The gap between ambition and execution is being tested as the humanoid robot sector transitions from speculative enthusiasm to hard-nosed financial scrutiny. On August 19, Unitree Robotics (688836.SH) made its debut on the STAR Market, becoming the first listed pure-play humanoid robot maker in A-shares. Priced at 150.8 yuan per share, the stock soared to an intraday high of 1,100 yuan, a 629.44% surge from its IPO price, pushing the company's market value past 440 billion yuan before closing at 845 yuan. However, the celebratory mood lasted just one day, with shares plunging 18.7% the following session and continuing to slide afterward.

The dramatic price swings reflect a market-wide reassessment of humanoid robot industrialization expectations. On one side, capital enthusiasm remains intense, with Unitree's issue market value hitting 61 billion yuan and other leading players crossing the 10-billion-yuan valuation mark. On the other side, practical constraints loom large, raising the question of whether robots can evolve from performing scripted shows to becoming reliable industrial workers that generate steady income. From Unitree's hardware mass production, to Zhiyuan and UBTech Robotics (09880.HK) betting on automotive factories, to Galaxy General and Astribot focusing on embodied AI models, China's humanoid robot sector is now split across divergent technical and commercial paths. But as the industry inches closer to real-world deployment, the challenges become increasingly concrete, spanning technical stability, cost reduction, data acquisition, and ultimately, whether the return on investment can be justified.

Unitree's Listing: The Post-Hype Reckoning

Unitree issued 40.45 million new shares at 150.80 yuan each, raising 6.1 billion yuan and corresponding to a market value of approximately 61 billion yuan at listing. The funds are earmarked for intelligent robot model R&D, robot body development, new product innovation, and manufacturing base construction. Financially, the company has maintained robust growth, reporting revenue of 1.15 billion yuan in the first half of 2026, a 48.54% year-on-year increase. Yet the market's initial reaction priced in far more aggressive expectations. On its first trading day, Unitree's peak valuation exceeded 440 billion yuan, more than seven times its issue value. The subsequent correction signals a shift from sentiment-driven trading to a fundamental review of the company's prospects.

Controversy quickly emerged around Unitree's R&D spending. The prospectus shows 2025 R&D expenses of 145 million yuan, representing an 8.53% R&D intensity ratio, notably lower than many tech peers. Additionally, the revenue mix raises concerns, with research and education customers contributing over 70% of humanoid robot revenue, commercial performances and showrooms around 17%, and genuine industrial applications accounting for less than 10%. This suggests Unitree has proven its ability to sell robots, but has yet to demonstrate their viability as large-scale production tools. The relatively low R&D ratio does not necessarily indicate weak technical capability, as Unitree's competitive edge lies in hardware engineering and manufacturing efficiency. Its in-house motors, reducers, and motion control algorithms, combined with a mature supply chain, enable competitive pricing and explain its early volume shipments. However, the market now focuses on the next phase. If humanoid robots remain confined to research labs, exhibitions, and tech enthusiasts, the billion-dollar valuation requires far stronger earnings support. To unlock larger markets, Unitree must prove its robots can operate in automotive, 3C electronics, and other industrial settings while delivering sustained value. The post-IPO price volatility is therefore less about one company's struggles and more about the industry's evaluation framework shifting from narrative-driven to financially rigorous.

Divergent Strategies: Players Charting Their Own Courses

Unitree's listing inevitably turns the spotlight on its competitors, with clear strategic divergence emerging across China's humanoid robot landscape. Zhiyuan Robotics leans toward a full-stack hardware and software approach with a focus on industrial deployment. Having completed over ten financing rounds backed by Hillhouse Capital, Sequoia China, Tencent Investment, and CDH Investments, Zhiyuan's primary market valuation is estimated between 15 billion and 18 billion yuan. The company has initiated Hong Kong listing preparations, with market reports suggesting a target valuation of 40 billion to 50 billion Hong Kong dollars. Notably, Zhiyuan's holding platform, Shanghai Zhiyuan Hengyue Technology Partnership, has acquired control of Shanghai Xinwei (688585.SH), adding a robotics segment to that company's operations.

