中国政策制定者对人形机器人行业泼冷水

中国的政策制定者对人形机器人行业泼冷水，指出该行业存在过热现象，需谨慎对待，决策者警告称，尽管人形机器人在技术和应用方面取得进展，但仍面临诸多挑战和不确定性，需要更多的研发和验证，决策者还强调了行业发展的可持续性以及对社会和经济的影响等问题，此举提醒业界要冷静思考，避免盲目跟风投资，确保行业的健康发展。

By Ma Jinnan

On November 27, during the National Development and Reform Commission's news briefing, when a relevant official stated the need to "actively prevent the simultaneous listing of highly repetitive products," the fervor of capital surrounding the humanoid robot industry seemed to be abruptly doused with a bucket of cold water.

This "cold shower" came suddenly, yet was not entirely unexpected—just half a month earlier, industry leader Unitree Robotics had just completed its IPO counseling filing, preparing for a market debut. At the same time, the entire industry was surging at an annual growth rate of over 50%. Market research institutions optimistically predict the formation of a hundred-billion-yuan market by 2030.

At this very moment, a dramatic contrast between regulatory caution and industry frenzy—two extremes of “hot and cold”—came vividly into play.

The humanoid robot sector is experiencing an unprecedented "big bang." According to public data, the number of registered humanoid robot companies in China surpassed 105 in the first half of 2025 alone, not only exceeding the total for all of 2024 but also marking an explosive increase of 183.78% year-on-year.

Even more astonishing is that there are currently over 150 humanoid robot companies nationwide, with more than half being startups or entrants from other industries. From internet giants to traditional manufacturers, from research institutions to entrepreneurial teams, capital of all kinds is rushing in, pushing humanoid robots to the very forefront of the wave.

In this feast of capital, the IPO progress of companies like Unitree Robotics has become a microcosm of the industry's fever pitch. On November 15, this leading company—which started with quadruped robots and has in recent years made a strong push into the humanoid domain—just completed its IPO counseling. Almost simultaneously, another well-known player, Leju Robotics, also filed for IPO counseling.

The enthusiasm of the capital markets seems to proclaim that the "golden era" of the humanoid robot industry has arrived, while regulatory “cooling” measures and industry “boiling point” continue to collide intensely.

The relevant official from the NDRC made it clear at the press conference that: the current humanoid robot industry’s technical approaches, commercialization models, and application scenarios are still not fully mature. As emerging capital accelerates its entry into the sector, it is necessary to take precautions against risks such as the flood of highly redundant products hitting the market and the compression of R&D space.

This statement sends a clear signal: at the policy level, concerns about unregulated expansion within the industry are intensifying.

It’s worth noting that the NDRC official specifically emphasized the balance between “speed” and “bubbles.” This is not simply about “cooling down” the industry, but rather about deep reflection on how to foster healthy growth. When more than 150 companies crowd into a track where the technology remains unproven and business models are still being explored, risks such as homogenized competition, resource waste, and valuation bubbles are quickly accumulating.

The intention behind this policy “chilling effect” is, in fact, to inject a dose of rationality into the current craze.

Policymakers themselves are caught in a dilemma. On one hand, humanoid robots, as the core carriers of embodied intelligence, have already been included in the national “15th Five-Year Plan” and identified as a critical direction for future industrial innovation—a key arena in the global technology race. On the other hand, overheating of the sector could lead to resource misallocation and bubble risks, so timely warnings and guidance are necessary.

This conflict is particularly evident in companies like “Unitree.” As industry leaders, they possess technology accumulation and market recognition, and thus deserve policy support. However, if a large number of enterprises lacking core competitiveness blindly jump on the bandwagon, investors’ interests may be harmed and genuine innovators could see their space squeezed out.

The deeper rationale behind this policy “cold water” is about mastering the art of balancing “strategic support” with “bubble prevention.”

As the industry’s “boiling point” meets policy “cold water,” the humanoid robot sector stands at a crossroads. Will this “cold water” douse the flames of innovation, or guide the industry onto a healthier development track? How should “companies like Unitree” find their rhythm amid both shifting policy and market dynamics?

The current technological landscape of the humanoid robot industry strongly resembles the "Warring States period," where a hundred schools of thought compete and consensus is far from reached. Presently, mainstream technological approaches include bionic design based on traditional mechanical structures, intelligent control systems powered by AI algorithms, and multi-sensor fusion for environmental perception. Each of these directions has its pros and cons yet none are fully mature, resulting in dispersed R&D investment by companies and a lack of collaborative momentum.

