Unitree has proven itself. Now it's the turn of Zhiyuan.

Written by | Huahua

Recently, ZHIYUAN ROBOTICS held an industry conference of an unusually high standard.

All 2,500 seats were taken, gathering customers, suppliers, and investors from 34 countries and regions. Inside the conference venue in Zhangjiang, Shanghai, simultaneous interpretation in English, Japanese, Korean, and Arabic was available.

This kind of grand occasion has always been exclusive to the annual galas of leading Internet giants and is hard to associate with a robotics startup that has been established for only three years.

At this conference, the founder, Deng Taihua, did something extremely rare in the industry: he made the company's core operating data public without reservation.

In 2023, ZHIYUAN's revenue was 300,000 yuan; in 2024, it soared to 60 million yuan; in 2025, it directly exceeded 1.05 billion yuan.

In just three years, the revenue achieved a nearly 20-fold crazy fission every year, and the growth rate can be called a miracle in the industry.

Immediately afterwards, Deng Taihua presented a more shocking five-year plan: In 2026, the revenue will achieve several-fold growth; in 2027, it will exceed 10 billion yuan; in 2030, it aims for 100 billion yuan.

A company that has been established for only three years, in an emerging track that has not fully exploded, is so high-profile in shouting out the goal of 100 billion yuan in revenue. This kind of high-profile behavior that almost overdraws the future has never been seen in ZHIYUAN's past public records and is also very rare in the entire embodied intelligence track.

However, if we look back at a key node 28 days ago, this high-profile behavior of ZHIYUAN makes sense.

On March 20th, Unitree Robotics submitted a prospectus for the Science and Technology Innovation Board to the Shanghai Stock Exchange.

This document completely changed the rules of the game in the embodied intelligence track.

I. Two Completely Different Report Cards

Unitree's prospectus is the first auditable commercial evidence submitted by the embodied intelligence industry.

In 2025, its revenue was 1.708 billion yuan, a year-on-year increase of 335%; the net profit after deducting non-recurring gains and losses was 600 million yuan, a year-on-year increase of 674%. It shipped over 5,500 humanoid robots and nearly 20,000 quadruped robotic dogs. What's even more enviable is that the gross profit margin of its humanoid robot business was as high as 62.9%.

These figures prove one thing: Unitree made money.

In an industry where almost all players are burning money, Unitree turned a profit as early as 2024. In 2025, its profit directly exceeded 600 million yuan, and its R & D expense ratio was only 7.7%, far lower than the industry average of 28%.

The core reason for achieving this virtuous cycle is that Unitree adheres to self - developed motors and reducers, controlling the BOM cost of its products at a level that competitors can hardly reach.

Unitree's development logic is: First, make the products cheap and practical, achieve large - scale mass production and profit, and then gradually iterate and optimize to improve the intelligence level of the products.

On the contrary, this is the development logic of ZHIYUAN ROBOTICS.

Deng Taihua, the founder of ZHIYUAN ROBOTICS, was once the vice - president of Huawei and the core operator of Ascend AI business. When he founded ZHIYUAN in 2023, he brought not only management experience but also a complete set of Huawei - style industrial thinking: first, build an ecosystem, lay out a platform, and invest in AI, and then reduce the cost through scale.

More than three - quarters of ZHIYUAN's R & D resources are invested in AI. In less than three years since its establishment, it has signed more than 17 joint - venture companies, invested in more than 15 early - stage enterprises, and released an open - source operating system and a developer platform. The fame of CTO Zhong Huijun is a halo, but Deng Taihua's large - platform approach is the framework.

ZHIYUAN's logic is: First, make the robot smart, and then think about making it cheap.

In the past two years, the two paths represented by Unitree and ZHIYUAN have been advancing in parallel. One starts from engineering, focuses on the robot body, mass production, and finding application scenarios, turning robots into products that can be bought, used, and replicated. The other starts from the system, builds platforms, creates ecosystems, and conducts integration, regarding robots as an entrance to general intelligence.

Each goes its own way, and there's no need to compare scores.

However, when Unitree put its prospectus on the table, the parallel paths intersected.

Once someone starts to report numbers, everyone will be put in the same coordinate system. You can tell stories at the partner conference, but the prospectus needs to pass the audit of the CSRC. You can shout out the goal of 100 billion yuan, but the numbers in the profit statement won't lie.

Unitree's profit of 600 million yuan and a gross profit margin of 62.9% have given the entire industry a ruler to measure everyone for the first time.

ZHIYUAN's revenue of 1.05 billion yuan is not low. It is the fastest robot company in China to reach a revenue of one billion yuan. However, Deng Taihua himself admitted that the company is still in the red, and the inflection point of profitability depends on the scale effect to dilute R & D costs. Not being eager to pursue net profit actually means that it is still burning money and needs time.

The problem is that Unitree's prospectus has compressed the time ZHIYUAN needs.

II. Who is the First?

Before the two report cards came out, there was a small episode about the ranking.

In January this year, market research firm Omdia released a report, ranking ZHIYUAN first in the global humanoid robot shipments in 2025, with 5,168 units, accounting for 39% of the global market. Unitree ranked second, with 4,200 units. ZHIYUAN's official WeChat account quickly reposted this report.

Unitree couldn't sit still. They issued a statement directly stating that the information circulating online was false. Later, Unitree provided its own figures in the prospectus, stating that it shipped over 5,500 humanoid robots in 2025. (Extended reading: Unitree, Refusing to Be Second)

This is more than the 5,168 units Omdia reported for ZHIYUAN. Moreover, Unitree pointed out a key detail: ZHIYUAN's statistics included wheeled and double - armed robots, not all humanoids.

On the surface, this dispute is about the difference in statistical caliber, but the subtext is that in this fledgling and wild market of embodied intelligence, the signboard of "who is the first" is more valuable than the numbers themselves. It directly affects the financing narrative, IPO valuation, and the confidence of partners.

For ZHIYUAN, there is a deeper sense of urgency in this ranking battle. Unitree's 5,500 shipments came with a profit of 600 million yuan, which means it makes money for every robot sold. For ZHIYUAN's 5,168 units, even if the caliber is questionable, there is continuous loss behind it.

With the same shipment scale, one is self - sustaining, and the other is bleeding.

After Unitree put this gap on the table with its prospectus, ZHIYUAN needs to come up with an equally powerful narrative to respond.

This conference 28 days later is the response.

III. The Huawei Approach: A Lever and a Burden

ZHIYUAN's conference is not just about showing off numbers. Deng Taihua did something more ambitious. He tried to redefine the coordinate system of the entire industry.

On - site, ZHIYUAN proposed an "XYZ Curve." From 2022 to 2025 is the "technology verification period," during which robots are solving the problem of whether they can move; starting from 2026, it enters the deployment and growth period, and robots need to solve the problem of whether they can work; after 2030, it is the productivity explosion period, and the trillion - dollar market will open up.

The subtlety of this framework lies in that it repackages ZHIYUAN's current state of high revenue but no profit as the inflection point when the industry has just moved from the X - curve to the Y - curve. In other words, the losses seen by the outside world now are the entry cost for the next stage.

To prove that the "deployment state" is not just a concept, ZHIYUAN conducted an 8 - hour continuous live - broadcast three days before the conference. Four ZHIYUAN ELF G2 robots continuously worked on the tablet production line at the Longcheer Technology Nanchang Factory, completing 2,283 tasks with a 100% success rate.

However, Deng Taihua said something very honest: The emotional value comes quickly and may also go quickly.

He admitted that in the past two years, the shipment scenarios were mainly concentrated in scientific research, entertainment, and data collection. The 8 - hour live - broadcast proved that ZHIYUAN's robots can work on the production line, but there are still hurdles such as cost, battery life, and stability between being able to work and large - scale deployment.

ZHIYUAN's bet is not on today's production line but on that Huawei - style ecosystem layout. In less than three years since its establishment, there are 17 joint - venture companies, more than 15 investment projects, an open - source operating system, and a developer platform.

Deng Taihua is using the approach that Huawei has accumulated over two decades to develop robots. First, he spreads out the ecosystem, binds the upstream and downstream, forms a leading position in the industrial chain, and then crushes the cost through the scale effect.

This approach has been verified in Huawei's hands.

However, there is a prerequisite. Huawei first proved its technological leadership and commercial reliability with its products (switches, base stations) before it was qualified to integrate the supply chain. ZHIYUAN's ecosystem is ahead of its products now.

For a three - year - old and still - losing company, the collaborative management of 17 joint - venture companies and the resource allocation of 15 investment projects are both a lever and a burden.

