Prominent investor Zhu Xiaohu has once again stirred debate by casting a critical eye on the robotics sector. During a two-hour lecture titled "Venture Capital in the AI Cycle" at Peking University's School of Innovation and Entrepreneurship, the managing partner of GSR Ventures delivered a stark message: among China's roughly 200 robotics companies, many can dance and perform martial arts, but few have built real businesses, and most will likely fail to survive a decade.
This blunt forecast has reignited controversy within the robotics community. The criticism appears aimed at companies that possess only flashy demonstration capabilities without achieving a commercial closed loop. Just months earlier, Zhu made headlines for "exit from humanoid robot companies in batches." At that time, he was even more direct: "I asked these CEOs, who are your potential commercialization customers? It felt like they were describing customers they had imagined themselves. Who would spend over one hundred thousand yuan to buy a robot to do these tasks?"
Such sharp remarks struck a nerve across the industry. Zhao Tongyang, CEO of UBTECH-backed Qinglong Robot, responded on social media: "It's like questioning a newborn baby and denying the future based on what you see now. That kind of talk is not what someone of his stature should be doing." This time at Peking University, Zhu's tone remained firm but his reasoning was more complete, and in the eyes of many in the investment community, it carried greater professional and financial logic, making it more persuasive.
Core Industry Pain Points
Zhu's statements directly exposed weaknesses in the sector. Based on a compilation of his on-site lecture by 36Kr, his judgment rests on several practical issues. As technology diffusion accelerates, the technology itself rapidly depreciates. Startups cannot survive merely by claiming "I also have a model." The robotics industry is transitioning from "can it be built" to "is there sustained customer demand." Motion demonstrations cannot prove unit costs, failure rates, maintenance expenses, or payback periods. Startups should instead look for tasks that are "just within reach"—such as underwater cleaning, warehouse sorting, or flexible robotic arms—where task boundaries are clear and customer needs are easy to quantify.
Once technology spreads, customers can also become competitors. Creating prototypes for large companies quickly erodes first-mover advantages. Ultimately, investors need an exit strategy, not just a good story. One must stand firmly on the grand direction of embodied intelligence while also identifying scenarios that can deliver value today. The "not surviving a decade" remark is not a statistical report but a venture capitalist's risk warning. The real test for robot companies is not on the product launch stage, but whether customers renew orders after six consecutive months of factory operation.
Companies in the GSR Ventures Portfolio
GSR Ventures was established in 2004 by founder Zhu Xiaohu. The firm's official website discloses managed capital of over $3 billion, with a long-term focus on enterprise services, consumer internet, healthcare technology, and artificial intelligence. Investments in Didi, Ele.me, Qunar, Xiaohongshu, and Horizon Robotics form the most recognizable labels of its early-stage investment track record. Consumer internet investing looks at user growth, retention, and network effects, while AI applications require attention to usage frequency, payment rates, and customer switching costs. For robotics, one must also monitor task success rates, equipment utilization, and after-sales expenses. Technology is merely the entry ticket; whether customers continue using the product over the long term determines the project's value.
Over the past few years, GSR Ventures has deployed capital across multiple embodied intelligence and robotics companies, with public reports mentioning Stardust AI, Songyan Dynamics, Flexiv Robotics, Xingjue Robotics, and Nimble Robotics. In March 2025, Zhu stated he was exiting a portion of early-stage projects. Stardust AI subsequently confirmed that GSR participated in its angel round in November 2023 and completed the exit in September 2024. However, exiting early humanoid robot projects does not mean GSR is abandoning robotics altogether. In May 2025, Zhu told Securities Times that he favored industrial robots and those that can "actually do work." Soon after, GSR joined the financing round for underwater robotics firm Shihang Intelligent. Shihang initially focused on underwater ship hull cleaning, then expanded into customs inspection, offshore solar panel cleaning, and aquaculture. The demand behind these scenarios is consistent: areas where humans cannot see clearly underwater, cannot easily descend, or face safety risks. The investment logic is straightforward: first solve a specific problem that someone is willing to pay for, then replicate capabilities to adjacent scenarios.
As for GSR's overall returns on AI and robotics investments, public information does not disclose a unified earnings statement, nor verifiable aggregate internal rate of return or return multiples. What can be confirmed are specific projects, financing rounds, and exits, but financing valuations should not be equated with investment returns. From an investment logic perspective, consumer internet projects demonstrated GSR's ability to identify platform-type companies. Robotics and AI remain in the portfolio selection phase, where exits, track switching, and continued betting exist simultaneously— consistent with the real state of early-stage investing.
Robotics Actually Represents Three Distinct Businesses
Many discussions lump industrial robots, humanoid robots, and underwater robots into a single chart, ultimately deriving one enormous market figure. When the concept gets inflated, judgment often becomes muddled. Industrial automation is already a mature business. According to data from the International Federation of Robotics (IFR), global industrial robot installations in 2024 reached approximately 542,000 units, with China installing about 295,000 units, accounting for 54% of the global total. China's operational stock of industrial robots exceeds 2 million units, and domestic manufacturers have increased their share of the local market to 57%. This segment has real customers, production line demands, and delivery records, with competition moving from growth markets into the deeper waters of cost, service, and domestic substitution.
