2026 is widely recognized by the industry as the decisive year for humanoid robot commercialization. On one side, Unitree Robotics passed its STAR Market IPO review in just 73 days, securing the position of the first humanoid robot stock on China’s A-share market, with offline directly operated stores opening successively in Beijing and Shanghai to probe the consumer market through physical products. On the other side, Agibot is following a platform-based ecosystem approach, splitting into six subsidiaries under a “6+1” organizational structure and laying out a full industry chain spanning leasing, components, data, and quadruped robots — transitioning from hardware manufacturing to an industrial conglomerate. The two companies that together account for more than 70% of global humanoid robot shipments have taken two divergent paths, laying the competing strategic approaches of China’s embodied intelligence track squarely on the table.
The IPO Divide: Unitree Reaches the Capital Market First, But Growth Concerns Lurk Behind the RMB 42 Billion Valuation
On June 1, Unitree Robotics successfully cleared its IPO review. From the acceptance of its prospectus on March 20 to approval at the review meeting took only 73 days, setting a new speed record for similar companies reviewed on the STAR Market. Based on IPO fundraising of RMB 4.202 billion (≈ USD 583.61 million) with issued shares representing 10% of the total, Unitree’s issuance valuation is set at RMB 42 billion (≈ USD 5.83 billion). Founder Wang Xingxing, with a direct and indirect combined stake of 33.36%, stands to see his personal net worth exceed RMB 14 billion (≈ USD 1.94 billion) after listing, making him a benchmark wealth-creation story on the embodied intelligence track. Meituan, as the largest external shareholder, holds 9.65%; Shunwei Capital (Lei Jun) holds 4.42%; and internet giants including Tencent and Alibaba have all already taken positions.
Looking back over a decade of entrepreneurship, Wang Xingxing left DJI in 2016 to start his own company, building on the XDog prototype he developed during his master’s program, and spent years cultivating the niche track of quadruped robots that few believed in. He first captured nearly 70% of the global quadruped market share, then in 2023 seized the AI opportunity to pivot into the humanoid track. Within three days of each other, Unitree and Agibot — which was founded around the same period — launched their first humanoid products, the H1 and the Yuanzheng A1 respectively, formally opening the dual-champion standoff. Unitree’s performance during the 2025 Spring Festival Gala became a critical inflection point in the company’s development: the viral scene of robots spinning handkerchiefs set the entire internet ablaze, directly driving full-year humanoid shipments to 5,500 units, with annual revenue surging 333.42% to RMB 1.699 billion (≈ USD 236.0 million), net profit of RMB 278 million (≈ USD 38.61 million), and a gross margin as high as 62.91% — making it one of the very few companies in the track to have achieved scaled profitability.
Beneath the bright lights, growth concerns have already emerged. In Q1 2026, Unitree’s revenue was RMB 423 million (≈ USD 58.75 million), with year-on-year growth slowing from 332% in the same period last year to 68.49%, and non-recurring net profit nearly halved year-on-year. The profit decline is primarily attributable to two rising expenditure categories: on one hand, the Spring Festival Gala and offline store expansion drove a sharp increase in sales expenses; on the other hand, the company has begun to increase investment in what was previously its weak area — embodied large model R&D. Over the past three years, Unitree’s cumulative R&D investment was only RMB 260 million (≈ USD 36.11 million), with an R&D expense ratio of 8.53% in 2025, far below the industry average of 23% to 26%. In the prospectus, Unitree explicitly earmarks half of its IPO fundraising proceeds — approximately RMB 2.1 billion (≈ USD 291.67 million) — for general large model R&D, being compelled to make up the shortfall in the robot’s “brain” after years of prioritizing hardware over AI.
To offset the ceiling of the scientific research customer market and find new incremental growth, Unitree has simultaneously launched an offline channel push. During the 73-day window of its IPO sprint, the company opened its first directly operated store at Beijing’s Wangfujing and Asia’s first embodied intelligence experience center in Shanghai’s Jing’an district, giving consumers the opportunity to experience humanoid robot interactive performances up close. This replicates Tesla’s early approach of cultivating the new energy market through offline experiences and is a key strategic move for Unitree to break free from its dependence on universities as a single customer source. Currently, 73.6% of Unitree’s humanoid revenue comes from universities and research institutes. These customers typically purchase one to two units at a time, with limited room for growth in the existing market. A revenue structure where commercial use accounts for 17.4% and industrial deployment for less than 9% means relying solely on B-end research clients cannot absorb the continuously rising production capacity.
Agibot Takes a Different Path: The 6+1 Ecosystem Fission, Building a Full Industry Map Without an Independent IPO
In contrast to Unitree’s focus on complete machines and sprint toward an independent IPO, Agibot — founded as recently as 2023 — has taken the completely opposite platform-based path. In July 2025, Agibot took a controlling stake in Shangwei New Materials. In June 2026, co-founder Peng Zhihui was appointed as chairman of the listed company, staking out a position in the capital market ahead of the game under the banner of “the first A-share humanoid robot stock” — though this claim has not received formal regulatory or capital market recognition, it nonetheless completed an early positioning in the capital markets.
In terms of organizational structure, Agibot has implemented a unique 6+1 development model: the parent company focuses on humanoid body R&D, while six founding partners each spin off and incubate six independent subsidiaries, each independently funded and market-operated, with healthy internal competition, covering the entire upstream, midstream, and downstream of the supply chain. Among them, Qingtian Leasing, as a robot leasing platform, completed six rounds of funding in six months with its valuation surging to RMB 7 billion (≈ USD 972.22 million), breaking the industry leasing track fundraising record. The platform has opened up to accommodate all brands including Unitree and AgilityBio, directly cutting into Unitree’s advantage in the leasing market. Linjiezdian focuses on proprietary dexterous hand R&D, spinning off the internal core components business to supply the external market. Mifeng Technology is dedicated to robot training data, transitioning from an internal service provider to an industry-wide data supplier. Deng Taihua personally helms Zhiyuan Kaituo, officially entering the quadruped robot track where Unitree originated. Zhiding focuses on commercial cleaning deployment; Shangwei New Materials handles the capitalization function.
