The booming interest in embodied AI is evident from the busy schedule of Gengxin Capital's Managing Director Mao Mingjun, who attributed his delayed response to the high activity in the sector at the start of 2026.
This sentiment reflects the heated climate in the embodied intelligence field this year. Companies like Qianxun Intelligence, Xiaoyu Zhizao, Paxini Perception, Excellent Vision, and Digua Robotics have successively secured funding, with Gengxin Capital participating as either the exclusive financial advisor or an investor.
According to IT桔子, over 100 investment deals have occurred in the downstream sector in less than three months, featuring a single high of 2.5 billion yuan and more than eight deals exceeding 1 billion yuan. February alone saw 54 financing events totaling over 12.4 billion yuan, indicating accelerated investment pace and density, with nearly 20 billion yuan raised in the first two months.
"First-tier embodied intelligence companies are the hottest, with those valued over $1 billion becoming even more sought-after despite their high valuations," Mao Mingjun noted, listing leading full-stack and "brain" companies such as Qianxun Intelligence, Zibianliang, Xinghai Tu, Yinhe Tongyong, and Excellent Vision.
Investors observe capital rapidly concentrating towards leading firms with technological and commercial closed-loops. State-owned capital has emerged as a major force, joined by industrial capital, strategic investors, and dollar funds entering the arena. Around the Spring Festival, robot orders and leasing demand increased, accelerating the path to commercial application.
However, beneath the excitement, challenges such as valuation bubbles, the capability of robotic "brains," and achieving commercial closed-loops remain for all players.
State-owned capital is playing a strong role, with preference for top-tier companies. Since February, single-round financing records have been repeatedly broken. Yinhe Tongyong Robotics completed a 2.5 billion yuan round, the largest this year; Qianxun Intelligence secured 2 billion yuan; Lingxin Qiaoshou finished a 1.5 billion yuan Series B; while Xingdong Jiyuan, Paxini Perception, Excellent Vision, Songyan Power, and Xinghai Tu each obtained financings in the billion-yuan range. Incomplete statistics show at least ten embodied intelligence unicorns valued over 10 billion yuan have emerged.
This investment wave covers "embodied brain" companies, integrated software-hardware full-stack firms, and core component suppliers, showing a trend of concentration at the top. "Compared to first-tier companies, other startups or mid-tier firms are finding it tougher," Mao Mingjun explained. "They lag in financing, valuation, and the advancement of their base models, missing out on this wave."
State-owned capital is a dominant force in this round, with various funds including industrial and strategic investment capital participating. Behind Yinhe Tongyong's 2.5 billion yuan financing are top-tier investors like the National AI Industry Investment Fund (the third phase of the "Big Fund"), Sinopec, CITIC Investment Holdings, SAIC Financial Control, and Yizhuang State-owned Capital Investment, marking the Big Fund's first move into embodied intelligence. Qianxun Intelligence's 2 billion yuan round attracted industrial capital from Synstellation Capital, TCL Capital, and Minghui Zhiyuan, as well as local state-owned funds like Chongqing Industrial Investment Mother Fund and Hangzhou Financial Investment.
"The proportion of state-owned capital is very high," Mao analyzed. "At this stage, only state-owned capital can invest large sums, including government-backed funds. Another group is major overseas dollar funds that hadn't previously invested in embodied intelligence and are now catching up."
Mao indicated that large language model companies have performed excellently in secondary markets, and the value of embodied intelligence base models far exceeds that of current large language models. Institutions recognize this core logic, leading to high investment enthusiasm for embodied model companies in the primary market.
Industrial capital is also deeply "locking in positions" to secure strategic layouts across the industry chain. For instance, CATL's Chendao Capital led the investment in Songyan Power; JD.com and SAIC heavily backed Zhuji Power; BAIC and Meituan increased their stakes in Xinghai Tu; and Baidu Strategic Investment and CRRC Capital appeared on the shareholder list of Zhi Pingfang. These industrial giants, leveraging advantages in funding, supply chains, and application scenarios, provide not only financial support but also full-chain assistance from R&D collaboration to commercial deployment.
Addressing the sector's high capital requirements, Magic Atoms President Gu Shitao emphasized that the key difference between embodied intelligence and ordinary robots is AI-driven technology. Implementing AI on physical carriers requires solving numerous software and hardware challenges simultaneously. Substantial upfront investment in software R&D is necessary before scalable growth can occur; costs related to computing power, tokens, effective data acquisition, and corresponding talent must be addressed proactively.
"Only with ample resources and solid foundations can technology achieve exponential leaps, making a 'ChatGPT Moment' possible," Gu stated.
Capital is no longer just paying for impressive technology; commercial deployment expectations are squeezing out bubbles. During the 2026 CCTV Spring Festival Gala, Magic Atoms' six humanoid robots performed a coordinated dance. At the Yibin branch venue, its humanoid robot MagicBot Gen1 acted as a "noodle chef," completing sequential tasks like lifting, draining, and serving noodles. Simultaneously, a formation of 100 MagicDog quadrupeds, styled as "robot pandas," performed synchronized routines.
