Nvidia is set to acquire the open-source model hosting platform, with its sights set on the company's pivotal strategic role within the open-source AI ecosystem. This acquisition would simultaneously bring in a rapidly growing robotics division, potentially helping Nvidia CEO Jensen Huang fulfill his vision for Physical AI.

Late last month, Hugging Face introduced the Microduck, a compact robot capable of autonomous walking that has also learned skills such as rollerblading and using laser pointers to play with cats. Hugging Face co-founder Thomas Wolf revealed that despite minimal promotion limited to X and LinkedIn, sales of the product have already surpassed 15,000 units, generating over $6 million in revenue.

This new robot is part of Hugging Face's strategy to offer affordable open-source AI robots to everyday consumers and developers. Wolf explained that the company's acquisition of Pollen Robotics in April 2025 facilitated this initiative. Pollen Robotics, which originally specialized in high-priced humanoid robots, has pivoted toward developing more accessible and affordable robotic products following the acquisition.

The Microduck's accompanying software is fully open-source, allowing developers to freely download and modify it, providing a low-cost experimental platform for developing more complex humanoid robots. Wolf noted that to train a robot capable of high-speed running, one could either experiment with a $40,000 humanoid prototype or use this $400 device for rapid iteration.

Meta, Apple, and OpenAI are all actively exploring the AI consumer hardware sector. Nvidia itself does not directly sell robots, instead providing models and software tools to robot manufacturers. Should products like the Microduck achieve market success, Nvidia's acquisition of Hugging Face could change this dynamic.

Wolf stated that Hugging Face initially projected sales of at least 50,000 Microduck units in its first year, corresponding to approximately $20 million in revenue. For comparison, 1X Technologies launched its home humanoid robot Neo last year, receiving only 10,000 pre-orders. Hugging Face has now revised its sales expectations upward, potentially boosting the company's overall annualized revenue.

Beyond hardware sales, robot training itself consumes significant computing power, which would also benefit Nvidia's business. According to the project's GitHub documentation, training the Microduck requires running simulation environments that depend on Nvidia GPUs and CUDA software.

Hugging Face previously launched the Reachy Mini robot for general users, developers, and researchers, but it lacks autonomous walking capabilities, relying primarily on vision, hearing, and speech for interaction. Wolf noted that the Reachy Mini also sells for $400 and has accumulated approximately 10,000 units sold.

The commercialization path for consumer-grade robots remains challenging, with the core difficulty being the complexity and variability of home environments that robots struggle to adapt to. The most successful home robot of the past two decades has been the Roomba vacuum, but its manufacturer filed for bankruptcy in December. Other companies are racing to fill this market gap, yet competing products are priced as high as $1,245.

Some may question whether consumers are willing to spend $400 on a robotic duck designed more for "causing mischief" than performing household chores. However, if the market indeed demands this type of entertainment-focused AI hardware, and low-cost compact robots become the mainstream experimental platform for humanoid robot development, Hugging Face could be poised for significant opportunities.