Hugging Face Draws Nvidia Takeover Interest While Launching a $399 Open-Source Robot

Nvidia has discussed acquiring Hugging Face in a transaction valued at roughly $13 billion, while the companies have not confirmed a signed agreement. The reports place Hugging Face, a central hub for open-weight AI models, at the center of Nvidia’s effort to expand beyond chips and computing hardware.

The status of the talks remains unsettled. The Information reported that Nvidia had agreed to buy the company for $12.9 billion, while Business Insider reported that negotiations had not produced a signed deal. Thomas Wolf, Hugging Face’s co-founder and chief science officer, declined to comment on the reports during a Bloomberg Tech interview, saying the company has consistently attracted acquisition offers.

Hugging Face has become a major distribution and collaboration layer for open-weight models. Developers use its platform to store, download, share, and work with models, giving Nvidia access to a community and infrastructure network that extends across research teams, startups, enterprises, and independent builders.

That position carries strategic value even as the company remains unprofitable, Ars Technica reported. Nvidia has also faced difficulty building its own cloud AI business, and acquiring Hugging Face could give it a ready-made software and developer channel alongside its dominant accelerator business. The deal would place Nvidia closer to the model ecosystem as OpenAI, Anthropic, and other AI labs develop more of their own infrastructure.

Hugging Face is broadening its open-source mission into robotics with Microduck, a small bipedal robot priced at $399. Developed with Pollen Robotics, the roughly 25-centimeter, one-eyed machine is designed for home experimentation and can be retrained to perform new behaviors through reinforcement learning.

Demonstrations show Microduck waddling, picking up objects with its beak, recovering after falls, crouching, kicking a ball, and moving on tiny roller skates. The robot can lift objects weighing up to 800 grams, and Pollen Robotics says preorders will ship before Christmas 2026. Bloomberg and TechCrunch described the device as part of Hugging Face’s effort to make robotics and physical AI more accessible.

Microduck gives Hugging Face a physical product that turns model training into an observable, hands-on activity. Developers can teach the robot tricks at home, test policies on a low-cost platform, and share models or behaviors through the same open ecosystem associated with Hugging Face’s software tools.

The product also places an approachable consumer object around difficult technical ideas such as reinforcement learning, embodied AI, and world models. Its small size and $399 price keep the entry point below research-grade robotic systems, while the open-source design invites experimentation rather than limiting owners to a fixed set of factory behaviors. The skating duck is deliberately playful, but its role is practical: it provides a compact test bed for physical AI.

An acquisition would connect Nvidia’s hardware reach with Hugging Face’s model repository, developer community, and robotics projects. Nvidia could use that combination to strengthen software adoption around its chips, build cloud services around open models, and compete more directly for developers who choose models before they choose infrastructure.

The transaction would also raise questions about stewardship of an ecosystem built on openness. Hugging Face hosts projects from many organizations and individual developers, so Nvidia ownership would bring a powerful commercial player into a platform valued for broad access. The reported price reflects Hugging Face’s strategic position more than its current profitability, according to Ars Technica, while the conflicting deal reports show that the business terms remain unresolved.

Hugging Face now faces two major storylines at once: a reported takeover approach from the world’s leading AI chip company and a push to make open robotics tangible for ordinary developers. Neither storyline has been fully settled, but together they show how the company’s influence has expanded from model hosting into hardware, training, and embodied systems.

The immediate milestones are clear. Nvidia and Hugging Face would need to confirm or abandon the reported acquisition, while Microduck must ship on schedule and attract a community willing to build new behaviors for it. If both efforts gain traction, Hugging Face will have strengthened its position as a bridge between open models and physical machines.