Two Months, 20 Billion in Capital Inflow: The Capital "Positioning Race" for Humanoid Robots

Stock trading relies on Golden Kylin Analysts’ reports—authoritative, professional, timely, comprehensive—helping you discover potential thematic opportunities!

Image/IC

“If the track wasn’t popular, I wouldn’t be so busy replying to you so late.” A few days after the interview news was released, Miao Mingjun, Managing Director of Gengxin Capital, responded this way.

This remark is also a microcosm of the booming embodied intelligence track at the start of 2026.

Since this year, multiple embodied intelligence companies such as Qianxun Smart, Xiaoyu Intelligent Manufacturing, Pasini Perception, Jiajia Vision, and Digua Robots have successively completed financing rounds, with Gengxin Capital participating as exclusive financial advisor or investor.

IT Juzi data shows that in less than three months this year, over 100 downstream investment events have occurred in the track, with the largest single financing reaching 2.5 billion yuan, and more than 8 deals exceeding 1 billion yuan. In February alone, 54 financing events took place, with total funding surpassing 12.4 billion yuan. The pace and density of financing are increasing, with nearly 20 billion yuan raised in the first two months.

“Top-tier embodied intelligence companies are the hottest; companies valued over $1 billion are even more sought after as their valuation rises,” said Miao Mingjun in an interview with Beijing News Shell Finance, listing leading full-stack companies and brain companies, including Qianxun Smart, Variable, Xinghaitu, Galaxy General, and Jiajia Vision.

In the eyes of many investors, capital is accelerating toward leading companies with complete technology and business loops. State-owned capital has become a strong main force, with industry capital, strategic investment funds, and other sources participating together. US dollar funds are also entering the scene. Meanwhile, before and after the Spring Festival, robot orders and leasing demand have surged, and commercialization pathways are accelerating.

However, behind the excitement, valuation bubbles, the “brain” capabilities of robots, and the threefold test of commercial loops are challenges faced by all players.

State-owned capital as a major force, more capital favors top-tier companies

In the first quarter of this year, the humanoid robot track saw a “battle for funding.”

Since February, record-breaking financing rounds have been frequent. Galaxy General Robots completed a 2.5 billion yuan new round, becoming the largest single financing so far this year; Qianxun Smart secured 2 billion yuan; Lingxin Qiaoshou completed a 1.5 billion yuan Series B; companies like Xingdong Jiyuan, Pasini Perception, Jiajia Vision, Songyan Power, and Xinghaitu all received large-scale funding of around 1 billion yuan. According to incomplete statistics, at least 10 billion-dollar unicorns have emerged in the embodied intelligence track.

Regarding the types of companies receiving investment, this round of enthusiasm covers “embodied brain” companies, “soft-hardware integrated” full-stack companies, and core component enterprises, showing a trend of top-tier clustering. “Compared to the first-tier companies, startups or mid-tier companies will find it harder,” said Miao Mingjun to Shell Finance. “They lag behind in funding, valuation, and the leading edge of brain models, so they can’t catch this wave of hotness.”

In this financing wave, state-owned capital is a major force, with multiple sources participating. For example, Galaxy General’s 2.5 billion yuan financing involves the National AI Industry Investment Fund (Phase III), Sinopec, CITIC Investment Holdings, SAIC Financial Holdings, and Yizhuang State Investment, among others. Notably, the National Big Fund made its first move into the embodied intelligence track; Qianxun Smart’s 2 billion yuan financing includes industry capital like Synstellation Capital, TCL Venture, Minghui Zhiyuan, as well as local state-owned funds like Chongqing Industry Investment Fund and Hangzhou Jin Investment.

“State-owned capital accounts for a very high proportion,” Miao Mingjun analyzed. “Because at this stage, only state-owned funds can make large investments, including some government-backed funds. There’s also a category of large overseas US dollar funds that haven’t invested in embodied intelligence before and are now stepping in.”