At the product deployment level, Agibot has abandoned the Spring Festival Gala traffic race, independently hosting its own “Robot Wonderful Night” as its counterpoint to Unitree’s Spring Festival exposure. At international exhibitions such as CES and MWC, rather than spectacular performances, it leads with real B-end deployment scenarios such as factory logistics and in-store navigation. On the volume production front, by the end of March 2026, Agibot’s complete machine production volume surpassed 10,000 units; going from 5,000 to 10,000 units off the line took only three months. The company has set a target of tens of thousands of units for full-year production, with a long-term ambition of breaking RMB 10 billion (≈ USD 1.39 billion) in revenue. Leveraging its full-stack proprietary software and hardware R&D advantages, Agibot has developed multiple embodied large models including GO-1, using the special ViLLA architecture to train on three types of data — simulation, internet video, and real machine data — in an attempt to improve the robot’s cross-scenario generalization with less real machine data, differentiating itself from Unitree’s technical approach of emphasizing real machine data collection.
Within Agibot’s internal industry framework definition, Unitree is still at the X curve (novelty performance and research procurement stage), while Agibot has already moved ahead into the Y curve (scaled productivity deployment). Behind these two positionings lies the fundamental divergence in the two companies’ underlying logic: Unitree believes in hardware products leading the way, making robots into standardized industrial goods and naturally stimulating consumer demand once costs come down; Agibot has laid out the full industry chain in advance, locking in high-value intermediate segments such as data, channels, and components to build an industrial ecosystem moat.
Twenty Players Competing on the Same Stage, the Hundred-Billion Club Expands, the Industry Bids Farewell to Single-Point Dividends
The strategic debate between Agibot and Unitree is only a microcosm of the white-hot competition currently underway in China’s embodied intelligence sector. In 2026, there are already 20 domestic humanoid robot companies that have achieved batch delivery. Nine companies including Galaxy General, Zibianliang, and Xinghaitu have entered the hundred-billion-yuan valuation club. Track fundraising continues to explode, with the entire industry chain raising more than RMB 256 billion (≈ USD 35.56 billion) in Q1 2026 alone.
Track differentiation has already become clear. Galaxy General, backed by national team funding, focuses on the practical route of factory transportation. Magic Atom, relying on the Dreame Technology ecosystem, has invested hundreds of millions (in RMB) to appear on the Spring Festival Gala, offering separate performance and commercial product lines. Songyan Power and AgilityBio focus on ten-thousand-yuan-level affordable small humanoids, competing for the entry-level consumer market. Zibianliang has secured investments from all three major internet giants — Alibaba, ByteDance, and Meituan — and focuses on general embodied large model R&D. Leju Robotics and UBTECH are deeply embedded in industrial deployment, each sprinting toward their respective IPOs. At the same time, manufacturing giants such as BYD, Luxshare, and Huaqin have entered from adjacent industries. Tesla’s Optimus is set to enter production in the summer of 2026, with overseas giants and crossover players continuing to compress the survival space of native companies.
The industry-wide wave of offline store openings also indirectly exposes a shared pain point across the industry: production capacity has rapidly moved into the tens-of-thousands-of-units era, but real-world deployment scenarios are severely lacking. At this stage, all humanoid robot companies are highly dependent on university research orders for revenue. Consumer household deployment generally requires another five to eight years of technological maturation, and industrial-scale deployment is still in pilot testing. UBS projects global humanoid shipments of only 30,000 units in 2026, with 150,000 units only a realistic prospect by 2030, as the industry remains in an extended validation cycle from zero to one. Quadruped robots remain the industry’s “cash cow” — both Unitree and Yushu Technology earned all of their early profits from their mature quadruped businesses, and humanoids are unlikely to independently support companies’ profit foundations in the short term.
The Final-Year Cliffhanger: Does Hardware Win, or Does the Ecosystem Prevail?
From simultaneously launching their first products in 2023 to a full-dimensional battle across capital, channels, and products in 2026, the contest between Agibot and Unitree has long since transcended a simple comparison of shipment numbers. Unitree holds IPO cash firepower and is stabilizing its shipment base through mature hardware and cost advantages, while doubling down on large models to fill its software gap. Agibot uses the platform split model to disperse R&D risk, locking in high-value segments including channels and data ahead of time.
With 2026 as the critical decisive year for real-world deployment, the industry’s “scenario discount coefficient” has become the yardstick hanging over every company’s head — for companies that cannot achieve real-world scenario deployment, valuation discounts will continue to widen. In the short term, Unitree enjoys the IPO dividend and mature hardware products to command advantages in both shipments and capital access. In the long-term competition, whoever can first break free from the single university market, replicate success with industrial mass-production deployment, and cultivate civilian consumer demand will be the one to define the future of the humanoid robot industry.
In the next three to five years, the outcome will be neither a single hardware manufacturer dominating the landscape nor a pure model company overtaking the field out of nowhere. Only companies that combine hardware mass-production capability with a real-world deployment ecosystem stand a chance to navigate through the industry’s early-stage bubble and establish a firm foothold in the trillion-yuan embodied intelligence wave.
[Disclaimer]: The above content reflects analysis of publicly available information, expert insights, and BCC research. It does not constitute investment advice. BCC is not responsible for any losses resulting from reliance on the views expressed herein. Investors should exercise caution.