Vbot’s 500 Million Financing: Insights from Founders on Bringing Robots into Homes

500 million financing behind the scenes: A conversation with Yu Yinan and Zhao Zhelun about the true integration of robots into homes

On May 11, 2026, Yu Yinan looks both urgent and relaxed, a unique combination that sets Vbot apart. He feels urgency due to the advancements in embodied intelligence technology overseas, which compel Vbot to accelerate its R&D efforts. On the other hand, he feels relaxed because Vbot has successfully transitioned from showcasing consumer-grade embodied intelligence products to actual deliveries, positioning itself as a leading brand in discussions about “consumer-grade embodiment.”

Recently, Vbot completed a Pre-A round financing of nearly 500 million yuan, marking the largest financing to date in the consumer-grade embodied intelligence sector. This round was led by Dongfang Jiafu, Huatai Zijin, and Fosun Ruizheng, with follow-up investments from existing shareholders such as Gaohui Fund and Today Capital. Just a day before the interview, the first batch of 500 Vbot “Big Head” super dogs had rolled off the production line and were shipped to consumers nationwide. The first direct experience stores in Shanghai and Beijing have also opened their doors. As consumer-grade robots begin to truly enter homes, this moment is both exciting and nerve-wracking.

After more than a year of entrepreneurship, Yu likens this venture to a game of Go. His first move was crucial: to avoid remote control, not pursue humanoid designs, and not chase rankings. Over the past year, the quadruped robot, Big Head, has completed its journey from definition to mass production. The second phase is quietly taking shape; Vbot’s humanoid robot prototype is currently under development and is expected to be unveiled in August. Yu noted that the most significant change over the past year has been the creation of a new category that allows embodied intelligence to enter households. There is a notable shift in discourse, as more industry players are now asking, “If you want to target consumers, how do you differentiate yourself from Vbot?”

Here are the details of the conversation between Vbot founder and CEO Yu Yinan and co-founder Zhao Zhelun.

Reflecting on the Past Year of Entrepreneurship

Phoenix Technology: Over a year into your entrepreneurship journey, what has been your biggest challenge?

Yu Yinan: There have been many difficulties. Initially, we thought making a robot would be simpler than building a car. However, we discovered that developing a robot is even more challenging. The reason is that cars are large, with a mature supply chain and industry standards; they benefit from a century-old industrial foundation that allows for incremental improvements and innovations. In contrast, robotics starts almost from scratch. While cars have sensors, batteries, computing platforms, actuators, and motors, our robot dog has 14 motors and 14 degrees of freedom, alongside a sizable battery (though still smaller than a car’s) and a powerful computing platform with sensors. Fitting all these components into a compact form presents enormous challenges in product development and engineering. We’ve invested significant effort into creating joint modules that are small yet have high torque density and are lightweight, necessitating extensive product and engineering innovation.

Once we made everything compact, we also needed to ensure the robot exhibited strong mobility, which required a close coupling with software algorithms. We didn’t develop hardware first and then hand it over to software; we iterated from the model to hardware together to achieve the current performance.

Phoenix Technology: At this stage, which do you consider more important: modeling capability or engineering capability? What are your strengths?

Yu Yinan: We excel in both engineering and modeling capabilities. The product demands stringent modeling requirements. Unlike ranking pursuits, product modeling must address user needs and open usage scenarios, where the challenges are unknown. In contrast, ranking is akin to an open-book exam. The current state of the industry resembles the AI 1.0 era from 2012 to 2015, where everyone is chasing various benchmarks. Achieving high scores doesn’t necessarily represent true technological capability. During my early career, I engaged in numerous benchmarking activities and found that these scores often lack practical value.

Our contemplation revolves around a path problem, akin to playing Go: where you place your first move, establish your main territory, and infiltrate the opponent’s territory. This is more of a strategic issue.

Phoenix Technology: Is it possible that other companies find it easier to secure financing through ranking pursuits, or do they have different needs?

Yu Yinan: Indeed, pursuing rankings facilitates financing. This is a matter of choice; we opted for a direct approach to bring a terminal product to market in collaboration with users. Even without focusing on rankings, we still managed to secure significant financing. However, I believe the current investment levels in this industry are still not high enough. There’s a massive disparity between China and the U.S. For instance, while China may secure 2 billion RMB, North America might raise 2 billion USD. Furthermore, if we compare today’s financing amounts to those in other industries, a company in the embodied intelligence sector typically raises only 3 to 5 billion yuan. In contrast, electric vehicle startups usually begin with 5 billion yuan, and without 10 billion in the bank, one cannot sustain operations. Looking back even further, during the mobile internet era, burning several hundred million a day was standard. This industry requires more capital investment.

