Tuesday 15 September 2026
Evening EditionRSS

ZOTPAPER

News without the noise

ZOTPAPER
Hardware & Devices

Hugging Face's Pollen Robotics unveils $399 Microduck robot for AI training

Bipedal duck-like robot offers open-source software stack, reinforcement learning, and a playful entry point into hardware AI — but hardware remains proprietary

edit
By LineZotpaper
Published
Read Time3 min
Pollen Robotics, a subsidiary of AI model hosting platform Hugging Face, has announced the Microduck, a $399 bipedal robot duck designed to teach users about hardware AI and machine learning. Standing just 25 cm tall and weighing 800 grams, the robot comes with an open-source software stack, onboard sensors, and a simulation environment for reinforcement learning, but its hardware is not open source. The announcement comes amid reports that Nvidia is in talks to acquire Hugging Face for approximately $13 billion.

Pollen Robotics, based in Bordeaux, France, introduced the Microduck on Thursday, describing it as a robot designed for immediate action and ready for training via reinforcement learning. The tiny robot — roughly the size of a plush toy — is intended as an affordable testbed for developers, researchers, and hobbyists to explore AI on real hardware.

Microduck ships with 1 GB of RAM, 32 GB of onboard storage, and a 50 Hz onboard policy loop that connects its AI to sensors for real-time decision-making. It features a front-facing camera, a built-in LiDAR unit for depth sensing, and an articulated beak capable of grasping objects weighing up to 100 grams. A video on the product's website demonstrates the robot picking up socks and a marker and placing them in a box. Microduck can also right itself when knocked over and, with an optional accessory pack, roller skate, play with a ball, and follow a laser pointer.

While the hardware is proprietary, Pollen Robotics has released the entire software stack as open source, including robot control, simulation, reinforcement learning training, and sim-to-real deployment. The company encourages users to share skills once the Microduck system goes live later this year. "Anyone can train new behaviors … in simulation using reinforcement learning," Pollen representative Santiago Pavon wrote in an email. "Adding purpose and autonomy to the robot is something we are still working on, and we expect the community to pick up and contribute."

The announcement comes against the backdrop of a reported acquisition deal. Business Insider, citing a person familiar with the matter, reported that Nvidia has agreed to buy Hugging Face for around $13 billion. The deal has not been confirmed, and Hugging Face and Nvidia have not commented. If true, it would represent Nvidia's first major push into AI model hosting and open-weight AI, a space where it already participates through its own Nemotron models.

Microduck is available for preorder now, with units expected to ship before Christmas 2026. Pollen Robotics hopes the community will help achieve autonomous behaviors such as voice recognition, simultaneous localization and mapping, navigation, and grasping — though the robot's small size and payload limit its practical applications.

§

Analysis

Why This Matters

  • Microduck lowers the barrier to entry for learning about AI robotics, offering a fully functional platform at a price point accessible to individual developers and educational institutions.
  • The open-source software stack, combined with simulation-based training, could accelerate experimentation in small-scale autonomous behavior, potentially leading to transferable techniques for larger robots.
  • The rumored Nvidia acquisition of Hugging Face would signal a major strategic shift for the GPU giant, further intertwining hardware acceleration with open-weight AI model distribution.

Background

Pollen Robotics was acquired by Hugging Face in 2023, as part of the company's expansion into robotics and hardware AI. Hugging Face has become the leading repository for open-source AI models, hosting hundreds of thousands of models and datasets. Nvidia, meanwhile, has been building its own ecosystem of AI hardware and software, including its Nemotron series of open-source models. The reported $13 billion acquisition — if confirmed — would be one of the largest in AI history, giving Nvidia direct control over the dominant model distribution platform.

Key Perspectives

Educators and developers: Microduck offers a low-cost, fully functional robot for teaching reinforcement learning, computer vision, and control systems. The simulation-first approach reduces the risk of hardware damage during training. Hobbyists and makers: The robot's playful design and optional accessories make it appealing for fun projects, though the closed hardware may disappoint those who prefer fully open-source platforms. Critics: The 100-gram payload limit severely restricts practical applications. The proprietary hardware means users cannot modify the physical design. Additionally, the rumored acquisition raises questions about Hugging Face's future independence and commitment to openness.

What to Watch

  • Preorder numbers and shipping timeline: If Microduck sells out quickly, it would indicate strong demand for affordable AI robotics platforms.
  • Confirmation of the Nvidia-Hugging Face acquisition: Any official announcement would reshape the AI hardware and software landscape.
  • Community contributions to the open-source software stack: The success of Microduck as a learning platform depends on whether users build and share new skills.

Sources

newspaper

Zotpaper

Articles published under the Zotpaper byline are synthesized from multiple source publications by our AI editor and reviewed by our editorial process. Each story combines reporting from credible outlets to give readers a balanced, comprehensive view.

Editorial ProcessSubscribe