Humanoid robots and embodied intelligence remain in the validation phase. Reuters cited IFR statistics indicating that global sales of humanoid robots for industrial and professional services in 2025 will total approximately 7,000 units. Compared to the 542,000 industrial robot installations, this figure remains very small, with many units sold to research institutions and enterprises for AI model training, while pilot deployments in automotive factories number in the single to low double digits. There is also a category of specialized and service robots: underwater robots, warehouse sorting, inspection, and cleaning often attract less attention, yet they are more likely to find clear paying customers first. The industry's shortcomings are also concentrated: reliability, generalization capability, full life-cycle costs, and data closed loops—none can be avoided. The China Electronics Society's 2026 policy interpretation has listed task success rate, efficiency improvement rate, safety reliability, and economic feasibility as key indicators, signaling a shift from showcasing prototypes to real job performance assessments.
Financial Reports Offer a Sobering Reminder
When examining companies that have publicly disclosed financial data, the true state of the industry becomes clearer. Revenue growth in robotics does not equate to high-profit operations. The more complex the delivery, the heavier the burden of R&D, production, sales channels, and after-sales costs. Inovance Technology has a broader business scope, so it cannot simply be understood that humanoid robots are already profitable. Estun Automation turned profitable in 2025, but margins remain thin. Siasun's revenue scale is considerable yet it still posts losses. UBTech's full-size embodied intelligence revenue is growing rapidly, but losses persist. These figures should not be used to simply rank companies, but they illustrate the industry's current stage: industrial automation has entered an era of operational efficiency competition, while humanoid robots are still moving from product delivery toward scaled validation.
How the Industry Will Reshuffle
In the next phase, demonstration-oriented products will decline, and scenario-specific products will increase. Being able to complete one job position holds more commercial value than performing ten movements on a stage. Competition among complete machines will also evolve into system-level competition. Sensors, actuators, controllers, data collection, and after-sales service—if any link fails, stable delivery becomes difficult. Sales models may shift from one-time equipment purchases to leasing, pay-per-task arrangements, and robotics-as-a-service, thereby lowering customers' initial investment. The industry will also see clear divergence: some companies will be absorbed by large corporations, some will become component and system suppliers, and only a few may grow into platform-type enterprises.
Policies are also pushing the industry toward real-world scenarios. The Ministry of Industry and Information Technology has proposed promoting applications in typical manufacturing scenarios and key livelihood industries, while the 2026 Real-Scene Special Training Action Plan further emphasizes real environments, real machine data, and batch deployment. The industry's main trajectory is increasingly clear: moving from "does it look human" to "can it reliably complete work."
Final Analysis
Zhu Xiaohu's latest judgment lacks romanticism but stays close to the essence of investing. The robotics industry certainly has a future. China possesses manufacturing scenarios, a complete supply chain, and a sufficient number of enterprises willing to experiment with automation. Industrial robots have already proven that machine substitution for labor is an industry that can last for decades. A future for robotics does not mean every robot company will reach that future. Zhu's words are harsh, but what he punctures is not just the robotics bubble—it is a habitual mindset throughout tech investment: the assumption that if technology is advanced enough, the market will automatically pay for it.
Looking back at solar, electric vehicles, and shared bikes, every trend has experienced a cycle from frenzy to consolidation. As a Securities Times commentary aptly stated: "Rather than arguing about whether the humanoid robot track has a bubble, it is better to discuss what stage the bubble has reached. Like fine beer, if you want to enjoy it, foam is an unavoidable part of the experience. The only thing to care about is whether the glass contains more beer or more foam." The capital market's most common mistake is to treat an industry's long-term potential as a company's current revenue prematurely. The truly difficult part does not happen in front of the camera but in the customer's workshop: how many hours can it work each day, who repairs it when it breaks, how much efficiency does it improve, and will customers continue buying three years from now.
The robotics industry is undeniably the future. China's advantages in supply chain, manufacturing capability, and market scale are tangible. In the first half of 2026, Chinese suppliers accounted for over 97% of global humanoid robot shipments—a number that itself reflects strength. But strength is one thing; turning technology into a viable business is entirely another. Zhu Xiaohu's "decade theory" is not a verdict, but it has pulled back the curtain on many companies in the industry. What it exposes is this: on a track where everyone says "yes," who dares to pause and ask, "Where does the money come from?" Among the two hundred plus companies, some will survive beyond ten years. The question is: who, and on what basis?
What are your thoughts on Zhu Xiaohu's criticism of the robotics industry? We welcome your rational and civil views in the comment section.
Disclaimer: This article is a financial hotspot analysis only. Data and materials are drawn from public sources, company announcements, and Tonghuashun IFinD. The views expressed are for reference only and do not constitute investment or consumption advice.