Phoenix Technology: Each company spends money differently. Now that you’ve raised funds, how do you plan to allocate them?

Yu Yinan: When things are unclear, it’s best not to spend recklessly. After a year of exploration, our technology is becoming clearer, especially as our data technology roadmap converges. We seem to see a pathway to the ultimate solution. Initially, remote control was the most straightforward method but proved inefficient and non-replicable. You cannot have a humanoid robot at a market, relying on one person to control it remotely daily; that approach simply doesn’t work. There were attempts to use grippers or gloves to collect data, but these methods fell short as well. The ultimate solution for robots has always been dexterous hands, capable of performing all human tasks. However, how to gather data using dexterous hands has yet to be solved. Recently, data collection gloves have advanced, but frankly, this approach may not be the ultimate solution.

Phoenix Technology: We’ve heard that a lot of collected data has a usability rate of less than 5%. Some also argue that third-person perspective data may not be as crucial.

Yu Yinan: Yes, a lot of money has been wasted. Currently, the popular data collection methods are primarily first-person perspectives. However, third-person perspectives are also essential because, from a first-principles standpoint, humans learn by observing others at work from a third-person viewpoint and then practice from a first-person perspective. These two perspectives need to be combined, and I believe technology is indeed converging.

Product Evolution Logic: From Model S to Model 3, From Ideal ONE to L9

Phoenix Technology: Vbot is about to deliver a large volume of machines to homes. The data you have may be unattainable for others, making it critical. Is the number of C-end users in this first batch in line with your expectations?

Yu Yinan: The market’s enthusiasm has surpassed our expectations.

Zhao Zhelun: Following our successful launch, many companies have initiated projects to create robotic dogs.

Phoenix Technology: Why do you think they reacted so quickly? What sparked renewed interest in robotic dogs?

Zhao Zhelun: The robotic dog category has existed in the industry for many years and is quite mature. There are fundamentally two standard forms of embodied intelligence: humanoid and quadruped. The consensus in the industry is that quadrupeds are more mature than humanoids. However, no one had previously identified the opportunity to penetrate the consumer market. Earlier, the definition of quadruped products always seemed to lack clarity—how do they interact with humans? What kind of collaborative relationship do they establish? This lack of clear definition meant that quadrupeds were often relegated to research, education, or remote-controlled toys, limiting their perception. Until our product launch last year, which effectively completed a definition cycle, the industry arrived at a consensus: under this definition, quadrupeds have the potential for incremental advancement, and the market could expand significantly beyond the previous annual sales of just a few tens of thousands of units. The leading players in the industry are optimistic about the potential for quadrupeds, believing that the surge will occur in the consumer market.

Phoenix Technology: So, what do you believe is Vbot’s competitive edge? We’ve seen companies like Honor suddenly emerge in the humanoid robot space, overshadowing earlier achievements.

Zhao Zhelun: There are three key components. First, the iterative process alongside users. Early user feedback significantly accelerates the refinement of our product definitions. Our iteration speed is exceptionally fast, largely due to the strength of our R&D team, but even more importantly, we receive authentic feedback from real users. Competitors may benchmark against our products, but recent OTA updates have led to noticeable changes in our offerings, driven by user input. Second, the overall data closed loop. We are the first company to truly enable users to utilize embodied intelligence products within their living spaces, forming a data closed loop. Previously, the only analogous data came from Tesla’s autonomous driving, which is in a traffic environment. However, the living space environment in residential communities and parks is vastly different. Our quadrupeds can be the first to gather data from these living environments, yielding far more data than traditional collection methods. Finally, the organizational advantage. Over the past year, we haven’t diversified our product lines; a team of around one hundred has successfully navigated the entire engineering loop—from initial product definition and design to hardware development, system development, software algorithm development, supply chain, production, sales, and marketing. This level of management is challenging for larger companies or startups that lack focus on this specific endeavor.

Phoenix Technology: You mentioned that user-driven product definition is reminiscent of the current state of new energy vehicles.

Zhao Zhelun: It resembles the early stages of the new energy vehicle sector. I’ve always viewed it as the amalgamation of two elements. One is the transition from Model S to Model 3. The Model S defined the architecture, determining the placement of the battery, motor, and sensors, essentially establishing the structure of electric vehicles. However, its drawback was its high cost. With the Model 3, the core definitions remained, but costs were significantly reduced by eliminating unnecessary components, leading to a surge in consumer adoption. This transition was facilitated by user feedback from the Model S and X and collaboration across the supply chain, bringing the price down from nearly 100,000 USD to 35,000 to 40,000 USD for the Model 3. We are quite similar in this respect. The other element is the progression from Ideal ONE to Ideal L9. The Ideal ONE was already a great product but received extensive user feedback. The L9 incorporated improvements such as the addition of a fridge, a rear screen, and air circulation, all derived from the challenges identified in the Ideal ONE, resulting in a much stronger product definition.

