As the "first humanoid robot stock on A-shares," Yushu Technology Co.,Ltd. (688836.SH) has become one of the most closely watched capital market events in August 2026. From a seed round valued at just 10 million yuan in 2016 to a valuation that has surged over 4,000 times in a decade, the company's listing on the STAR Market has drawn nearly ten million investors to participate in its new share subscription.

Behind the fanfare lies a significant valuation debate triggered by the 61 billion yuan issuance market value and a 219-fold issuance price-to-earnings ratio. On one side stands a decade of multi-thousand-fold valuation growth backed by major internet giants, government industrial funds, and renowned VC/PE firms. On the other side are the constraints of reality: revenue heavily dependent on scientific research scenarios, notable gaps in embodied large models, and a marked slowdown in performance growth since 2026. The listing of Yushu Technology is not merely a capital market milestone for a humanoid hardware manufacturer—it has fully exposed the deep-seated competition over valuation pricing, commercialization, and technology route selection across the entire embodied intelligence sector.

"The STAR Market listing of Yushu Technology marks a profound shift in A-share hard-tech assets from 'digital intelligence' to 'physical intelligence'," said Tian Lihui, a finance professor at Nankai University. He noted that Yushu Technology represents the "embodied" form of AI interacting with the physical world, filling a critical gap in "physical AI" assets on the A-share market. In fact, Yushu Technology's IPO has set a record for the fastest approval on the STAR Market—just 73 days from acceptance on March 20 to approval on June 1. A person close to regulators noted that the robot industry is a vital engine driving industrial transformation and economic growth, as well as a strategic high ground in global frontier technology competition.

Kang Yu, general manager of the capital market department at Shoucheng Holdings, which manages the Beijing Robot Industry Development Investment Fund—the state-owned single largest shareholder of Yushu Technology—views the A-share listing as a typical signal of long-term policy support for the physical AI sector. Ding Jian, chief researcher at Debon Securities' high-end manufacturing research center, added that embodied intelligence is a highly technology-intensive industry requiring enormous funding for large model training, real-world data collection, hardware iteration, and talent acquisition, making IPO fundraising an essential financial backbone for continued industrial development.

While the embodied intelligence track offers vast imaginative potential, achieving large-scale commercialization in practice remains difficult. As Yushu Technology's listing approached, discussions and controversies have never ceased: Can embodied intelligence deliver on investors' high expectations? When will robots that currently can only perform flips achieve large-scale commercial deployment? How should valuation systems be established for embodied intelligence companies when valuations run ahead of earnings? How can universal embodied intelligence technical bottlenecks be overcome? Industry consensus holds that this track will eventually produce trillion-yuan-scale leaders. Many companies currently command valuations in the tens of billions, but if they ultimately become industry leaders, today's valuations may not be expensive. The core challenge lies in predicting who will emerge as the final winner—all valuation judgments must be grounded in deep analysis of fundamental strength and growth potential.

Behind the scenes, Yushu Technology has attracted 46 shareholders, including major internet companies, market-oriented VC/PE institutions, local state capital, and government guidance funds. Founder Wang Xingxing remains the controlling shareholder, directly holding 21.44% of post-listing shares and controlling 68.78% of voting rights through differential voting structures, while serving as chairman, general manager, and chief technology officer. The collective presence of internet giants is particularly striking: Meituan's affiliated entities hold approximately 8.68% of post-listing shares, worth about 5.297 billion yuan at the 61 billion yuan valuation. Wang Xing, Meituan's founder, has stated that robots are a key vertical for Meituan's investments, potentially representing the ultimate solution for last-mile food delivery. Xiaomi's Lei Jun through Shunwei Capital, Alibaba's affiliates, Tencent's investment platform, and ByteDance's investment vehicle all hold positions ranging from 2.18 million to 16.11 million shares.

"Internet giants' cloud services, large models, and digital ecosystems remain confined to the virtual world, while embodied intelligence serves as a critical carrier for AI's physical deployment—potentially the next-generation universal interactive terminal after smartphones," one industry insider commented. Market-oriented investors include Sequoia China, holding approximately 6.4% of post-listing shares worth about 3.906 billion yuan; Temasek through its Vertex and Xiangfeng vehicles; and other notable institutions such as Matrix Partners China, Source Code Capital, Shenzhen Capital Group, and Lightspeed China Partners. Zheng Juncong, founding managing partner of Vertex Ventures, noted he invested in Yushu Technology three times, praising Wang Xingxing as a technology-focused founder he greatly admires.

