At the beginning of 2026, the embodied intelligent track will usher in an unexpected and reasonable "hurricane".

On March 2, according to people close to Galaxy GM, Galaxy General Robots has completed a new round of financing of 2.5 billion yuan. On the same day, the humanoid robot star company "Songyan Power" announced the completion of the B round of financing, with a cumulative scale of nearly 1 billion yuan. In just two months since the beginning of this year, companies such as Qianxun Intelligence, Independent Variable Robot, Zhisquare and Xinghaitu have successively announced financing. According to public data, since the beginning of 2026, the total amount of financing disclosed in the field of embodied intelligence has been close to 15 billion yuan.

This is a capital feast and a new round of market wars.

Recently, I participated in the theme interview of embodied intelligent entrepreneurs planned by CANPLUS and "Daily Economic News", and had an in-depth conversation with 6 alumni of the China Robot Industry Ecological Camp: Zhao Guangzhi, co-founder of Luming Robot, Liu Yang, co-founder of Force Infinite, Xu Xiaomin, co-founder of Amio Robotics, Bai Guochao, founder of Changingtek Robot, Liu Dong, founder of Xingyuanzhi, and Ji Chao, CEO of Lingdong General Motors and chief scientist of iFLYTEK Robotics. As entrepreneurs on the outlet, some of them have just released the world's first adaptive left-handed dexterous hand; some financing is fast, winning 200 million yuan in financing in the first month of establishment; and others are capital darlings who won four rounds of financing in a row last year...... In the capital frenzy, they are both witnesses and witnesses.

Interviewed China Robot Industry Ecological Camp Alumni Company Image source: Reporter Zhao Wenqi Mapping

"The capital market is still enthusiastic about embodied intelligence, and it is also betting on real money." Xu Xiaomin, co-founder of Amio Robotics, told reporters, "Most embodied intelligence companies are not short of money, mainly the mentality of 'hoarding money', which shows that everyone generally recognizes the huge space of this track and bets on longer-term goals." ”

Liu Yang, co-founder of Force Unlimited, summed up the signals conveyed by this round of financing boom in three words: consensus acceleration, intensified differentiation, and narrowing of windows.

In the interview, almost all founders mentioned that 2026 is no longer the year of competing for demos (product samples), but a thrilling leap from "running through the laboratory" to "practical scene landing". In this capital frenzy, even if the camp is strong, no one dares to take it lightly. Some people are hoarding food to prepare for war, some are anxious to fall behind, and the spotlight of capital is converging towards the head at a speed visible to the naked eye.

The embodied intelligent "multi-hero" competition officially fired the gun.

The "anxiety" of entrepreneurs: like the takeaway "100 reunion war" ten years ago

"Last year it was 'a hundred flowers blooming', and this year's more accurate description is 'a hundred rivers returning to the sea'." Liu Yang's words show the most essential changes in the embodied intelligent track since the beginning of 2026. If 2025 is the year of the industry's "show muscles" - companies are vying to release stunning demos such as bipedal walking and backflips, then there is only one keyword for 2026: delivery.

This step will become the first life-and-death hurdle in front of startups.

Because of this, in Liu Yang's view, this round of hot money is not "general rain", but "targeted irrigation". Financing is accelerating the concentration of leading enterprises, and enterprises that have received billion yuan of financing either have a clear closed loop of commercialization or have built irreplaceable barriers in core technologies such as "brains". The financing window of mid-waist enterprises is narrowing at a speed visible to the naked eye.

"Last year, everyone walked more steadily, and this year it is more about who can do it longer." Liu Dong, founder of Xingyuanzhi, told reporters that the core challenge facing startups is to cross the gap from "laboratory run-through" to "practical scenario implementation". "Last year's demo, this year investors will ask you very realistically about your landing scenario and conversion rate, if there is no answer, no matter how cool the technology is, it is difficult to continue to transfuse."

"In 2025, a large number of so-called 'commercial orders' in the industry will essentially be public relations display procurement and data collection cooperation, rather than a real productivity substitution. The cruelty of 2026 is that industrial customers are no longer satisfied with 'seeing your robot do somersaults'. What they want to see is: Can your robot reduce costs and increase efficiency after running in my factory for three or six months? Can repurchase occur? So far, frankly, no company in the industry has completely passed this 'proof-of-concept' stage. This is a life-and-death line. Liu Yang said.

"Shuqiao" dexterous hand Image source: Changingtek Robot official website

At present, although the capital heat in the field of embodied intelligence in China is high, the hidden worries of the industry have emerged. Liu Yang bluntly said that there are currently more than 200 physical intelligent enterprises in China, including more than 100 humanoid robots, but research institutions have clearly observed that "some companies are actually no longer good, but they are still operating at low speed." He believes that just like the takeaway "Hundred Regiments War" ten years ago, only a few that are commonly used by everyone have survived in the end.

Xu Xiaomin also observed that there were many bubbles in the commercialization of the industry last year, but now the wind direction has changed: "Now most companies have entered the deep water area to explore model capabilities, and more time is needed for general capabilities." However, the commercial closed-loop and industrial landing capabilities are better judged at present, and this year everyone will pay more attention to real orders and mass production and landing capabilities. ”

When orders and mass production become the key words, it is no longer only the advancement of the algorithm, but also the maturity of the entire supply chain system.

Bai Guochao, founder of Changingtek Robot, said frankly that this year's leading enterprises will face three major challenges: mass production cost and yield control, scenario profit model running, and independent and controllable core components. In his view, the focus of industry competition has also shifted from prototype competition to the showdown of key components such as core actuators and dexterous hands, scene landing efficiency and supply chain integration capabilities.

In the view of Ji Chao, founder and CEO of Lingdong GM and chief scientist of iFLYTEK Robotics, in the future, the head and central enterprises may be differentiated in the core of competition, and the actual technology gap between enterprises may be the focus of everyone's controversy.

Zhao Guangzhi, co-founder of Luming Robot, told reporters that in the past year, the embodied intelligence track has continued to make progress at the model, hardware and application levels, and a consensus is gradually forming in the industry: what really limits the further development of embodied intelligence is real, reusable data.

"Entering 2026, data problems have become an unavoidable core bundle when embodied intelligence expands the boundaries of capabilities. 2026 will be a year of large-scale accumulation of embodied data and market explosion. Zhao Guangzhi said.

Capital focus shift: "brain" has become the core battlefield of embodied intelligence in 2026

When the industry enters the stage of "whether it is good or not", the division of technical routes begins to become a key variable that determines the fate of enterprises. A phenomenon worth paying attention to is that the new physical intelligent 10 billion yuan-level unicorns around the Spring Festival all belong to the direction of "embodied brain", which marks that the focus of capital has shifted from "ontology" to "brain".

Liu Yang bluntly pointed out to reporters the "embarrassing truth that is not often discussed" in the industry: "The vast majority of robot ontology companies do not have the ability to independently develop large models. The training cost of basic large models is too high, which is not something that a robot start-up can cover by investing hundreds of millions of yuan or billions of yuan. Therefore, you will see that many ontological enterprises can only rely on the general large models of technology giants such as Alibaba and Byte to empower. But the question is, when the big technology companies personally end up making robots, what will these companies that 'borrow their brains' to survive? ”

This difference in route essentially corresponds to different judgments on "intelligence sources": whether to mount a general large model to "borrow wisdom", or to train an embodied native model from scratch? Is it to synthesize data in a simulation environment to accelerate iteration, or to collect data in real scenarios to build a flywheel?

Force Infinite Choice adheres to the full-stack self-developed embodied brain. "The deep integration of our Hyper-VLA end-to-end multimodal large model and causal world model is not grafted on someone else's pedestal, but is completely autonomous and controllable. On February 25, we officially announced that Wang Yizhou, a former founding member of NVIDIA DriveAV and a doctor from the University of Berkeley, became CTO (Chief Technology Officer), forming a dual engine of 'cybernetics + large model' with the cutting-edge large model that our young scientist Chen Jiayu is focusing on. Liu Yang believes that in 2026, the core battlefield of embodied intelligence will focus on three key dimensions: who has an independent and controllable "brain", who masters the largest scale of real scene data, and who can take the lead in running through the business closed loop from "proof of concept" to "repeated procurement".

In Liu Yang's view, the PC (personal computer) era was Intel as a chip and Microsoft as an operating system; In the era of smartphones, Foxconn did hardware, and Apple and Google did iOS and Android. "In the era of embodied intelligence, the ontology (the shell of the robot) will eventually be leveled into a standard product by China's strong supply chain capabilities, and whoever masters the universal embodied brain will have the final say in this trillion-level ecology, which is also the ultimate bet of Force Infinity." Liu Yang said.

Image source: Xingyuanzhi official website

"For Xingyuanzhi, this year will be a key year for us to widen the gap. We firmly believe that when the industry moves from hardware to intelligence, whoever can provide the smartest, most stable and most understanding of the physical world will define the next generation of robots. Liu Dong said.

Ji Chao believes that with the continuous evolution and falsification of technical architecture, enterprises will pay more attention to data this year, and related solutions will also bloom, and the perception of "the importance of data" has basically formed a general consensus in the industry. The main challenge may be how to choose the appropriate technical path to truly verify the scaling law of the data to the embodied model, and to build the company's own embodied brain, which may also be the key work of most companies in recent years, and the importance of the business model will also increase.

Bai Guochao mentioned that since the beginning of this year, customers have shifted from simple consulting to batch orders, the efficiency of industrial collaboration has been greatly improved, and the market demand for the autonomy of core components is unprecedentedly urgent. Under the boom, the industry needs to return to its hard-core essence. Changingtek Robot will continue to focus on the R&D and mass production of dexterous hands and core actuators, support the scale of the industry with independent and controllable technology, and stand firm on the track and make real value in this round of critical cycle.

2026 is not the first year of commercialization, but the first year of the knockout round

"2026 is not the 'first year of commercialization', but the superposition of 'the first year of commercialization verification' and 'the first year of knockouts'." For entrepreneurs at the center of this storm, Liu Yang's judgment has aroused widespread resonance in the industry.

From the end of 2025 to the beginning of 2026, a number of overseas star robot companies have fallen into an existential crisis: Silicon Valley company K-Scale Labs dissolved due to running out of funds on the eve of mass production, collaborative robot pioneer Rethink Robotics suffered a second "death", and iRobot, the originator of sweeping robots, filed for bankruptcy protection.

Xu Xiaomin defines 2026 as a "big test": "For companies with clear vertical scenarios, 2026 is a big test. Most companies do not lack customers and orders, and the key is whether they can run an effective technical route and polish a complete delivery system. ”

Ji Chao analyzed the possible differentiation of the industry from a more macro perspective: "Due to the different ecological niches of each company in the track, there may be a more differentiated development path in 2026." Commercialization is an important part, but it may not be the goal of all embodied companies this year. For most companies, it may be more important to make good use of existing resources and adjust their business models to amplify their own advantages. The market will be more carnival and more rational. ”

Liu Dong predicts that there will be significant polarization in 2026: "Industrial and special scenarios will usher in small-scale outbreaks. In these scenarios, customers have a strong willingness to pay, strong generalization ability, high reusability, and high fault tolerance rate, and we will see the birth of 'hidden champions' in some vertical fields. However, in commercial and home scenarios, there is still the 'last ten meters' to go in commercialization, and it will take time for 'robots to enter the home on a large scale'. ”

In his opinion, a bigger watershed in the future lies in the presence or absence of scene data. Enterprises that only show technology in the laboratory and do not bind real physical scenarios will feel the chill of financing in the second half of the year. Because capital will find that these technologies are just castles in the air.

Image source: Force Unlimited official website

After the hustle and turn, what's really important?

Liu Dong believes that embodied intelligence is taking off its scientific and technological clothes and putting on a "battle robe" to the battlefield. The excitement of the Spring Festival Gala will eventually dissipate, and the enthusiasm of capital will return to calm. The last ones left at the table must be those robots that really work in workshops, warehouses, and all walks of life.

"We actually judged internally very early that the competition for embodied intelligence will eventually fall on the data infrastructure. Models can be iterated and algorithms can be upgraded, but once the real-world data system is established, the accumulation is irreversible. Whoever can consistently and stably produce high-quality real machine data will have a better chance of pushing the model capabilities to a higher limit. Zhao Guangzhi said.

In Zhao Guangzhi's view, this year's pursuit of the entire industry is to truly scale the data of real scenarios, and open up the link of data production, quality evaluation, training and transformation. Real data is not a supplementary variable, but a core variable that determines how far embodied intelligence can go. In the next few years, what can really widen the status gap for embodied intelligence companies may be "whose data is more real, more stable and more scalable".

The 2026 "Duoxiong" competition has begun. Whoever can be the first to cross the gap from the "laboratory" to the "training ground" will be qualified to define what the next generation of robots will look like. And the answer lies in the day and night operation of every real scene.