Phoenix Technology: Are there specific examples of this in the “Big Head” design?

Yu Yinan: First, the majority of our core components are self-developed. This has multiple benefits. In terms of cost control, we can produce joints at less than half the industry standard cost. For quality control, because everything is self-developed, we can capture all issues in fine detail and continuously improve throughout the iteration process. Our failure rates are one to two orders of magnitude lower than those of other companies. Thirdly, efficiency is enhanced; our development pace is not only self-reliant but also benefits from supplier collaborations. A high rate of self-development significantly boosts project advancement speed. If we were to rely on purchasing third-party solutions, it would not only be more expensive but also lower quality, and compatibility issues could prolong the entire project. In terms of software algorithms, our AI team is not large; I’ve learned that some of our competitors have teams several times our size. However, our strength lies in specialization; every member is an expert in their field. This minimizes communication friction and problem analysis costs. Thus, we compete not only on development speed but also on the entire iterative cycle from development, integration, and testing to identifying root problems. Furthermore, the underlying issue of computational power per person matters; it’s not about total computational power but how much each individual can utilize effectively. The ideal approach is to control team size while maximizing computational power, empowering each individual to harness sufficient computational resources effectively. This contrasts with companies that may have ten times our workforce but lack our level of computational power.

Phoenix Technology: You previously mentioned that if robots surpass human limits, you would develop humanoid robots. Are you preparing for that now?

Yu Yinan: We have already begun work on this, currently in the stage of developing our first engineering prototype. This has always been part of our long-term strategic roadmap. Using the Go analogy again: once your first move is reasonably solid, you can start preparing for the second position. On the quadruped side, some engineers are finalizing their work, and we are reallocating some resources to initiate early planning for humanoid robots. This includes preliminary industrial design, structural stacking, and selecting core components. Our experience with quadrupeds has led to some upgrades in our thinking about the architecture. For example, our robot dog currently has an exoskeleton structure, with the skeleton on the outside and internal components protected. However, humans are not structured this way. We have an internal skeleton, surrounded by muscles, fat, and skin. Hence, we prefer an internal skeleton architecture for humanoid robots. This offers numerous advantages: greater flexibility in appearance, higher freedom of movement, and enhanced safety due to a softer body. However, it also presents greater technical challenges in structural strength, design, and matching structure with appearance. But this distinction is akin to the difference between an “electric tractor” and an “electric Porsche.”

Phoenix Technology: Once developed, will you pursue challenges such as speed, control, or competitions?

Yu Yinan: I haven’t thought much about that. A good product should evolve across multiple dimensions. Today, many companies focus on rankings, whether for models or competitions, but outside of achieving fame, these pursuits may lack significant purpose. I’m not suggesting they are meaningless—just like in the automotive industry with F1 and rally racing, they do drive technological advancements. However, we prioritize core technology over mere appearances. The essence is technological progress. For instance, achieving higher torque in a smaller volume, minimizing weight—all these involve innovations in joint architecture, motor architecture, and gear systems. Additionally, increasing computational density raises questions about heat dissipation. For instance, current batteries are centralized, resembling a brick in the chest cavity, resulting in large chests for many robots. Is it possible to create a distributed power architecture, similar to a human’s circulatory and nervous systems, allowing for more dynamic and flexible energy distribution? These challenges pose massive technical hurdles, but overcoming them would lead to more advanced products. Pushing the limits of technology is our core pursuit.

Regarding Competitive Barriers: Technology Lacks Barriers, but Organizational Strength Matters

Phoenix Technology: As more companies express interest in the consumer market, how will you respond to the competition?

Yu Yinan: Competition is always present in any field. You cannot find a sector devoid of competition; it’s merely a matter of who enters the battlefield first. At least today, there is a consensus in the industry that the C-end market, particularly for quadruped robots, represents a more significant battleground. We entered this battlefield earlier than others and currently lead in defining product discourse in this market. When most manufacturers aim to define a quadruped robot or a home embodied robot, they inevitably need to reference and compare their offerings to ours.

Phoenix Technology: However, if competitors emulate the small details you’ve developed, can’t they catch up by poaching your talent?

Yu Yinan: While competition exists, enabling continuous advancement is essential; technology itself does not provide barriers. As you mentioned, if competitors recruit a single individual from our team, they might quickly grasp what we are doing. However, truly producing physical products and following the Physical AI route involves multiple barriers. You cannot simply poach an entire team, and the ability to recruit individuals does not create a comprehensive competitive advantage; at best, it supplements localized capabilities. This competitive dynamic is ongoing. We also hire from others when we need specific skills. Ultimately, every new hire must integrate into our organization, enhancing our organizational strength. Genuine competitive power derives from improving organizational capabilities and efficiency—whether you can accomplish tasks using a fraction of the cost or time compared to others who take years to achieve the same outcomes.