Local governments have also thrown their weight behind the sector. The Beijing Robot Industry Development Investment Fund holds 3.44% of post-listing shares, worth approximately 2.101 billion yuan, making it the largest single state-owned shareholder. "Our fund was established to help Beijing build a humanoid robot industrial cluster, supporting local hard-tech companies through capital," Kang Yu explained. "We take a long-term patient capital approach, accompanying enterprises throughout their growth while leveraging our group's offline industrial resources." Other state-backed investors include the China Internet Investment Fund, China Mobile's industry fund, Beijing's Zhongguancun Science City, Shanghai's science and technology innovation fund, and Xiamen's Junshi Venture Capital.

Notably, at the IPO stage, a significant new shareholder emerged: Liang Wenfeng, founder of DeepSeek, a leading domestic large model company. Liang participated in the strategic placement, receiving 933,400 shares, and added another 258,200 shares through his quantitative firms Huanfang Quantitative and Zhejiang Jiuzhang Asset. In total, Liang holds approximately 1.19 million shares through three entities. The pairing of two Hangzhou "Six Little Dragons"—both leading domestic AI enterprises—has captured market attention. During the pre-listing investor conference, Wang Xingxing confirmed that DeepSeek's participation stems from a strategic cooperation memorandum covering collaborative R&D for general artificial intelligence, deep cooperation on high-performance general-purpose robots, and joint work on AI large models. This capital-to-business bond directly targets Yushu's core bottleneck in robot development. Given the geographic proximity of both companies in Hangzhou, Yushu may leverage DeepSeek's large model capabilities to accelerate breakthroughs in embodied brain technology.

From a performance perspective, Yushu Technology is the first humanoid robot company in China to achieve profitability. According to the prospectus, revenue grew from 159 million yuan in 2023 to 393 million yuan in 2024 and 1.699 billion yuan in 2025, achieving a compound annual growth rate of 226.78%, with 2025 growth reaching 332.64%. The company turned profitable on a net profit basis in 2024, with non-GAAP net profit of 78 million yuan that year and 591 million yuan in 2025—a 674.29% increase. Industry peers have also maintained strong revenue growth, though most remain loss-making. UBTech (9880.HK) generated 2.001 billion yuan in 2025 revenue, up 53.33%, while narrowing its net loss to 703 million yuan. CloudMinds achieved 337 million yuan in 2025 revenue, up 227%, becoming the second profitable humanoid robot company with 28.7 million yuan in net profit. Leju Intelligence posted 258 million yuan in revenue, up 365.2%, with a 70 million yuan net loss. Agibot generated 1.05 billion yuan in revenue, up 16.5 times year-on-year, with an 83.54 million yuan net loss.

However, entering 2026, Yushu Technology's revenue growth has shown signs of moderation. The prospectus reveals first-quarter 2026 revenue of 423 million yuan, up 68.49% year-on-year, with expected first-half revenue of 1.052–1.128 billion yuan, representing 35.62%–45.41% growth. Due to increased R&D and sales expenses, first-quarter non-GAAP net profit fell 52.55% to 40.25 million yuan, with first-half non-GAAP net profit expected at 236–283 million yuan, down 21.97%–6.43% year-on-year—an improvement over the first-quarter decline. The company attributes the slowdown to the rising base and intensifying competition as industry heat gradually cools from its early-2025 peak. Yushu is expanding robot applications to diversify revenue sources. Its humanoid robots serve three scenarios: scientific research and education, commercial consumer use, and industrial applications, with research and education remaining the primary revenue driver. In the first nine months of 2025, these segments generated 438 million, 103.5 million, and 53.6 million yuan, representing 73.6%, 17.39%, and 9.01% of revenue, respectively. This indicates heavy reliance on universities, research institutes, and AI laboratories purchasing humanoid hardware platforms for algorithm training and validation, funded largely by research grants. Market-oriented orders from commercial consumers and industrial applications remain limited, particularly for real-world industrial deployment. The company's more mature quadruped robots generate a higher share of market-based orders, with commercial consumer, research and education, and industrial application revenue accounting for 42.3%, 31.58%, and 26.12% respectively in the first nine months of 2025. Yet even for quadrupeds, real industrial applications contribute the smallest share.

