Leikeji Embodied Intelligence Group | Editor: Chongjia | Supervisor: Luo Chao
In recent years, humanoid robots have been frequently seen at various occasions, especially technology exhibitions. They are often the focal points of these events, attracting countless spectators to watch and take photos.
However, although we have always believed that it is a general trend for humanoid robots to enter households, their retail prices remain high for now. There is still a long way to go before they truly reach ordinary families. Against this backdrop, "desktop robots," characterized by their compact size and focus on office or study desk scenarios, have gradually come into view, capturing the attention of some users.
Naturally, opinions on desktop robots are polarized. Proponents view them as entry-level carriers for large models to land in the physical world; critics consider them merely expensive electronic figures disguised with AI, destined to gather dust once the novelty wears off.
So, will desktop robots become a hot track in the field of embodied intelligence?
Desktop Scenarios Become a New Sub-track for Embodied Intelligence Vendors
Before diving into this topic, it is necessary to briefly define what constitutes a desktop robot. At its most basic level, we believe a desktop robot must meet the criteria of embodied intelligence + desktop scenario. Additionally, it needs to possess active AI perception capabilities, effectively having a "brain" powered either by local chip computing resources or the cloud. Furthermore, as a "robot," it must have mechanical structures. Typically, we do not classify AI speakers or AI cameras as desktop robots.
Compared to full-sized humanoid robots, desktop robots have lower manufacturing costs and narrower capability ranges, which is clearly an advantage for startups.
As mentioned earlier, a major pain point for humanoid robots today is price. The high cost of full-sized robots ultimately stems from costs that have not yet been fully reduced. Normally, a humanoid robot contains high-performance chips to support local computing power; simultaneously, it requires a brain composed of large models + agents, testing algorithms and training data. Moreover, dexterous hands and mechanical structures represent unavoidable costs.
In contrast, commercializing robots in desktop scenarios is much easier. The physical boundaries of a desk are very clear, and the environment is relatively simple, limited to the user's desktop area and space. This means there is no need to solve complex bipedal balance issues or perform global navigation and obstacle avoidance; essentially, only edge fall prevention needs to be addressed. Consequently, R&D teams can focus limited computing power, batteries, and other hardware resources entirely on intelligence and interaction.
Moreover, from a supply chain perspective, desktop robots also enjoy cost advantages. Their main control SoCs, display screens, microphones, camera modules, batteries, etc., can almost directly reuse mature solutions accumulated over the past decade-plus in the smartphone and IoT industries. As a result, component procurement costs are significantly lower than those for full-sized robots.
Recently, we introduced a low-cost bipedal robot solution developed by the University of California, Los Angeles (UCLA). Specifically, it drives a small robot using second-hand smartphones, keeping the body cost under CNY 3,000. Its ability to compress costs so drastically lies in fully utilizing the various components inside phones: cameras for visual recognition, SoCs for inference computation, WiFi for connectivity, and batteries for endurance. Robots of this size are well-suited for desktop applications. While conventional desktop robots won't necessarily use phones as drivers, many components can indeed reuse the smartphone supply chain.
Another practical reason many startups enter the desktop robot sector is that full-sized robots represent the industry's technical high ground, crowded with heavyweights in intense competition, with most still burning cash. In comparison, the technical barriers for desktop robots are relatively lower, trial-and-error costs are less prohibitive, capital requirements are smaller, and crucially, they allow vendors to quickly launch products directly sellable to C-end consumers, generating continuous cash flow.
Desktop Robots Focus on Emotional Companionship and Efficiency Tools
Turning our attention back to the products themselves, we find that current desktop robots primarily offer two types of functions: emotional companionship and efficiency enhancement.
First is emotional value and companionship, an attribute emphasized by many robots nowadays. In our view, the emotional companionship provided by desktop robots is mainly manifested through "desktop pet" forms that can move, speak, and interact.
For example, the popular Letianpai desktop robot is a small bipedal robot whose "head" is primarily a circular screen displaying bouncing "eye" patterns, positioning itself as a cute pet.
The Letianpai desktop robot offers rich interaction methods. Besides touch and voice control, it can recognize user gesture commands via its camera. In terms of movement, it can walk forward and backward, as well as move side-to-side. Frankly speaking, it doesn't have any particularly groundbreaking features; its core selling point is emotional companionship, capable of chatting and responding with actions. It runs on Android, supports flashing Linux, and allows customizing large models, offering high playability for advanced users.
