From Indirect Benchmarking to Head-to-Head Competition: Unitree vs. Zhiyuan in 2025

Technology Author: EqualOcean News, Yongli Lu Editor: Hanchen Meng, Yiran Xing Feb 09, 2026 04:47 PM (GMT+8)

From Implicit Rivalry to Open Competition: Who Will Define the Rules of the Future?

robot

On one side, Unitree Robotics (宇树科技) issued an urgent public clarification, stating that its actual shipments of fully humanoid robots exceeded 5,500 units in 2025, directly countering a third-party ranking that placed the company second in the industry. The move marked an explicit effort to reclaim narrative control over shipment data and industry positioning.

On the other, Zhiyuan Robotics (智元机器人) abruptly announced that it would forgo its planned appearance at the 2026 Spring Festival Gala, instead unveiling its own event—“Robot Wonder Night”, scheduled for February 8 and billed as the world’s first fully robot-led showcase. By building an independent stage rather than competing for traditional mass exposure, Zhiyuan deliberately differentiated its branding strategy from the conventional Spring Festival route adopted by peers.

These closely timed developments in early 2026 were not isolated public relations maneuvers. Rather, they signaled a visible escalation in competition between Unitree and Zhiyuan, the two leading players in China’s humanoid robotics sector.

Since 2025, the rivalry between the two companies has unfolded not through sudden confrontation, but through a series of progressively layered moves. Beneath actions that appeared independent on the surface lay a broader contest over industry leadership, commercialization pathways, and capital market narrative dominance.

I. Strategic Positioning (May–July 2025): Implicit Competition for Sector Footholds

As early as the first half of 2025, the moves made by Zhiyuan Robotics and Unitree Robotics appeared parallel and independent on the surface, yet beneath them lay a subtle contest over the binding of technology and application scenarios.

In late March, prominent investor Zhu Xiaohu publicly raised concerns about a potential “bubble” in humanoid robotics, questioning the near-term viability of commercialization. His remarks functioned as an industry-wide stress test, setting the tone for the first half of the year: a dual-track approach in which technical demonstrations aimed at public visibilityand efforts toward tangible commercial validation progressed simultaneously.

In the early stages of the “bubble debate,” Unitree repeatedly emphasized—through statements attributed to early investors—that it had achieved sustained profitability since 2020 and maintained healthy financials. This messaging was intended to distance the company from narratives of capital-intensive, loss-driven experimentation.

However, a setback at the world’s first humanoid robot marathon in April triggered a reputational challenge. When Unitree’s G1 humanoid robot fell at the start of the race, the company attributed the incident to a third-party self-developed algorithm and stated that it was not a direct participant—responses that were widely criticized as deflecting responsibility. Subsequent long-distance running livestreams intended to demonstrate recovery were met with further skepticism, including accusations of overly idealized environments and potential pre-recording. A series of controversies led to Unitree being labeled by some commentators as a “performance-oriented” rather than application-driven company.

In response, Unitree repositioned the G1 robot on May 25 through a live broadcast appearance on CCTV’s mecha combat arena, subjecting the system to high-intensity physical confrontation. By publicly demonstrating impact resistance and high-dynamic stability under extreme conditions, Unitree reframed these technical capabilities as indicators of industrial-grade reliability, securing a critical form of validation for entry into high-requirement entertainment and live-performance scenarios.

afd0df11b80349ed9314c24af58677c5_941x498.jpg.jpg

Source: Unitree Robotics (Official Website)

On May 28, Unitree Robotics completed its corporate restructuring into a shareholding company, a move widely interpreted as a standard preparatory step toward a future public listing. Almost simultaneously, founder Wang Xingxing publicly stated that orders for the company’s humanoid robots had “surged beyond expectations,” signaling strong near-term commercial demand.

In parallel, Unitree partnered with multiple industry players to establish an Embodied Intelligence Pilot and Testing Base, aimed at accelerating the transition from technological validation to productization. Taken together, these developments conveyed a clear message to the market: Unitree had established a relatively complete chain spanning technology verification, order acquisition, industrial collaboration, and capital market readiness.

By contrast, Zhiyuan Robotics pursued a markedly different approach, deploying a series of coordinated moves to systematically demonstrate its capacity to define ecosystems and shape future industry structures.

Shortly after the “bubble debate” gained traction, Zhiyuan convened its first supplier conference, announcing that it had cumulatively mass-produced and delivered 1,000 units, while simultaneously assembling a supply chain network of more than 200 partners. This was positioned as evidence of its ability to scale production and manage costs—key prerequisites for sustainable commercialization.

Between April and May, Zhiyuan further intensified its ecosystem-building efforts. It launched Genie Studio, a one-stop development platform, released open-source simulation and evaluation tools, and announced plans to host a global robotics competition. Through open tooling, benchmark-setting, and developer aggregation, Zhiyuan sought to rapidly establish itself as the central node of a broader technical and industrial ecosystem—positioning itself not merely as a product supplier, but as a rule-setter shaping the future competitive framework of humanoid robotics.

The competition between the two companies became particularly symbolic in their contest for the entertainment and live-performance segment, a key early commercialization scenario. Unitree emphasized performance through physical confrontation, using robot combat demonstrations to validate extreme capabilities and pave the way for high-performance shows. Almost concurrently, on May 21, Zhiyuan explicitly designated entertainment and live exhibitions as a core commercial application for its Lingxi X2 robot, formally initiating partner recruitment.

One approach emphasized technical training and validation; the other focused on scenario definition and market entry. Together, these contrasting strategies crystallized the underlying competitive tone.

By June–July, the rivalry escalated into an IP positioning battle. Zhiyuan moved first, partnering with Pepsi China to launch a customized robot IP, “Pepsi Blue,” followed by the debut of a second customized model during the World Artificial Intelligence Conference (WAIC). Concurrently, it signed a cooperation agreement with the Shanghai Theatre Academy, aiming to construct a replicable commercialization model through a combination of brand partnerships and institutional endorsement.

Unitree, meanwhile, adopted an alternative path. On July 30, it announced a strategic partnership with Giant Star Group, a company associated with pop icon Jay Chou, targeting high-value scenarios such as concert choreography and fan interaction. By binding its technological capabilities to top-tier entertainment IPs, Unitree sought to leverage scarce cultural and traffic resources for amplified market impact.

At this stage, the divergence in strategic positioning became increasingly clear. Zhiyuan emerged as an “ecosystem architect,” pursuing broad expansion through standardized cooperation frameworks and platform-centric growth. Unitree positioned itself as a “premium resource integrator,” leveraging technical branding to anchor scarce IPs and extract outsized value through exclusivity. Though their paths differed, both strategies ultimately converged on the same objective: securing narrative authority and strategic high ground in the humanoid robotics market, setting the stage for more direct confrontation in the second half of the year.

II. Direct Confrontation (H2 2025): A Multi-Dimensional Contest Across Capital, Technology, and Commercialization

Entering the second half of 2025, the humanoid robotics sector began to shift from a phase characterized by technical showcases and scenario exploration toward one defined by capital acceleration and commercial validation. Competition between Unitree Robotics and Zhiyuan Robotics transitioned from implicit benchmarking to an openly visible, head-to-head contest. Actions on both sides intensified across three dimensions: capital strategy, technology and product development, and commercialization with ecosystem building.

At the capital level, Unitree adopted a distinctly conservative and regulation-aligned approach. In late May, the company completed its conversion into a shareholding entity, followed by the closing of a Series C financing roundaround June, co-led by China Mobile, Tencent, Alibaba, Ant Group, and Geely, valuing the company at approximately RMB 12 billion post-money. During the same period, founder Wang Xingxing disclosed that Unitree’s 2024 revenue had exceeded RMB 1 billion, with the company having achieved sustained profitability.

In July, Unitree formally filed for IPO counseling, and by November had completed the counseling acceptance process, moving rapidly along the standard domestic A-share IPO pathway. The underlying logic was to leverage its disclosed profitability to demonstrate regulatory compliance and investment value as a “hard-technology manufacturer,” positioning itself as a candidate for China’s first A-share-listed humanoid robotics company.

That said, Unitree’s plan to formally submit its IPO application in the fourth quarter of 2025 did not materialize as scheduled. In early January 2026, market rumors emerged suggesting that a proposed “fast-track listing channel” had been suspended. Unitree subsequently clarified that it had never applied for such a channel and reaffirmed that its listing process remained on track.

Zhiyuan, by contrast, pursued a more proactive and aggressive capital strategy. Even before Unitree completed its shareholding reform, Zhiyuan posted openings for a securities affairs director on recruitment platforms—widely interpreted by the market as an explicit pre-IPO signal.

In July, as attention focused on Unitree’s fundraising and valuation, Zhiyuan announced plans to acquire a controlling stake in Shanghai Weaver New Material (上纬新材), a listed A-share company, for RMB 2.1 billion. The transaction—often described as a “reverse acquisition”—triggered intense market reaction, with the target company’s share price reportedly rising by over 1,000% within ten days.

While Zhiyuan denied any intention of pursuing a backdoor listing, the move effectively enabled it to bypass the lengthy IPO queue and gain immediate access to a mature A-share capital platform, significantly enhancing flexibility for future fundraising through instruments such as private placements.

In the private market, public records indicate that Zhiyuan completed at least four financing rounds in 2025. Following a Series B round led by Tencent in March, the company’s valuation reportedly reached RMB 15 billion. In August, Zhiyuan further introduced LG Electronics as a strategic investor and partnered with Hillhouse Capital to establish an industry fund, launching the “Zhiyuan A Program” accelerator. This combined strategy—platform acquisition, strategic investor onboarding, and ecosystem formation—was designed to rapidly assemble an industry-capital alliance centered on Zhiyuan itself.

On the technology and product front, Unitree concentrated on hardware performance leadership and price accessibility. In July, it released the R1 lightweight humanoid robot priced at RMB 39,900, breaking through the consumer-grade pricing threshold. In October, it followed with the H2 full-size humanoid robot, forming a “dual-tier” product lineup that spans consumer, research, and high-end industrial applications.

Technically, Unitree upgraded its G1 anti-gravity balance algorithm, enabling recovery from a fall within 0.3 seconds, and secured four gold medals at the World Humanoid Robot Sports Competition. These achievements further reinforced its positioning as a leader in motion control and dynamic hardware performance.

image.png.png

Source: Unitree Robotics (Official Website)

In December, Unitree launched a dedicated robot “App Store”, marking a critical step toward platformization. The platform enables developers and users to upload and download motion models—including dance routines, martial arts movements, and task-oriented actions—which can be deployed to robots with one-click installation. This initiative signaled Unitree’s attempt to shift humanoid robots from standalone products toward a developer-driven platform ecosystem, expanding functional diversity through modularized motion capabilities.

Zhiyuan, by contrast, pursued a more system-oriented strategy centered on its “One Architecture, Three Intelligences” full-stack framework and a comprehensive, scenario-spanning product portfolio. Rather than positioning itself as a pure hardware manufacturer, Zhiyuan articulated a longer-term ambition to become an ecosystem integrator and standards contributor for the embodied intelligence era. Throughout the second half of 2025, the company advanced in parallel across technology stack upgrades, product line expansion, and multi-scenario validation.

In July, Zhiyuan released its open-source operating system Lingqu OS, followed in August by the industry’s first open-source robot world model platform, and in September by the open-sourcing of GO-1, a general-purpose embodied foundation model. Through progressive open-source releases, Zhiyuan sought to attract global developers and establish a shared technical foundation. It subsequently launched the world-model platform GenieEnvisioner and the no-code creation tool Lingchuang, significantly lowering development and application barriers and broadening ecosystem participation. On the product side, Zhiyuan formed six major series covering diverse scenarios: Yuanxing (industrial and special-purpose), Lingxi (interactive), Jingling (task execution), D1 (quadruped), Juechen (commercial cleaning robots), and OmniHand (dexterous hand systems).

On the commercialization front, Unitree continued to explore full-spectrum deployment spanning consumer, entertainment, and industrial use cases, laying a solid foundation for long-term growth. In the first half of the year, Unitree entered a strategic partnership with Giant Star Group, a company associated with pop icon Jay Chou, and in November the two parties established a joint venture, Yuxing Entertainment, with equal ownership. The collaboration deepened Unitree’s presence in the entertainment sector. In December, Unitree’s humanoid robot performed high-difficulty aerial flips at a Leehom Wang concert, with the footage later reposted by Tesla CEO Elon Musk, generating significant global exposure.

In industrial applications, Unitree partnered with companies such as CATL and Baotong Technology, deploying humanoid and quadruped robots in scenarios including material handling, equipment inspection, and mining site patrols. Publicly disclosed metrics—such as over 40% efficiency gains and cost reductions of up to 70%—highlighted tangible progress in real-world deployment.

At the end of December, Unitree opened its first national offline flagship store at JD Mall in Beijing, complementing its existing official online store and completing an online–offline integrated sales loop. Collectively, these moves illustrate Unitree’s trajectory: starting from hardware-centric technical strength, leveraging high-scarcity, top-tier exposure events to achieve mass visibility, and gradually transitioning toward ecosystem-oriented operations and a closed-loop, omnichannel commercialization model.

Zhiyuan Robotics advanced commercialization across multiple dimensions—including application scenarios, sales channels, customer segments, ecosystem partnerships, and R&D—adopting a high-intensity, top-down execution strategy to accelerate real-world deployment of humanoid robots and to build an embodied intelligence industry ecosystem.

On August 18, 2025, Zhiyuan launched its full robot product lineup through its proprietary online store and JD.com, releasing nine models spanning eight application scenarios, including reception and guidance, entertainment and live performances, industrial manufacturing, and logistics sorting. Moving away from its earlier focus on purely industrial use cases, Zhiyuan adopted a “start easy, scale hard” commercialization strategy: prioritizing lower-complexity scenarios such as entertainment, exhibitions, and research/education to validate product capabilities in controlled environments, before progressively extending into more demanding industrial and logistics settings. This approach reduced trial-and-error costs and de-risked large-scale deployment.