"Research-oriented products like Yushu's primarily serve laboratory platform construction and academic needs," noted one industry investor. "Functional products targeting real industrial needs—such as material handling and low-precision operations—are expected to enter small-batch delivery this year, but full penetration of factories and homes will take time." Ding Jian observed that currently deployable opportunities are closed-loop vertical scenarios that can scale with lightweight models, though single-scenario growth faces ceilings. "Showroom performances, factory material handling, express sorting, and power inspection may be near-term scaling scenarios," he said. Li Xuenan, finance professor at Cheung Kong Graduate School of Business, proposed metrics to gauge commercialization progress: industry client revenue share, actual operating hours per robot, failure and human intervention rates, customer repurchase behavior, after-sales maintenance costs, and cost savings delivered to customers. "Once these data points emerge, we'll know whether robots are merely novel tech products or genuine production tools." Kang Yu cautioned against forcing immature projects to meet capital market expectations, while Wang Xingxing emphasized that humanoid and quadruped robot markets hold enormous potential—near-term focused on research and cultural performances, with long-term goals of entering factories and households as model generalization improves.

The valuation debate centers on whether Yushu Technology should be priced as a hardware manufacturer or an AI company. At 61 billion yuan with a 219-fold issuance P/E ratio, the valuation appears significantly elevated compared to the general equipment industry's average P/E of 38.56 times—roughly 5.7 times higher. But under an AI/embodied intelligence platform narrative, capital markets may tolerate higher valuations given the long-term growth ceiling. At the August 7 investor conference, Wang Xingxing noted that based on 2025 audited non-GAAP net profit of 591 million yuan, the issuance price corresponds to a 92.92-fold P/E. The 219.23-fold figure arises from excluding 349 million yuan in share-based compensation expenses classified as non-recurring items, bringing GAAP net profit to 278 million yuan. Gao Guolei, chairman of Shanghai Zhanghe Investment, acknowledged that the issuance valuation certainly contains bubbles, but as long as market enthusiasm for the "first humanoid robot stock" persists, the valuation can hold—whether it sustains until commercialized earnings materialize remains uncertain. Ding Jian argues that current total market capitalization of all humanoid robotics companies is under 400 billion yuan, a tiny fraction of AI hardware overall, with the embodied intelligence track potentially reaching hundred-trillion-yuan scale. "Yushu's listing, followed by other humanoid companies, will return pricing power to domestic embodied intelligence enterprises, potentially triggering systematic revaluation of the entire sector," he said.

Li Xuenan offered a framework: Yushu's valuation comprises hardware operating value plus growing AI and platform value. "The capital market's willingness to assign a high valuation essentially prices in expectations of future value. The core question is whether Yushu can convert today's 'future price' into customers, orders, data, profits, and cash flow over the coming years." Debon Securities constructed a model comparing the sector to smartphones rather than new energy vehicles. Assuming 15 million cumulative global humanoid shipments by 2035, with 5 million units sold that year at an average price of 100,000 yuan per unit, applying 2 times price-to-sales for hardware (benchmarked against Apple) and 30 times P/ARR for software (benchmarked against Tesla's 500 yuan monthly FSD subscription), combined with a 20% market share assumption for global leaders, the neutral scenario yields a 470 billion yuan combined hardware and software valuation for industry leaders by 2035, discounted back to approximately 255.6 billion yuan in 2026. The optimistic scenario approaches 400 billion yuan. However, parameter adjustments to hardware PS multiples and software pricing produce wide valuation swings.

Institutional valuations vary substantially. CCB International expects Yushu's post-listing market value to potentially re-rate to 60–100 billion yuan, projecting a 32-fold price-to-sales ratio for 2026 corresponding to 109 billion yuan. Some market participants offer optimistic forecasts of 200–300 billion yuan based on long-term industry space. Notably, lead underwriter CITIC Securities provided a reasonable valuation range of 50.6–55.9 billion yuan for 6–12 months post-listing—nearly 10% below the 61 billion yuan issuance value. Market expectations had priced the issuance at approximately 104 yuan per share based on fundraising size, but the actual price came in 45% higher. These contradictory figures highlight pricing divergence between primary and secondary markets, professional institutions, and market sentiment. The core risk for Yushu and other embodied intelligence companies remains commercialization. Gao Guolei suggests that third-party physical simulation training systems combined with industrial partners providing deployment scenarios could help create mature, reliable robots for specific applications—only through such concrete deployments can Yushu's orders become sustainable. Competitive pressures also loom: while Yushu leads the first tier, the industry landscape remains fluid, with Tesla, Agibot, UBTech, XPeng, and Xiaomi all potential threats. Additionally, Yushu acknowledges its "heavy body, light brain" weakness in the prospectus, allocating nearly half its IPO proceeds to embodied brain R&D and bringing in DeepSeek as a strategic investor to address this gap. "For embodied intelligence industry chains, valuation logic differs by segment. As a body company already profitable through domestic supply chains, Yushu will become the industry's universal valuation benchmark," Kang Yu noted. "Over the next year or two, multiple embodied intelligence companies will list. Post-listing stock price fluctuations will directly reflect capital market expectations, testing corporate management and founder psychology more severely than primary market valuations."

