new robotics networking system built on Jetson Thor and the Holoscan Sensor Bridge. The goal is to establish an open network standard for AI enabled robots, one that could reshape how robotic systems communicate, scale, and evolve across industries.
That vision comes to life on the CES show floor. Humanoid will publicly demonstrate the HMND 01 Alpha Robot at the Schaeffler Booth 7301 in the West Hall of the Las Vegas Convention Center. Founder Artem Sokolov will present the robot performing a near real world industrial task: bin picking metallic bearing rings in a cluttered environment, executed in a near production setting. Demonstrations will take place on Tuesday, January 6 at 10 a.m., 11 a.m., noon, 2 p.m., and 3 p.m.
It is a small task on the surface, but a telling one. Behind that bin picking demo sits a broader signal about where humanoid robotics is headed. Faster development cycles, AI native systems, simulation driven design, and robots that are increasingly ready to leave the lab and step into everyday environments.