Unitree's path is completely different. From the first day of entrepreneurship, founder Wang Xingxing chose an outer - rotor brushless motor to replace hydraulic drive, a typical cost - reduction path in the manufacturing industry.

Over the years, Unitree has proved that robots can be manufactured cheaply and sold on a large scale like consumer electronics. Its quadruped robotic dogs account for nearly 70% of the global market share, and the stable cash flow supports the R & D investment in humanoid robots.

However, Unitree also has its own concerns. In the prospectus, more than 2 billion yuan out of the 4.2 billion yuan in planned fundraising is intended for the R & D of embodied large models. This is an open admission that the advantage of being cheap has an expiration date, and the core of the next - stage competition is to be smart.

Unitree is trying to evolve from a precision manufacturing company into an AI company, using the money earned in the manufacturing industry to buy a ticket into the AI field.

Both companies have gone halfway. Unitree has proved that it can make money but has not proved that it can be smart. ZHIYUAN has proved that it can grow large but has not proved that it can make money.

IV. It's Not Just About Two Companies

If we only focus on the tussle between ZHIYUAN and Unitree, we will miss a bigger picture. Embodied intelligence is entering a window period of intensive listings.

According to incomplete statistics, in 2026, more than 20 embodied intelligence companies have clarified their listing plans. Unitree is aiming for the Science and Technology Innovation Board, and ZHIYUAN is eyeing the Hong Kong stock market (according to previous media reports, the target valuation is HK$40 billion to HK$50 billion). Whoever completes the listing first will have an additional ammunition depot in the next round of competition.

Looking at Unitree's shareholder structure, we can find some interesting names. Wang Xing and the Meituan system hold a total of 9.65% of the shares, and Lei Jun's Shunwei Capital holds 4.42%. On ZHIYUAN's side, Tencent led the Series B financing, and JD.com, SAIC, BYD, and Hillhouse have all joined in.

The capital line - ups behind the two companies reflect two different jigsaw puzzles.

On Unitree's side, the Lei Jun system brings manufacturing experience and the consumer electronics supply chain, and the Wang Xing system brings local - life scenarios and distribution networks. Coupled with Unitree's own motion control and mass - production capabilities, these three pieces together answer the questions of where to use the robots and who will operate them after they are manufactured.

On ZHIYUAN's side, there is Tencent's cloud - computing and AI capabilities, JD.com's warehousing and logistics scenarios, and the manufacturing genes of SAIC and BYD. Together, they answer the questions of where the robot's brain comes from, where the body is manufactured, and from which scenarios to start.

The competition logic of robots is fundamentally different from that of pure software products. An app can achieve a single - point breakthrough, and a product can penetrate a certain type of users. However, robots are a systematic project. Without any one of hardware, software, scenarios, and service networks, it is difficult to achieve large - scale production.

This is why the competition in embodied intelligence may ultimately be between ecosystems rather than between companies.

Whoever can piece together the jigsaw puzzle the fastest will have an advantageous position in the next round.

The meaning of the word "fastest" is being redefined. It is no longer about showing off skills at the press conference and concept videos but about the revenue, profit, and shipments in the prospectus. When robots were just a concept, the narrative could define a company's position.

When robots start to become commodities, the delivery ability will re - rank everyone.

In this view, Deng Taihua's high - profile behavior comes at a cost.

In the technology - driven stage, a company can establish its position by talking about the vision, the system, and the future. Capital is also willing to pay for the ambiguity because the future space is large enough. However, once the industry enters the verification stage, the focus will shift. Whether the growth is sustainable, whether the model is valid, and whether the input - output ratio is predictable.

Numbers are the only things that can be repeatedly verified.

Unitree has changed the reference system of this industry with its prospectus. Before it, no one needed to answer the question of whether they were making money. After it, everyone has to answer.

[Beyond the Page] Words:

ZHIYUAN's goal of 100 billion yuan and the XYZ Curve are essentially trying to establish another reference system. Don't measure me by today's profit; look at the productivity explosion tomorrow.

Is this statement reasonable? Yes. Can it be fulfilled? It's still unknown.

ZHIYUAN's high - profile behavior is more like a head - start rather than confidence. Companies that originally relied on future - oriented narratives to build advantages now need to improve their numerical performance. And companies that have been focusing on product delivery have started