UBTech Robotics has long bet on industrial scenarios. Rather than starting from robot bodies like many startups, UBTech has systematically developed its Walker series while pushing into automotive manufacturing, 3C electronics, and logistics. In partnerships with car manufacturers, the company emphasizes operational stability and task performance within actual production lines. Galaxy General takes a different route, focusing on embodied AI models with a wheeled chassis and dual-arm configuration that sidesteps the balance, load, and endurance limitations of bipedal robots, concentrating instead on visual spatial understanding and complex grasping. In early 2026, the company completed a 2.5 billion yuan financing round involving the National AI Industry Investment Fund, Sinopec, Bank of China, and SAIC, with post-investment valuation reportedly exceeding 20 billion yuan. Astribot emphasizes reinforcement learning and motion control, completing multiple financing rounds in 2026 with its post-strategic-round valuation surpassing 10 billion yuan, backed by national industrial capital, Sequoia China, IDG Capital, and other strategic investors. LimX Dynamics has also attracted significant attention, completing a pre-IPO round in July 2026 at a post-investment valuation of approximately 15 billion yuan, after raising about 200 million USD across two rounds within six months. Another notable player is Leju Robotics, which has filed for a ChiNext IPO with a pre-application valuation of around 4.3 billion yuan and plans to raise 2.6 billion yuan.

Given these diverse trajectories, the industry can no longer be compared simply by asking who builds a better robot. Unitree prioritizes hardware cost and production efficiency; Zhiyuan and UBTech focus on industrial clients; Galaxy General and Astribot wager on embodied intelligence; Fourier Intelligence extends from rehabilitation medicine into humanoid robotics; and LimX Dynamics concentrates on motion control and complex environments. Capital is placing bets across these various approaches, but the ultimate arbiter of value will be commercial outcomes.

The Ultimate Challenge: Beyond Building, Toward Profitable Operations

Automotive and 3C electronics factories represent the most closely watched application scenarios for humanoid robots today. UBTech and Zhiyuan have begun deploying robots for training within automotive production lines. However, entering a factory is not the same as replacing workers. Many current projects remain in joint R&D or proof-of-concept stages, with procurement volumes ranging from single digits to dozens of units. Tasks are typically limited to label application, quality inspection, material handling, and sorting, where error tolerance is higher, while core positions requiring high-cycle precision assembly remain largely unassigned to humanoid robots. The technical obstacles are multifaceted. The first challenge concerns dexterity. Tasks such as wire harness insertion, screw tightening, and seal installation demand sophisticated tactile feedback and fine motor control, which vision systems alone cannot reliably deliver under occlusion, reflection, and millimeter-level tolerances. The second involves reconciling AI reasoning speed with production line rhythm. Embodied large models require inference time, while industrial operations demand millisecond-level control. Many companies now adopt a layered architecture where large models handle task understanding while smaller models manage real-time control. Further down the chain, hardware durability, battery life, failure rates, and maintenance costs become decisive factors. Factories will not adopt robots simply because they appear intelligent; if equipment frequently breaks down or a single failure halts an entire production line, large-scale adoption becomes impractical.

These realities have driven some companies to make pragmatic compromises. Galaxy General employs wheeled chassis to reduce the balance and energy demands of bipedal locomotion, while UBTech and Zhiyuan integrate quick-change tooling to combine mature industrial tools with robots. Technically, these approaches solve certain problems, but commercially, they introduce a new dilemma. When wheeled bases, specialized fixtures, and remote intervention systems become prevalent, what remains of the humanoid robot's advantage over traditional robotic arms and autonomous mobile robots? This is the core question for industrialization. Manufacturing customers ultimately compare total cost of ownership rather than how human-like a robot appears. Once equipment price, depreciation, maintenance, computing power, system integration, and on-site servicing are fully accounted for, if robots remain more expensive than human labor or conventional automation, even the most advanced technology will struggle to secure large-scale orders.

The humanoid robot industry currently finds itself in a phase where technology advances rapidly, yet commercial viability remains unproven. The capital markets have already assigned high valuations, but the true inflection point depends on robots demonstrating three things in real-world settings: sustained long-term operational stability, reduced human intervention in complex environments, and cost recovery within a reasonable period for enterprise buyers. Unitree's listing marks the first large-scale exposure of this sector to secondary market scrutiny. For the entire industry, the real competition may only be beginning, as the focus shifts from who can manufacture robots to who can make them genuinely create value.

How do you view the development of the humanoid robot industry? We welcome your comments and discussions.