Summary of Key Points
This report discusses two major events in the AI industry that occurred within the same week and seem unrelated on the surface, but actually signal a turning point for the industry: On one hand, AI chip giant NVIDIA spent $14 billion to acquire Hugging Face, the world's largest open-source AI platform, with an outrageous premium of 86 times its annual revenue of $150 million. On the other hand, Hugging Face launched a plastic toy duck named Microduck for just $399, which wobbles around like a drunk penguin but sells out every 4 seconds. Together, these two events indicate that the AI industry has moved beyond the “large models on screens” and has officially shifted its focus to the real physical world.
---
Explanation in 5 Simple Dimensions
1. The wildly successful Microduck is not a high-tech product; it’s just an inexpensive toy with “intelligence equivalent to that of an ant”
Don’t be fooled by the term “embodied intelligence.” This duck’s technical complexity is far below what most people would expect. It uses an old chip that was released five years ago, and the chip’s AI-specific components are hardly utilized at all. It relies on basic CPU power to run a small neural network with only 10,000 “neurons” (parameters)—compared to GPT-4’s 1.8 trillion neurons, a difference of a hundred million times. The duck can’t chat or use large models; all its movements are pre-programmed in a computer simulator before being implemented in the physical device. It can only perform a few fixed actions (walking, falling, getting up, picking up a rag), and it doesn’t even have a voice dialogue function; it merely emits strange short sounds. Essentially, it’s a toy for geeks that might become boring after a few minutes of use.
2. The $399 price is the real game-changer: It lowers the barrier to embodied intelligence from million-dollar laboratories to the average consumer’s desk
The real value of Microduck isn’t its ability to walk; rather, it makes embodied robots, which were previously accessible only to universities and top tech companies, affordable for ordinary people at around 2,000 RMB. Similar previous robots were either expensive prototypes like Tesla’s Optimus or costly research equipment in labs, out of reach for most students and enthusiasts. Microduck not only sells for a low price but also has all its code, simulator, and development tools open-source. You can “walk” the duck online without buying the physical device, modify the code to make it perform new actions (like doing somersaults or climbing stairs), and developers have even created an AR version to play with in your living room. China’s manufacturing capabilities played a crucial role in making this possible: A company in Shenzhen transformed the 3D-printed prototype into a mass-produced product in just 5 months, with a 99.9% quality rate after 20 days of assembly and testing. Without China’s advanced manufacturing, this price would be unattainable.
3. NVIDIA’s $14 billion acquisition of Hugging Face is a wise move: It targets the millions of AI developers
Many question the rationale for such a large investment in a company with annual revenue of $150 million. However, NVIDIA is acquiring the ecosystem surrounding Hugging Face: the platform has 13 million developers worldwide, over 3 million open-source AI models, and hundreds of thousands of datasets. Whether it’s overseas giants like Meta and Google or domestic companies like DeepSeek and Alibaba/Tencent, if you develop open-source AI, you generally have to upload your models to Hugging Face for others to use. NVIDIA previously only sold GPU chips, like a company that builds highways but sells the cement. Now, by acquiring Hugging Face, it controls the entire ecosystem (chips, development tools, open-source models, and distribution channels). All developers will naturally choose NVIDIA’s solutions, as it offers the most comprehensive and compatible environment. This means NVIDIA can easily earn money without even needing to force them to use its hardware.
4. The simultaneous release of the duck and the massive acquisition signals a shift in the AI industry’s focus
The AI industry has been competing for years to create the largest and most human-like models on screens, but this path has reached its limits. 98% of large models lack practical applications. These two events together announce that the next frontier is the real physical world. The embodied intelligence industry is still in the early stages, similar to GPT-2 in 2019—interesting but far from practical products like ChatGPT. However, Microduck represents the first physical AI product that millions of ordinary developers can easily access and use. This sudden influx of developers will drive rapid innovation.
5. The impact on ordinary people: The popularity of affordable household robots will be faster than you think
This isn’t just a game for industry giants; China’s supply chain is quickly replicating Microduck, using domestic components to further reduce costs. In the next one or two years, you’ll see similar robots for a few hundred to a thousand RMB on e-commerce platforms. They will initially be used as teaching tools and geek toys, but eventually, they may evolve into household robots that can help with tasks or accompany you. The open-source nature means they won’t be monopolized by any one company. Small startups and student teams can modify the code to create their own functions, eliminating the need for massive R&D investments. The pace of innovation in physical AI will likely exceed that of smartphones.
In summary, these developments indicate a significant shift in the AI industry, with the next phase of development focusing on the real world, not just on screens. The popularity of affordable household robots will accelerate rapidly, thanks to China’s manufacturing capabilities and the open-source nature of the technology.