Miao Mingjun pointed out that large language model companies perform very well in the secondary market, but the value of embodied model frameworks far exceeds current large language models. Institutions have recognized this core logic, leading to high enthusiasm for investing in embodied model companies in the primary track.

Industry capital is also deep “locking in” positions, completing upstream and downstream chain layouts. For example, CATL’s affiliated Chendao Capital led a round of investment in Songyan Power; JD.com and SAIC heavily invested in Zhiqi Power; BAIC and Meituan increased their stakes in Xinghaitu; Baidu’s strategic investment and CRRC Capital appeared among the shareholders of Zhipian. These industry giants leverage advantages in capital, supply chain, and application scenarios, providing not only funding but also full-chain support from R&D collaboration to commercialization.

Regarding the high capital demand in the track, Gu Shitao, CEO of Magic Atom, believes that the biggest difference between embodied intelligence and ordinary robots is AI-driven. AI technology must be physically implemented, which requires solving numerous hardware and software issues simultaneously. Early-stage software R&D demands large investments, and only then can scale growth be expected; costs for computing power, tokens, acquiring effective data, and talent are all upfront challenges for companies.

“Only when the resources are solid and abundant can technology achieve exponential breakthroughs, leading to the ChatGPT Moment—an industry-changing breakthrough like ChatGPT,” said Gu Shitao.

More than just flashy technology, commercialization expectations are “bubbling”

On the stage of the 2026 CCTV Spring Festival Gala, six humanoid robots from Magic Atom performed a dance together. At the Yibin sub-venue, their humanoid robot MagicBot Gen1 transformed into a “noodle-making master,” completing tasks like “lifting dough—controlling water—noodle pulling.” Meanwhile, a formation of hundreds of MagicDog quadruped robots performed synchronized group shows.

Beyond stage tricks, Magic Atom announced on March 10 that it secured 500 million yuan in funding. “The Gala appearance brought us noticeable market attention and business growth,” a Magic Atom representative told Shell Finance. On the night of the Lunar New Year’s Eve, 100 “machine pandas” appeared on the Gala and sold out immediately; in B2B and G-end businesses, inquiries and cooperation intentions surged exponentially. After the holiday, the company received numerous customer inquiries across industries such as tourism, government, enterprise, and education, involving product procurement, customized solutions, and leasing performances, rapidly expanding the geographic and industry scope.

“Humanoid robots can generate quick revenue in the short term mainly through entertainment performances, leasing, and guided tours,” Gu Shitao said. “This isn’t necessarily a bad thing; this market is booming. We see strong demand for performances and guided tours in Japan, Korea, and Singapore, and the same in Europe, the Middle East, and the US.”

Gu Shitao revealed that among confirmed orders for 2026, humanoid robots and quadruped robots each account for half. Quadruped robots have more mature application scenarios, such as inspection, security, and emergency rescue for B2B delivery; humanoid robots, on the other hand, generate quick revenue through entertainment and leasing.

An investor told Shell Finance that since the second half of last year, capital has been more focused on business implementation (orders, repurchase rates, UE models) and the engineering delivery capability of teams. Having only technological halo is no longer enough to support high valuations; the bubble of commercialization expectations is being pricked, and funds are concentrating on leading companies with real orders.

This means that capital is no longer just paying for flashy tech demos but is increasingly valuing real orders and scene implementation. The rise of leasing models has become a key to opening up commercialization for humanoid robot manufacturers.

On March 18, Qingtian Leasing, a robot leasing platform, announced the completion of angel and angel+ rounds, with total funding reaching hundreds of millions of yuan. CEO Li Yiyan and Chief Strategy Officer Wang Mingfeng told media outlets including Shell Finance that their current robot leasing mainly targets entertainment scenarios, covering commercial performances, openings, guided tours, and other scattered orders, as well as projects like robot exhibition halls and robot fighting competitions. They are also exploring new scenarios such as robot dog patrols, security checks, and robots for shift work and dining customer attraction.