Beyond stage demonstrations, Magic Atoms announced a 500 million yuan financing on March 10. "Our Gala appearance brought significant market attention and business growth," the company reported. On New Year's Eve, the 100 "robot pandas" sold out immediately upon release. ToB/G端 inquiries and partnership intentions saw exponential growth. Post-holiday, the company received numerous client inquiries from sectors including culture and tourism, government, enterprises, and education institutions, covering product procurement, customized solutions, and leasing for performances, with rapidly expanding geographical and industrial coverage.
"Humanoid robots can generate quick revenue short-term through entertainment performances, leasing, and guided tours/shopping," Gu Shitao said, noting this isn't necessarily negative as the market is growing vigorously. "We see strong demand for performances and guided services in Japan, South Korea, Singapore, as well as Europe, the Middle East, and the US."
Gu revealed that in confirmed 2026 orders, humanoid and quadruped robots each account for half. Quadrupeds have more mature application scenarios like inspection, security, and emergency response for ToB delivery, while humanoids quickly generate income via entertainment and leasing.
An investor noted that since the latter half of last year, capital has focused more on commercial deployment—orders, repurchase rates, and unit economics—as well as teams' engineering delivery capabilities. Technological appeal alone no longer supports high valuations; bubbles in commercial expectations are being deflated, with funds concentrating towards leading firms and those with real orders.
This shift means capital prioritizes genuine orders and scene implementation over flashy tech demos. The rise of leasing models has become a key for humanoid robot manufacturers to unlock commercialization.
On March 18, robot leasing platform Qingtian Zu announced completing angel and angel+ rounds, raising a cumulative total in the hundred-million-yuan range. CEO Li Yiyan and Chief Strategy Officer Wang Mingfeng stated that current leasing focuses on entertainment scenarios, including commercial performances, store openings, and guided tours for sporadic orders, as well as project-based scenes like robot exhibitions and robot fighting tournaments. Qingtian Zu is also exploring new applications such as robotic patrols, security checks, office clock-ins, and restaurant promotion.
"Our orders increased 70% month-over-month during the Spring Festival. We've observed clients previously opting for short-term leases shifting to long-term leases post-holiday, such as restaurants in malls gradually adopting robots for customer attraction and interaction as regular setups," Li Yiyan said. As robots deploy in more scenarios, leasing platforms' reachable customer base will expand, further broadening client range and enriching monetization paths.
Leasing a Unitree robot for dance performances isn't cheap, and competition is shifting from "hardware" to "intelligence." "The Gala reignited interest in robot performances; March and April are relatively off-peak, so overall order volume is acceptable," said Hu Jingling, founder of Kunming Shenlin Technology. Current leasing market volume is similar to last year.
Yan Junru, Business Director at Yaowu Technology's robot division, shared this sentiment, noting post-Gala inquiries for Unitree's Gala-style performances emerged quickly, but "order volume is normal, not as hot as last year."
Post-Gala, Kunming Shenlin received many leasing requests for Gala-style dance performances, which are challenging to fulfill as Unitree hasn't open-sourced the specific dance code. Hu Jingling also has practical concerns: "The performance is spectacular but highly difficult. Usually, if robots perform pre-loaded dances, malfunctions are covered by free repairs from the manufacturer; but if we develop our own choreography and issues occur during shows, we can't get free repairs."
Cost is another factor; the Gala-featured Unitree H2 humanoid robot stands about 180 cm tall, making intercity transport relatively expensive. "We plan to buy one or two units this year, not more."
Four humanoid robot brands competed on the Gala stage. According to Hu's observations, Unitree dominates the leasing market currently, as "its martial arts performances offer stronger visual impact, and compared to skits, martial arts and dances are more suitable for general commercial events."
Multiple robot lessors indicated that application scenarios for humanoid robots haven't changed much, still centering on tourist site filming, opening events, and annual party performances—where "creating a lively atmosphere is key."
Despite catalyzing by billion-yuan financings, the humanoid robot sector faces challenges. The industry remains in early commercial stages, with most companies yet to achieve profitability or even large-scale product delivery.
"Last year was 'a hundred flowers blooming'; this year is more accurately 'all rivers flow to the sea'," said Liu Yang, co-founder of Original Force Unlimited, summarizing the financing surge with three terms: accelerated consensus, intensified differentiation, and narrowing windows. Liu believes 2026's core battleground in embodied intelligence will focus on three key dimensions: who possesses a self-controlled "brain," who holds the largest scale of real-scene data, and who can first achieve a commercial closed-loop from "proof of concept" to "repeat purchases."
"Bubbles mainly accumulate from the mismatch between 'technological concepts' and 'valuations'. Many companies have model capabilities but lack hardware carriers, or have hardware but lack generalization ability. Funds will flow more to leading firms and those with real orders," an investor commented.
This implies that humanoid robot companies must not only consider commercial deployment but also compete in the transition from "hardware" to "intelligence." The Zhiyuan Research Institute predicts that over 230 domestic robot-related enterprises will face consolidation adjustments in 2026. Newding Capital stated bluntly that among the more than 100 robot companies that received funding in 2025, only 10-20 may survive by 2026, indicating an industry elimination rate of 80-90%.
For 2026 industry trends, Gu Shitao offered three predictions: "First, companies will gradually start selling products, sinking deeper into scene implementation. Second, there will be divergent route choices in AI direction, and capital will select; some will succeed while others fall behind. Third, commercial application deployment will also see reshuffling, leading to a new round of industry consolidation."
"The opportunity is certain, but the winners are not yet determined," Gu concluded.
Comments