On the technical frontier, the industry's core bottlenecks involve physical interaction data collection and standardization, embodied large model training efficiency, and hardware iteration speed. "Data is the most critical constraint," Ding Jian emphasized. Unlike autonomous driving or text large models, embodied intelligence requires real physical action interaction to generate effective training data, yet no unified industry data standard exists, creating severe data silos that directly limit model iteration. Kang Yu added that physical AI must combine real-machine interaction data to substantially improve model performance, with current approaches split between human-centric and robot-centric data collection. Miao Tianyi, managing partner of Puzhuo Capital, stated: "Superior motion control alone no longer builds long-term barriers. The AI brain is core to next-generation product iteration. Deep software-hardware integration—not simple assembly—is the biggest current technical challenge." PwC projects three waves of Chinese smart robot adoption: 2026–2028 for high-risk, heavy-labor, and standardized commercial services; 2029–2032 for penetration into supermarkets, retail, and repetitive service scenarios as technology and standards mature; and post-2032 for semi-skilled manufacturing replacement, timing depending on technological breakthroughs and cost declines. Li Xuenan predicts embodied intelligence will first deploy in scenarios with clear tasks, controllable environments, and calculable returns—manufacturing, warehousing, logistics, energy inspection, and hazardous operations—while home environments with their non-standardized nature and safety, cost, and generalization hurdles will require much longer cycles.

Comparing domestic and international approaches, Ding Jian observed: "Overseas players like Tesla and Figure AI pursue a high-profile route—heavy investment in universal embodied large models aiming for one-step full-scenario humanoid deployment, which is difficult in the near term. Domestic companies follow a fast-iteration approach: first deploying vertical scenarios to generate cash flow, then upgrading to universal humanoid platforms. This path better matches industrial development reality." China's richer industrial scenario diversity—spanning manufacturing, exhibitions, education, and special operations—enables faster data accumulation. Kang Yu concurred that China's advantage lies in abundant offline physical scenarios for rapid physical interaction data collection, supplemented by supportive policy and capital resources. Li Xuenan predicts the industry's biggest change in coming years will involve evaluation standards. In June 2026, the Ministry of Industry and Information Technology and SASAC launched a special action for humanoid robots and embodied intelligence real-scenario training, pushing robots into real production and living environments to establish a closed loop of "real-scenario training—data accumulation—product iteration—scale deployment," requiring key products to begin "operation mode" in representative scenarios by year-end. This signals shifting industry evaluation standards.

The industry remains in a multi-point bloom stage. Yushu's first-mover A-share listing secures a leading position, but the landscape is far from settled. Tesla's Optimus, despite not yet being commercially sold, represents the strongest potential competitor given its technical capabilities and talent pool. Among domestic players, Agibot, with valuations exceeding 20 billion yuan, is considered ahead in "brain" technology; UBTech's Walker S series has entered automotive manufacturing and logistics scenarios; XPeng and Xiaomi have also crossed into robotics. "Yushu leads domestically in hardware and motion control, with clear first-mover advantages in commercialization and capital market processes, plus strong brand and talent aggregation effects. But the industry remains in full bloom, with Yushu, Agibot, and UBTech each holding distinct vertical advantages," Ding Jian said. Looking at long-term industry evolution, he envisions two potential ecosystems: closed-source integrated hardware manufacturers benchmarking Apple, earning through proprietary machines plus dedicated large models and differentiated skill subscriptions; and open-source humanoid brain platforms benchmarking Android, where third-party manufacturers integrate universal foundation models and compete through differentiated hardware. Both routes will coexist long-term without absolute dominance. China's embodied intelligence industry possesses three strategic assets: the world's most complete industrial support system lowering entry barriers, dual policy and capital empowerment supporting scale investment, and massive manufacturing scenarios and data providing fertile ground for technological upgrading and commercialization. "China's robots previously excelled at building complex machines and continuously cutting prices. The next race is about deploying machines into factories, warehouses, power grids, and mines—getting them to genuinely work every day," Li Xuenan stated. "China's vast manufacturing base itself is a globally scarce embodied intelligence training ground—data generated in operations, products iterated in scenarios, costs reduced through application. Once the 'scenario-data-product-cost' cycle runs, today's manufacturing advantages will grow into the industrial moat of the embodied intelligence era."