Summary in Plain Language
This report sheds light on the true situation of Zìbiànliàng Qìjī (Self-Variable Robot), the most controversial unicorn in the domestic embodied intelligence sector. Founded just two and a half years ago, this company has yet to produce any mass-produced products or generate stable sales revenue. Despite this, it has managed to attract investments from four major internet giants—Meituan, Alibaba, ByteDance, and Xiaomi—as well as over 30 other institutions, including state-owned funds and Sequoia Capital. Its total funding has exceeded 5 billion yuan, and its valuation has soared to 20 billion yuan. It has even secretly submitted an IPO application to the Hong Kong Stock Exchange. However, recent incidents have exposed the company’s predicament: its claims of outperforming a leading American humanoid robot company, FigureAI, were revealed to be a marketing gimmick. The company used a fixed robotic arm to compare with a full-size humanoid robot and its one-hour sprint result to compare with FigureAI’s 200-hour marathon average performance. This reveals that, despite being rapidly promoted by capital, its actual capabilities fall far short of its ambitious claims.
Detailed Explanation
1. How did it raise 5 billion yuan with no products? By hitting the perfect story template favored by investors
If you’re in the venture capital industry, you can’t help but be impressed by Zìbiànliàng Qìjī’s background. The founder holds a bachelor’s and master’s degree from Tsinghua University and a Ph.D. from the University of Southern California, and is one of the first researchers to develop the core component of AI models, the “attention mechanism.” This is like someone who contributed to the invention of the engine’s core patent in the automotive industry. Before returning to China, the founder also managed a quantitative fund and has both technical and business expertise. His partner is a key figure in developing multi-modal open-source AI models in China. Their combination is truly top-tier in the AI community.
The company’s chosen technological approach is at the forefront of innovation. While other robots piece together various components (vision recognition, motion control), Zìbiànliàng Qìjī claims to be creating a “end-to-end, globally unified model” that can understand the physical world and learn to perform tasks autonomously. This would be like providing all robots with an operating system that enables them to function independently—whether they’re used for sorting, household chores, or industrial applications. The potential is vast.
Additionally, Zìbiànliàng Qìjī is the only company in China to have received investments from all four major internet giants. The investment rush is driven by the fear of missing out on the next generation of general-purpose AI technologies. Over 30 institutions have pushed its valuation from several billion to 20 billion yuan, betting on its potential to become a leader in setting industry standards.
2. The “outperforming American giants” marketing stunt was a carefully designed scam
Industry bloggers explained this clearly: It’s like riding a modified electric bike on a 100-meter straight road, winning a marathon in 10 seconds, and then claiming to be 45% faster than the champion while costing 70% less. FigureAI’s sorting performance of 1,248 items per hour was achieved using a full-size humanoid robot over 200 hours of continuous operation, involving multiple robots working in sequence. Humanoid robots face much greater challenges, such as navigating the environment, finding parcels, and bending over to pick them up.
Zìbiànliàng Qìjī used a fixed robotic arm, not a humanoid robot, to perform the task. It only conducted a one-hour live demonstration and didn’t test the system’s stability over extended periods. This misleading marketing was necessary because the company lacked tangible evidence of its capabilities. Its competitors either sold tens of thousands of robots or received substantial orders, while Zìbiànliàng Qìjī had nothing to show for its investments. It relied on such sensational headlines to maintain its 20-billion yuan valuation.
3. The 5 billion yuan in funding is being spent rapidly on expensive data
Many wonder where the 5 billion yuan went, as no mass production has occurred. 90% of the funds were invested in two key areas:
- Talent and computing power: The team consists of top AI engineers from prestigious universities, with salaries at the industry’s highest level. Developing the model also required expensive GPUs.
- Data: Unlike typical AI models that can train using free online text, robot models need real-world data to learn tasks like folding clothes or cleaning tables. Zìbiànliàng Qìjī’s initial “149 yuan for 3-hour robot cleaning” service failed due to the robot’s inefficiency. Now, it’s spending money to collect data by compensating residents to allow robots to enter their homes for training, turning what could have been revenue into data costs.
4. The rush for an IPO at two and a half years indicates the pressure of a high valuation
Typical tech companies take five to six years to go from inception to IPO. Zìbiànliàng Qìjī’s haste indicates that investors’ patience is running out. Other companies in the embodied intelligence sector are showing tangible results: Yushu sells over 5,000 consumer robots annually with revenue of 1.7 billion yuan and a market value of 40 billion yuan; Zhiyuan Robot has produced its 15,000th general-purpose robot; even Zhi Pingfang, which has the same valuation as Zìbiànliàng Qìjī, has received orders worth 500 million yuan. Zìbiànliàng Qìjī’s only significant achievements are partnerships with SF Express for sorting and BMW for pilot projects, with no large-scale sales or stable revenue.
Investors initially valued it at 20 billion yuan, assuming it would become a platform that could sell models to multiple customers with almost zero marginal costs. However, its current projects require constant engineering adjustments and data collection, making it a contractor rather than a platform company. Its 20-billion yuan valuation is unsustainable.
5. Standing at the finish line, but the real test is yet to come
Zìbiànliàng Qìjī has top-tier talent, cash, and support from major companies. However, embodied intelligence requires practical applications in real-world scenarios. Its claims about a universal model need to be validated by customer purchases. Its valuation is based on future prospects. Only when its next-generation robots and subsequent sales data are released will we know if its claims were based on technical vision or a bubble. Its journey is just beginning.