Nvidia CEO Jensen Huang spent part of his GTC Berlin keynote playing with a flock of tiny walking robots called Microducks. The clip went viral, and preorders followed fast. According to TechCrunch AI, Hugging Face CEO Clem Delangue tweeted that preorders passed 20,000 units by the next day.
The robots are expected to ship before Christmas. That puts them in the running for this year’s hottest holiday gadget, and Delangue thinks they could become the best-selling robot of the year.
What Exactly Is a Microduck?
Microducks are small, open source robots made by Pollen Robotics. Hugging Face bought Pollen in 2025. TechCrunch AI reports that Nvidia then acquired Hugging Face in September, so the Microduck is now technically an Nvidia product.
This is Hugging Face’s second consumer robot. The first was Reachy Mini, which is also very cute.
5 Things Worth Knowing About the Microduck Launch
- They respond to voice commands. On stage, the ducks stomped around, sat down, flopped over for a “nap,” and did somersaults when told to. They chirp with tiny R2-D2-style sounds while they do it. It’s a simple trick list, but it’s exactly the kind of personality that sells robots to regular people.
- They learned to move in simulation. The Microducks were trained on Nvidia’s technology stack. Virtual copies of the robots learned basic skills like walking, sitting, and standing inside a simulation, and those skills ship on the real hardware. This “sim-to-real” approach is how Nvidia wants robotics to scale: train cheaply in a virtual world, then move the result into a physical body.
- You can teach them new tricks. Pollen says owners can train their Microducks on new behaviors and share them with the community. That turns a toy into a platform. If enough people build and swap tricks, the robots get more capable over time without anyone buying new hardware.
- They’re open source. This fits Hugging Face’s whole identity. The company built its reputation as the place to share open AI models, and it’s bringing that same idea to hardware. Open source robots invite hobbyists, students, and researchers to tinker, which usually builds a much bigger ecosystem than a closed product would.
- Adults are the real audience. Delangue suggested the ducks might not even be mainly for kids. The promo videos show grown-ups playing with them too. That makes sense. The people most excited about a programmable, AI-trained robot are probably developers and tech fans, not eight-year-olds.
Why This Matters
What stands out here is the strategy, more than the cuteness. Nvidia has been talking up “physical AI” for a while now, and most of that talk has been about factory arms and humanoid robots that cost as much as a car. Microducks come at it from the other end: cheap, friendly, and built for the living room.
A viral consumer robot does some useful things for Nvidia:
- It shows off the stack. Every duck is a walking demo of Nvidia’s simulation and training tools.
- It builds a developer base. People who teach a duck new tricks are learning robotics workflows that carry over to bigger machines.
- It makes AI feel friendly. TechCrunch AI pointed out that the launch landed in a rough week for the industry, with AI blamed in a string of cyberattacks. A waddling duck is a much nicer headline.
The Caveats
Plenty is still unknown. The source doesn’t say what a Microduck costs, which regions it ships to, or how hard it is to train new behaviors. Twenty thousand preorders is a big number for a hobby robot, but preorders aren’t deliveries. Hitting a Christmas ship date at that volume is a real manufacturing test for a small robotics team.
Open source hardware also tends to reward tinkerers more than casual buyers. If the community side takes off, the ducks will keep getting more interesting. If it doesn’t, they could end up as a fun gadget with a short trick list.
What Comes Next
If the Microducks actually ship on time and the trick-sharing community grows, Nvidia and Hugging Face will have something the robotics industry has struggled to build for years: a consumer robot people actually want. Watch the delivery timeline closely. You can find the original video and more details in TechCrunch AI’s full report.