Preorders for the Microduck robots have passed 20,000 units, according to Hugging Face's Clem Delangue. The waddling, open source robots went viral after a video showed Nvidia CEO Jensen Huang playing with them during a keynote at the GTC Berlin conference. The robots can sit, nap, and somersault on voice command. They are expected to be delivered in time for Christmas. Pollen Robotics produces the devices, and Hugging Face acquired the company in 2025. Nvidia bought Hugging Face in September.
Nvidia chief executive Jensen Huang has set out a vision for Europe built on robotics, AI agents and modular data centres. Speaking at Nvidia's conference in Berlin, he said Europe could both use and produce artificial intelligence. His first major announcement was a pre-assembled data centre unit called the DSX Pod, which he said could be running within 14 days. Huang also named humanoid robotics as the next major industry and said Europe was well placed to lead it. Two Greenpeace protesters interrupted the speech with a banner.
Google has made TorchTPU, its version of the popular PyTorch toolkit, available as an open-source project. The tool lets AI apps run on Google's own chips, called TPUs, and move to Google Cloud with a simple configuration change. Amazon released a similar version, TorchNeuron, a day earlier for its own Trainium chip. Most AI apps have been built to run on Nvidia's chips, and both companies are working to reduce that advantage. Developers are not expected to leave Nvidia's hardware en masse, though.
AU Now app
Your suburb’s news, alerts and Shorts — in one app
Alerts, Shorts and your saved stories — free on Google Play.
Amazon Web Services has released a tool called TorchNeuron that lets AI developers move apps built with PyTorch onto its Trainium chip. Amazon says a single configuration change is enough, with no code rewrite needed. The move targets the software advantage that has made Nvidia the default choice for AI hardware. AWS argues that Trainium offers better value for startups than specialist cloud providers tied to Nvidia. Amazon also says it will contribute code to PyTorch to make it less dependent on Nvidia chips.
Nvidia has launched new open-weight models designed for always-on agents and edge computing. The Nemotron 3.5 Super VL is a 120-billion-parameter model aimed at enterprise agents, which Nvidia says can match the accuracy of the larger Nemotron 3 Ultra in instruction-following and tool-use tasks. Two lightweight Bolt models are designed to run locally on new AI PCs and DGX Spark workstations. Nvidia says the Super VL completes certain agentic tasks faster than competing models from Anthropic and OpenAI, and lists companies including EY, Salesforce and Perplexity as users.