Despite the multitude of approaches, core technological bottlenecks remain prominent. There are still critical challenges with high-precision servo motors, lightweight materials, and motion control algorithms for complex environments—issues that effectively "choke the neck" of development. With over 150 companies entering the field and each one independently exploring technical routes, not only is there a waste of R&D resources, but it also slows down breakthroughs in key technologies. The warning from the NDRC about R&D space being "squeezed" most likely refers to this very phenomenon: when technological directions are highly "fragmented."

Although the market widely predicts that the humanoid robot industry will reach a hundred-billion-yuan scale by 2030, the path to commercialization remains arduous. In industrial settings, high costs and poor adaptability persist; in home scenarios, limited by safety concerns and subpar interactive experiences; and to enter specialized fields such as healthcare and education would require an even longer period of validation.

In short, the limitations of application scenarios are just too pronounced.

Given such significant constraints on practical applications, the business models inevitably remain very vague as well.

Capital markets and investors should be keenly aware that most companies are still in the "burning cash for R&D" phase, lacking clear commercialization pathways. Industry surveys show that the average cost of a humanoid robot product is as high as 500,000–1,000,000 yuan, while market acceptance generally sits below 200,000 yuan. This kind of "cost inversion" forces companies to rely on continuous financing just to keep operations running.

When the capital fever fades, those companies lacking sustainable revenue generation will face a survival crisis. This is the rationale behind the NDRC's judgment that the "commercialization model is not yet fully mature," and it remains the Achilles’ heel for sustainable industry development.

So, what is the NDRC really warning about when it says there’s an “excessive clustering of homogeneous products entering the market”?

In this industry, more than half of the companies are either startups or entrants from other sectors. Many of these firms lack technological foundations and have rushed into the field riding on hype and capital alone, resulting in a significant issue of product homogeneity.

The market is now flooded with mid- to low-end products that share similar functions and performance, while truly innovative high-end products are few and far between.

This “hundreds of players crossing a single-log bridge” phenomenon inevitably leads to a threefold waste of resources. First, there’s redundant investment in R&D, with many companies crowding into similar technological paths; second, there’s cutthroat competition for talent, with core engineers’ salaries driven to unreasonable heights; third, market resources are spread too thin, making it hard to achieve economies of scale.

The NDRC’s “dousing with cold water” is not an isolated case; this has already become an inevitable pattern in the development of emerging industries in China.

Looking back at history, both the “subsidy fraud wave” in the new energy vehicle industry and the “inflated bubble” in AI chips have passed through similar phases of “barbaric growth.”

A decade ago, the new energy vehicle industry experienced an explosive boom under policy incentives, attracting a massive influx of enterprises, which led to severe overcapacity and technological stagnation. In the end, the industry had to go through a shakeout via “subsidy cuts” and “raising the entry threshold.” Similarly, during the chip frenzy from 2018 to 2020, hundreds of companies raced to develop mid- to low-end chips, resulting in severe product homogeneity, and most companies were eventually weeded out by the market.

History shows that emerging industries often go through a stage of “wild, chaotic growth” in their early development, which is inevitably followed by a painful restructuring and “survival of the fittest.” The warning from the NDRC aims to help the humanoid robot industry avoid repeating the same mistakes by guiding the sector to shift from “quantity expansion” to “quality improvement” ahead of time.

As policy “cools down” an industry at its “boiling point,” leading companies represented by Unitree Robotics now find themselves at a pivotal crossroads.

These industry pioneers, holding strong technological foundations and already in the process of going public, face the dual challenge of tightening regulations and the opportunity of industry reshuffling. They might break through amid this two-pronged interplay of regulation and market forces.

In the face of potentially stricter listing reviews, the top priority for companies like Unitree is to accelerate the disclosure of core technologies and the deployment of patents. Take Unitree Robotics as an example: their Unitree H1 humanoid robot is equipped with self-developed, high-performance joint motors, for which multiple core patents have already been filed. Such technological breakthroughs can no longer remain as mere “black tech” locked away in laboratories; they must be transformed into “tangible assets” that appear in prospectuses.

While racing to accelerate, companies like Unitree need to focus on building their own moats with patents.

As for IPO strategies, the old approach of “storytelling” through conceptual packaging no longer works; what’s needed now is “technical validation.” In their prospectuses, companies must clearly demonstrate the maturity of their technology roadmap, the rate of localization of key components, and the makeup and capabilities of their R&D teams. This is particularly important for those that have historically depended on imported critical components—they must show real progress in overcoming proprietary “bottleneck” technologies.

The true breakout strategy is not to constantly try to counteract policy, but rather to go beyond it.

According to a relevant official from the National Development and Reform Commission, the “commercialization model is not yet fully mature,” meaning that listed companies must present a clear commercialization path and profit expectations. For example, during the IPO counseling process, Unitree Robotics has begun systematically organizing its case studies for applications like industrial inspection and education & research, and has quantified its market penetration rate and revenue forecasts for the next three years.