Lingtong NIA-F01 humanoid robot takes emotional companionship further. It is essentially a figure robot, with an appearance completely resembling a figurine. However, it incorporates cameras, micro-skeletons, and other sensors. Unlike traditional figurines, it can perform different actions based on commands and engage in voice interactions with users. Users can also create and edit personas, voices, etc., for it. In short, it is a highly customizable figurine, mastering the art of emotional companionship.
Second is efficiency tools. A representative example is the myCobot series launched by Elephant Robotics, which resembles a small robotic arm. The myCobot is not a plug-and-play consumer product but requires secondary development. For instance, it can be equipped with grippers or suction pumps to become a tidying robot for desks; it can also interface with mice and keyboards to simulate human computer operations, helping users complete repetitive tasks.
The proliferation of desktop robots is largely supported by mature underlying technologies for corresponding scenarios. Now, multimodal large models enable them to understand complex contexts and natural language. Coupled with the dissemination of machine vision technology, robots can use front-facing lenses to capture and understand user faces and gestures. Regarding motion, major vendors' desktop robots generally use micro-servo motor solutions, allowing the robot bodies to perform anthropomorphic actions like shrugging, shaking heads, and stomping feet.
Looking at domestic and international markets, many brands have entered this field. Domestically, Letianpai is a brand that has moved quickly with clear product definitions. They deeply integrated large models into a compact body, enabling chat and expressions, entering the market at a price of CNY 1,799. Other brands like Lingqi Technology and Kechu Future are also focusing on different forms. Some major humanoid robot manufacturers, such as Unitree, Lingtong, and Xuanji, have also joined in producing such products, including the aforementioned figure-style desktop companion robot NIA-F01.
In overseas markets, crowdfunding platforms are the main battleground for such products. For example, the previously popular Looi Robot featured innovative product design, using the user's smartphone as the robot's brain and face, with the base responsible only for power and support, keeping the price down to USD 129 and achieving high crowdfunding amounts. Additionally, the AI open-source community Hugging Face launched the Reachy Mini desktop robot, primarily targeting developers.
Overall, currently more domestic brands are entering the desktop robot field, while overseas brands are still in a nascent stage.
Is the Endgame for Desktop Robots Physical Agents?
To be honest, the main selling points of desktop robots currently available to ordinary consumers are still concentrated on emotional companionship. There are plenty of desktop robot solutions focusing on efficiency tools and practical functions, but many are aimed at commercial use, often providing just a robotic arm solution requiring secondary development. While highly customizable, apart from industry users, basically only geek enthusiasts can master them.
Emotional companionship is certainly an important track for the robotics industry, but desktop scenarios inherently have limitations compared to full-scenario environments. If emotional companionship is the sole major selling point, desktop robots will remain cyber-pets flying the banner of embodied intelligence. After all, for ordinary users, it is just a smart speaker with a screen that can take a few steps. Moreover, even entry-level desktop robots often cost over CNY 1,000, making them more expensive than smart speakers.
In our view, for desktop robots to truly become a track with massive incremental growth, their product forms and core functions must undergo a critical upgrade.
In terms of product form, desktop robots should move towards modularity. Current products still resemble abrupt electronic devices too much. Unless unexpected changes occur, future desktop robots will deeply integrate with users' monitor stands, desk lamps, or even wireless charging pads. When needed, they appear promptly as assistants; when not needed, they quietly hide away.
Furthermore, modular design will give them greater plasticity. Users might assemble them like computers, attaching a robotic arm module for organizing small items, or swapping in a tracked chassis or bipedal legs for moving across the desk.
In terms of functional evolution, desktop robots must significantly upgrade their capabilities in the physical world. Although current robots look impressive in demonstrations, many run according to pre-set programs, essentially acting as script tools. We believe the true value of desktop robots to users must land on performing actual work. At least personally, an ideal desktop robot could operate computers, organize items, adjust lighting, etc., for me—evolving into a "Physical Agent" capable of doing things beyond AI assistants on phones/computers and smart speakers.
All in all, although current desktop robots ride the wave of embodied intelligence hype, they are still in the early stages of industry development. Consumer-grade products focus mainly on emotional companionship and have not yet truly unleashed productivity-enhancing capabilities.
In our opinion, desktop robots shouldn't just be expensive movable figurines or smart speakers with eyes. Since they occupy precious real estate on the user's desktop, they should provide commensurate efficiency value.
Unless unexpected changes occur, as the multimodal capabilities of large models descend and smartphone supply chain components continue to be reused, the desktop robot track will see more new products in the coming years. However, to make ordinary users willing to pay for them, there is still a long road ahead.