Nvidia selects Unitree for humanoid research platform

CNBC reports that Nvidia has selected Chinese robot maker Unitree to supply the humanoid body for the chipmaker's first publicly available humanoid robotics system for researchers. The package pairs Unitree's nearly 6-foot-tall H2 humanoid with Nvidia's Jetson Thor hardware, which includes the Blackwell GPU, and will include Nvidia's humanoid-focused AI models Isaac GR00T and simulation tools, CNBC reports. Sales, primarily to research institutions including Stanford and ETH Zurich, are set to begin later this year, according to the report. CNBC also quotes Nvidia CEO Jensen Huang describing a reference humanoid, the "Nvidia Isaac Root," with hands providing 25 degrees of freedom and the robot offering 31 degrees of freedom, and noting a weight of 150 pounds in his keynote remarks.
What happened
CNBC reports that Nvidia has selected Chinese humanoid-maker Unitree to provide the body for the chipmaker's first research-focused humanoid robotics system. Per CNBC, the system pairs Unitree's almost 6-foot-tall H2 humanoid with Nvidia's Jetson Thor hardware that contains the Blackwell GPU, and it will include Nvidia's humanoid-focused models Isaac GR00T and simulation tooling. CNBC says sales, targeted primarily at research institutions including Stanford and ETH Zurich, are due to start later this year. CNBC also quotes Nvidia CEO Jensen Huang describing the reference humanoid, the "Nvidia Isaac Root," with hands offering 25 degrees of freedom, the robot having 31 degrees of freedom, and a weight of 150 pounds.
Editorial analysis - technical context
Bundling a high-performance inference platform (Blackwell-class GPU inside Jetson Thor) with a pre-built humanoid body removes substantial mechanical and systems-integration overhead for labs. Industry-pattern observations: research groups often accelerate iteration when vendors deliver integrated hardware-software stacks, because less time is needed to integrate sensors, actuators, and on-device compute. For practitioners, that typically shifts effort toward model development, simulation, and control software rather than low-level hardware integration.
Context and significance
Industry context: Nvidia has been pushing what CEO Jensen Huang and coverage term "physical AI," arguing for large market potential; CNBC reports Huang referenced the market scale and showcased the integrated reference robot during a keynote. The combination of a mainstream chip vendor providing on-device Blackwell-class compute with a commercially produced humanoid body represents a step toward lowering the barrier for institutions to run embodied-AI research at scale. Observed patterns in similar vendor-driven research kits show faster adoption by university labs and startups that lack in-house robotics engineering capacity.
What to watch
Indicators observers should follow include availability and pricing for the integrated kit (CNBC reports sales will start later this year), latency and thermal behavior of Blackwell GPUs in continuous humanoid workloads, the maturity of the Isaac GR00T model suite for real-world control tasks, and which research groups publicly publish work using the platform. Also track Unitree's investor mentions in the coverage (PitchBook is cited as listing Qiming Venture Partners among backers) for signals about funding or corporate milestones.
Scoring Rationale
This is a notable product-level announcement: Nvidia bundling Blackwell-class on-device compute with a commercial humanoid body lowers integration friction for research groups. It is not a frontier-model release, but it meaningfully affects embodied-AI experimentation and lab deployment.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems

