Author | Hao Xin

Editor | Wu Xianzhi

When discussing the booming field of embodied intelligence, two companies cannot be overlooked—Unitree and Zhiyuan.

Both are valued in the tens of billions, leading the way in the IPO process. Unitree has raised over 1.5 billion yuan in cumulative funding, with a valuation of 12 billion yuan after its Series C round in 2025. Zhiyuan has completed 11 rounds of financing, reaching a valuation of 15 billion yuan as of March last year.

From more meaningful metrics—shipment volume and revenue—Zhiyuan and Unitree have formed a “duopoly,” dominating most of the current domestic market share.

According to official and third-party statistics, both Unitree and Zhiyuan generated sales revenue exceeding 1 billion yuan last year, with shipments of humanoid robots reaching approximately 5,000 units.

One investor described the two companies to us like this: “Unitree understands the present of embodied intelligence; Zhiyuan envisions its future.”

This investor's remark reflects two current strategies: Unitree has flourished in frontline markets with its sales network, while Zhiyuan is leveraging capital to build an industrial cluster for embodied intelligence.

The “present” of embodied intelligence means products achieving a commercial closed loop, with robots entering industrial production lines. The “future” means building technological ecological barriers and reshaping the relationship between humans and robots.

As the excitement of the Spring Festival fades, this track (track/field) has never lacked conceptual hype—what it lacks is patience for industry-specific implementation.

The Capital Integration Game

Capital is the biggest driving force behind embodied intelligence, but when it comes to companies skilled at playing the capital game, Zhiyuan stands out.

Last year, Zhiyuan stunned the industry by using 2.1 billion yuan to acquire a controlling stake in Shangwei New Materials.

Traditionally, a tech company's growth path runs from angel round to IPO, but Zhiyuan, between its Series B and C rounds, directly used money raised from the primary market to buy a controlling stake in an A-share listed company on the secondary market.

Zhiyuan's brilliance lies in avoiding “backdoor listing” suspicions while enjoying the capital advantages of being a publicly traded company. Before its official Hong Kong IPO, it can already use Shangwei New Materials for acquisitions and financing. Zhiyuan has elevated capital operations to a new level—essentially using the secondary market platform for primary market integration to boost secondary market valuation.

But Zhiyuan, well-versed in capital operations, didn’t stop there.

In June last year, Zhiyuan and Hillhouse Capital established a billion-yuan-scale embodied intelligence industrial investment fund. They also announced that over the next three years, through “Project Zhiyuan A,” they would invest billions to incubate over 50 high-potential early-stage projects and build a trillion-yuan industrial ecosystem.

According to incomplete statistics from Photon Planet, Zhiyuan is currently associated with 53 companies. Among them, it has four wholly-owned subsidiaries, three internally incubated and independently financed companies, ten joint ventures with other manufacturers, and 24 equity investment companies.

Through Zhiyuan, we can see capital strategies employed by many major firms. With Zhiyuan Robotics as the core, it integrates technology, manufacturing, scenarios, and capital through four layers—controlled incubation, joint ventures, equity participation, and industrial funds—to build an embodied intelligence industrial cluster.

Through joint ventures, such as partnerships with Lens Technology and LY Technology, Zhiyuan gained mature consumer electronics manufacturing capabilities and supply chain management systems. This allows it to consider mass-production processes during product design, shortening the cycle from prototype to product. Such exclusive partnerships provide an edge over competitors in ramping up production.

By taking equity stakes in companies like HuiXi Intelligence and Qianjue Robotics, Zhiyuan established a complete technological matrix covering the “brain,” “cerebellum,” senses, and application scenarios. Since core components like dexterous hands and joint modules are provided internally or by equity-linked companies, software-hardware collaborative optimization (collaborative optimization) efficiency is higher than competitors relying on outsourced assembly.

Notably, Zhiyuan has internally incubated three subsidiaries: Shanghai Critical Point Innovation Intelligence Technology Co., Ltd., Shanghai Mifeng Embodied Intelligence Technology Co., Ltd., and Qingtian Lease (Shanghai) Technology Co., Ltd.

The first two focus on dexterous hands and data platforms—bottleneck businesses in the embodied intelligence track (field). Qingtian Lease targets the most profitable business at this stage—robot leasing—to expand Zhiyuan's sales market.

Additionally, Zhiyuan has brought in industrial shareholders like SAIC, BYD, and Longcheer, directly tying clients into a community of shared interests. Thanks to this, SAIC’s factories became testing grounds for Zhiyuan robots, accumulating vast scenario feedback data and practical experience. Through joint ventures, Zhiyuan has also pre-locked benchmark client (benchmark clients) in core sectors like automotive and 3C manufacturing for embodied intelligence.

The Vanished Industrial Closed Loop

While capital moves aggressively, the market tells a different story.

Currently, robot sales are tiered: at the top are original equipment manufacturers (OEMs) like Unitree and Zhiyuan, which hold pricing power and channel management rights and can directly sign large client projects; the second tier consists of regional agents, including provincial and municipal representatives as well as industry-focused agents, who generally possess local client resources; the third tier comprises industry solution integrators, which procure robots from OEMs to create integrated hardware-software solutions sold to end clients; the final tier is leasing service providers, whose core business model involves bulk purchases from OEMs and leasing to small and medium-sized clients under subscription-based, time-based rental agreements.

Data from robot bidding projects in 2025 shows that billion-yuan-scale large orders are concentrated among government, enterprises, and university research institutions. Practitioners in embodied intelligence note that government mega-projects are usually placed directly with OEMs, accounting for nearly half of sales revenue—profits that regional agents cannot access.

“Most agents aren’t making money. The commercialization of embodied intelligence is still in its early stages, and all signs indicate it hasn’t penetrated across industries yet.”

Even orders can be deceptive.

According to industry insiders, “ambiguous nature” is the most common form of inflation.

For example, Company A signs a strategic cooperation agreement with Company B, agreeing to “purchase XX robots over the next three years.” Press releases hail this as a major order, but legally, it’s merely an intention. Without breach clauses, promised orders can be canceled if market conditions change—more like a worthless check that cannot be counted as real revenue.

Another common scenario involves resource swaps. Many robot purchasers already have capital and strategic partnerships with the backing companies. This can lead to mutual “pumping”—investors, seeking impressive order growth or joint benchmark cases, choose to purchase robots.

When local state-owned assets introduce companies, they often promise “first purchase, first use” to keep the firm local, creating jobs and tax revenue. Whether the purchased robots perform well is a matter for later operations.

The most critical issue is that capital-driven embodied intelligence has generated countless false demands.

As capital flows into embodied intelligence, a wave of companies has emerged. Many robotics firms start with financing plans, then deduce required order volumes and revenue targets before seeking clients in the market. This results in significant wasted capacity and R&D, targeting capital market preferences rather than industrial pain points from the outset.

Fueled by capital, the outcome is clear: robots perform well in demos but struggle when deployed in real industrial environments.

You’ll notice that the industrial closed loop in the embodied intelligence sector has vanished.

A healthy industrial closed loop typically starts with real demand, leading to products or solutions that address pain points, achieve cost reductions and efficiency gains, generate profits, and then reinvest in R&D to create more demand.

However, in today’s embodied intelligence field, this loop has been distorted into one driven by capital narratives: “inflated” orders are fabricated out of thin air, driving capacity expansion and valuation increases to complete the next funding round, cash out, and support even grander capital stories.

Too Much and Too Little

Setting aside capital bubbles, robot leasing is currently the primary revenue source across the entire track (field).

Theoretically, once a company launches a product, it can generate revenue through direct sales and agent channels.

For example, according to its official website, Zhiyuan offers the most comprehensive product line, including wheeled-legged, humanoid, quadrupedal, and commercial robots, covering nearly all scenarios.

But market feedback tells a different story. A robot agent told us, “Unitree robots are the most common in the market; Zhiyuan’s are relatively rare. Some manufacturers are only known by name and have never been seen in person.”

Robot leasing and maintenance costs are becoming a burden for agents.

We learned that Unitree, having expanded its sales network earlier, has more leasable robots in the market.

A half-size Unitree G1 robot rents for 4,000–5,000 yuan per day; the rarer Zhiyuan robots command higher prices—a half-size Zhiyuan X2 rents for 6,000–8,000 yuan per day. Zhiyuan’s full-size general-purpose humanoid robots are rarely seen in the market, with its Expedition series renting for at least 10,000 yuan per day.

“Leasing” robots doesn’t end with shipping them out.

The robot agent cited the common scenario of opening ceremonies: for a typical 2–3-minute performance, they need at least two staff on-site for the entire event—one to operate the robots and one for maintenance.

Robot maintenance is a significant expense. For such events, robots typically need to return to the factory for servicing after every two or three performances.

Even during the manufacturer’s promised one-year warranty period, agents still bear maintenance costs if robots malfunction.

We’ve seen a maintenance bill for a Unitree robot: despite only having scratches on the protective shell and normal wear and tear, the total cost exceeded 5,000 yuan.

The recent Spring Festival Gala acted as a “magnifying glass” for capital. Embodied intelligence companies spent hundreds of millions of yuan to showcase robots performing martial arts, street dance, and flips, solely for exposure to drive new funding rounds.

Behind the audience’s applause were pre-programmed routines, fixed routes, countless rehearsals, and perfect performance environments. Those robot “stars” didn’t need to avoid obstacles, handle emergencies, or face extreme industrial environments like factories and construction sites.

Shortly after the Spring Festival holiday, financing in the embodied intelligence track (field) surged wave after wave.

According to incomplete statistics, eight domestic humanoid robot companies have now entered the 10-billion-yuan valuation tier, including UBTECH, Zhiyuan Robotics, Unitree Technology, Yinhe General, Xinghaitu, Zibianliang, Zhipingfang, and Qianxun Intelligence.

This track (field) has too many—and too few—players.

Too many are eager to spin tech-driven stories; too few are willing to refine products in factories.