Three types of validation are indispensable: technology validation (are product performance metrics up to par?), scenario validation (what’s the client feedback from trial runs?), and commercial validation (can a replicable profit model be formed?). Companies that are still stuck in grand narratives about “future markets worth hundreds of billions,” yet can’t explain “where next year’s revenue will come from,” will inevitably face much more stringent review and questioning.

When it comes to technology paths, emphasis should be placed on “differentiated competition”—we cannot simply aim to be “the next Tesla”.

In a track crowded with more than 150 companies, technological differentiation could be the key to survival. Unitree Robotics has chosen to avoid direct competition with Tesla’s Optimus, instead focusing on the niche of high-dynamic motion control. Its H1 robot has already achieved difficult feats such as backflips. This is a case of “leveraging your strengths against others’ weaknesses”—a strategy that leading players ought to adopt. However, it’s also important to note that “backflips” must not remain mere spectacle or showmanship; the unique technical capabilities behind these skills must be applied to suitable practical scenarios in order to close the loop and deliver real value.

As a company, it's essential to consider: what are the fundamental differences between our technical approach and mainstream solutions? Which unique pain points does this difference address? Can it build a sustainable technological moat? For example, some companies focus on flexible bionic structures, others are dedicated to multimodal interaction systems, while some tackle extreme environment adaptability—such differentiated positioning is what helps avoid the red ocean of homogeneous competition.

In the face of cooling policy environments, the “hot survival strategies” of companies like Unitree are essentially about finding a dynamic balance between regulatory constraints and market opportunities.

On one hand, it’s vital to align with policy directions, strengthen research and development of core technologies, and avoid homogeneous competition. On the other hand, companies must seize market windows, accelerate commercialization, and secure first-mover advantages.

Given the pain points of underdeveloped application scenarios, leading enterprises need to shift from the grand narrative of “general-purpose humanoid robots” to the pragmatic path of “specialized solutions”. Some companies have already established complete product lines in the education sector, while others are focusing on medical rehabilitation scenarios, and so on…

When selecting scenarios, it’s vital to follow the “three-high principle”: high demand (clear customer pain points), high value (strong willingness to pay), and high barriers (significant technical thresholds). For example, areas like industrial inspection, specialized operations, and high-end education should be strategic priorities for top-tier companies, who should target application scenarios with “surgical precision”, extricating themselves from the illusion of being “general” in time and strategically withdrawing to “specialized” sectors—steering clear of the red ocean and opening up new blue ocean opportunities.

Once a company’s technological strength is robust enough, its business model clear, and its ecosystem layout comprehensive, stricter regulations may even serve as a “filter” to weed out competitors.

The so-called "pouring cold water" by the NDRC reflects a profound shift in industrial policy thinking. The approach has moved from early-stage “broad encouragement”—akin to “releasing water to raise fish”—to a more targeted “precision drip irrigation” model of focused support. This shift is an inevitable policy response to the true developmental rules and characteristics of the industry.

At its core, this change is about building a differentiated policy system of "selective support and selective restriction." For companies with genuine core technologies and clear business models, the country will certainly provide substantial support. As for those enterprises marked by severe homogeneity and lacking competitiveness, the aim is to encourage natural elimination via market mechanisms. The proposals from relevant authorities at the National Development and Reform Commission—such as “accelerating the establishment of industry standards and evaluation systems” and “improving embodied intelligence entry and exit mechanisms”—are concrete examples of this policy direction.

Through risk warnings, expectation guidance, and better institutional frameworks, the policy is designed to help the humanoid robotics industry achieve a qualitative transformation—from “barbaric growth” to “refined cultivation.” Although this transition may dampen some speculative fervor in the short term, in the long run, it will undoubtedly benefit the industry’s healthy and sustainable development.

In this breakthrough battle, every step taken by companies like "Unitree" is crucial to the future of the industry. Their success will not only mark the triumph of individual companies, but also serve as a symbolic turning point for China’s embodied intelligence industry, moving from “barbaric growth” to “refined cultivation.” When the policy’s “cold water” meets the enterprises’ “fervent drive for survival,” the humanoid robotics industry will face its true "tempering into steel" moment.

It’s worth believing that the policy’s “cold water” is not about “sharing the benefits equally,” nor is this “cooling-off” a sign of an impending winter. Companies that value technological innovation, operate with sound business acumen, possess an industrial mindset, and have real-world application scenarios—those with a “starry sky vision” who remain “down-to-earth”—will, in fact, welcome the arrival of a “springtime.” (Author | Ma Jinnan, first published on TMTPost App)