In the industrial segment, Zhiyuan secured scaled orders from leading enterprises such as Fulin Precision and China Mobile, positioning itself as a solutions provider embedded in core processes like precision manufacturing and telecom services. These contracts not only implied meaningful revenue potential, but also served as validation of product reliability in complex environments. At the same time, certain transactions—particularly those involving investor-related parties (e.g., Longcheer Technology)—prompted market debate regarding order independence, with some observers questioning potential related-party dynamics.

Within commercial services, Zhiyuan accelerated penetration into research/education and entertainment scenarios, supported by a series of livestreams and promotional demonstrations. Notably, its Yuanxing A2 robot completed the world’s first 24-hour outdoor walk and a 100-kilometer inter-provincial journey, earning a Guinness World Record. These events functioned both as market education and as high-visibility proof points of product performance.

To transition from single-order delivery to scaled market penetration, Zhiyuan placed strategic emphasis on building platformized content and interaction capabilities. Its no-code Lingchuang platform allows users to generate robot actions by simply uploading videos, significantly lowering content production barriers for scenarios such as live performances and in-store services. Complementing this, the Lingxin platform—designed to endow robots with personalized voices and memory—targets interaction-intensive consumer-facing applications including guided tours and companionship. Together, these platforms form the “soft engine” of Zhiyuan’s business model, enabling repeatable deployment and ecosystem-driven expansion beyond hardware sales.

Building on this foundation, Zhiyuan Robotics, in collaboration with industry partners, launched the “Qingtian Rent”robot leasing platform. The platform deeply integrates skill packages generated through tools such as Lingchuang, enabling customers to rent humanoid robots and bundled services on a scenario-by-scenario basis, much like leasing conventional equipment. This model supports flexible deployment across use cases including shopping mall promotions, live event performances, and hotel services, lowering upfront adoption costs while accelerating market penetration through service-based delivery.

This development marks Zhiyuan’s transition from a pure humanoid robot hardware manufacturer toward an ecosystem platform operator that connects hardware, content, and end-user demand. By adopting a Robot-as-a-Service (RaaS), asset-light model, Zhiyuan is positioned to rapidly penetrate a wide range of industries, while reshaping the rules for scalable, scenario-based deployment of humanoid robots in the mass market.

III. Endgame Scenarios: Convergence or Divergence?

From the silent positioning and implicit benchmarking of the first half of the year to the year-end disputes over “shipment leadership,” culminating in contrasting moves around the Spring Festival Gala stage, the rivalry between Unitree Robotics and Zhiyuan Robotics reached a visible peak. This was not a superficial war of words, but a public confrontation rooted in fundamentally different organizational genes and strategic philosophies—an open contest over how the future industry ecosystem should be structured and defined.

Against the backdrop of persistent “bubble” skepticism, unresolved technical bottlenecks, and uncertain commercialization pathways across the sector, Unitree chose a strategy of inward consolidation. By repeatedly emphasizing sustained profitability, adhering strictly to the conventional IPO process, and leveraging scarce, high-impact events—such as top-tier concert performances—to build brand authority, Unitree positioned itself as a disciplined hard-technology practitioner. Its core objective has been to strengthen fundamentals within existing institutional rules, compete for the benchmark status of a “first listed humanoid robotics company,” and secure predictability and downside protection in capital markets.

Zhiyuan, by contrast, adopted an outward-expansion strategy. Through acquiring a listed company as a capital platform, co-establishing funds and joint ventures with industry leaders, and accelerating ecosystem formation, it sought to rapidly weave a network with itself as the central hub. Initiatives such as open-sourcing its operating system, launching the no-code Lingchuang platform, and introducing the Qingtian Rent leasing model were all aimed at lowering industry entry barriers. Collectively, these moves reflect Zhiyuan’s ambition to position itself as industry infrastructure and a standards contributor, rather than merely a hardware supplier.

Importantly, this competition is not a zero-sum game. The divergence between Unitree and Zhiyuan represents parallel explorations within China’s embodied intelligence sector. Unitree has focused on pushing the physical performance limits and cost baselines of humanoid robots, effectively strengthening the industry’s “body.” Zhiyuan, meanwhile, has concentrated on constructing the “neural network” of future intelligence—platforms, ecosystems, and scalable business models.

Whether competing for national-stage visibility or debating shipment metrics, each round of confrontation has contributed to market education, boundary expansion, and expectation recalibration. Together, these dynamics are accelerating the sector’s transition from technical demonstration toward commercial reality. The most likely endgame may not be the dominance of one player over the other, but the emergence of a layered, collaborative industry ecosystem, shaped through ongoing competition, differentiation, and eventual convergence.