“During the Spring Festival, our order volume increased by 70% month-on-month. We also observed that customers who previously focused on short-term rentals are now shifting toward long-term rentals, such as shopping malls and restaurants using robots for customer attraction and interaction as routine setups,” Li Yiyan said. “As robots are deployed in more scenarios, the accessible customer base will continue to expand, and business scenarios will further diversify, broadening monetization paths.”

The high cost of renting a Yushu robot for dancing, and shifting competition from “hardware” to “brain”

“The Spring Festival Gala revived robot performances, and March and April are relatively slow seasons, so overall order volume is still decent,” said Hu Jingling, founder of Kunming Shenlin Technology Co., Ltd., in an interview. He noted that the overall order volume in the robot leasing market is about the same as last year.

Yan Junru, Business Director of Ruotech’s robot division, shared similar feelings. “After the Spring Festival, we received inquiries from clients wanting the same performance as Yushu’s Gala show, but the orders are normal, not as hot as last year.”

After the Gala, Kunming Shenlin received many leasing orders requesting performances of the same dance as the Gala. However, fulfilling this is not easy because Yushu has not open-sourced the code for the Gala dance. Hu Jingling also expressed practical concerns: “The performance is spectacular, but the difficulty is high. Usually, if a robot performs a self-developed dance, the company will repair faults for free; but if the dance is self-developed and faults occur during the show, repairs won’t be free.”

Cost is also a concern. The Yushu H2 humanoid robot, about 180 cm tall, has relatively high transportation costs. “This year, we plan to buy 1-2 units, not more,” Hu Jingling said.

At the Spring Festival stage, four humanoid robot brands competed. Hu Jingling observed that the market for robot leasing still mainly revolves around Yushu, whose martial arts performances deliver a stronger visual impact. Compared to skits, martial arts and dance are more suitable for general commercial performances.

Several robot leasing companies said that the scene for humanoid robot performances hasn’t changed much, mainly used for scenic spot filming, opening events, and annual meetings—“festivity is key.”

Massive funding has catalyzed the humanoid robot track, but it’s not all smooth sailing. The industry remains in early commercialization, with most companies not yet profitable or having achieved large-scale product delivery.

“Last year was ‘a hundred flowers blooming,’ but this year is more accurately described as ‘all rivers flowing into the sea,’” said Liu Yang, co-founder of Yuanli Wuxian, summarizing the recent financing boom with three words: consensus acceleration, increasing differentiation, and shrinking window. Liu believes that by 2026, the core battleground of embodied intelligence will focus on three key dimensions: who controls autonomous “brains,” who has the largest real scene data, and who can first establish a business loop from “concept verification” to “repeat purchases.”

“Most of the bubble is in the mismatch between ‘technology concept’ and ‘valuation.’ Many companies only have model capabilities but lack hardware carriers, or have hardware but lack generalization ability. Capital will flow more to leading companies with real orders,” an investor told Shell Finance.

This also means that humanoid robot companies need to think beyond just commercialization; they must compete from “hardware” to “brain.” The Institute of Zhi Yuan predicts that by 2026, over 230 domestic robot-related companies will face contraction or adjustment. New Ding Capital states that of the more than 100 robot companies that secured funding in China in 2025, only 10-20 may remain by 2026, with an industry淘汰率 of 80%–90%.

Regarding the industry outlook for 2026, Gu Shitao offered three predictions: “First, companies will start gradually selling products and sinking into scene implementation. Second, in AI, different routes will emerge, with capital choosing sides; some will succeed, others will fall behind. Third, there will be a new wave of reshuffling in commercial application deployment.”

“Opportunities are certain, but winners are still uncertain,” Gu Shitao concluded.

Reporters: Wei Boya, Zhang Xiaohui, Beijing News Shell Finance