Phoenix Technology: During this process, how do you communicate with investors? Their expectations seem to change constantly.

Yu Yinan: This is where your resolve comes into play. If you continuously follow market sentiments, the company will inevitably be pulled in various directions. You must have your own roadmap and maintain your composure. It shouldn’t be about what’s popular today; otherwise, you won’t earn investors’ respect. When they invest in you during a trend, they may become anxious once the trend fades, and such emotions can backfire.

Phoenix Technology: What principles and values guide you in the pursuit of building robots?

Yu Yinan: From day one, we established a core principle: to create robots for living spaces. This is our foremost principle. Breaking it down, we focus on two main aspects. On the left, we prioritize premium hardware. What defines premium hardware? Look at the items on this table—MacBook, iPhone, Porsche sports cars, Mercedes G-Class. Regardless of current sales volumes, these products set the benchmark for their categories. They define an era. On the right, we view AI or software as fluid; it evolves over time. What you see today will differ in just a few months as it upgrades and improves. We designed our robot dog with a full-year extension plan, evolving from a platform product into a versatile application-focused product, continuously enhancing its value. On top of these two cores, we build our technological competitiveness. The most critical aspects of hardware are energy, power, and computational capability. In terms of algorithms, we focus on models, AI talent, and data. Application-wise, as Zhao Zhelun mentioned, we continuously loop back with users, analyzing their feedback to form our product development roadmap. These elements are constants. On top of these constants, we craft our narrative for investors: what we were like in the past, what we are now, and what we aim to become, including the technological reserves and advanced technologies we plan to develop. We occasionally pursue rankings to validate our technological competency, but our logic is built bottom-up, layer upon layer, rather than merely reacting to current market fads.

Phoenix Technology: How many of your initial users are C-end consumers, and how many are industry insiders?

Zhao Zhelun: Almost all are consumer users; industry insiders make up less than 10%.

Phoenix Technology: What do their profiles look like?

Zhao Zhelun: We recently conducted a user survey, and it yielded interesting results. Nearly 90% are families with children. Currently, over 80% of the purchasing decision-makers are male, but in our offline experience stores, we noticed that women and mothers also show significant interest; it’s just that our initial outreach predominantly reached men. Our target demographic consists primarily of middle-class families, concentrated in the top 20 GDP cities. The top six cities—Beijing, Shanghai, Guangzhou, Shenzhen, Chengdu, and Hangzhou—account for half, with the other fourteen cities making up the remaining half. From our public testing data, a fascinating observation emerged: there’s a disparity between what users claim to enjoy and their actual usage. Some users who seem keen to showcase their experiences may not use the product long-term; they might engage only during OTA updates or out of novelty. Conversely, long-term users are predominantly families with children, particularly those aged between three and twelve. Children are genuinely interested in this technology and require companionship. This generation of children in China often feels bored, as parents are busy and have limited time for interaction. Our product indeed serves as a companion. Additionally, with strong AI capabilities, it can assist children in English language learning and answering general knowledge questions. Furthermore, it can accompany children outdoors, meeting both children’s and parents’ needs.

Phoenix Technology: Will the price decrease further?

Yu Yinan: We anticipate a reduction of about 10% to 20% annually for the next five years. As production scales up and our architecture continues to optimize, the costs will naturally decline. The current major cost drivers are joints, which are heavily tied to scale, and the computing platform, which relates to chip performance. Each new generation of chips, with higher integration levels, can reduce costs by 20-30%. Sensors generally adhere to Moore’s Law, and structural components and batteries will also decline in cost yearly. The only component that has seen price increases lately is memory. The overall logic is as follows: when your production volume truly increases, and your architecture is sufficiently optimized, the costs will converge towards the cost of raw materials—copper, rare earths, aluminum, and PCBs.

Zhao Zhelun: While the current price for our generation may not be low for C-end consumers, our cost pressures are quite significant. Compared to similarly priced robotic dog products, our materials are of much higher quality. Our battery, motor, sensors, and computational power far exceed those of competitors. Essentially, we’ve developed algorithms for products originally supplied to universities for research, allowing consumers to use them right out of the box. This represents a brand-new category, which is why initial costs might be higher.

On Organizational Innovation: Recruitment Focuses on Learning Ability, Not Computational Power

Phoenix Technology: In this era, there’s a trend towards AI-native organizations. Are your engineers no longer hand-coding?

Yu Yinan: We have transitioned away from manual coding; now, we primarily utilize AI. I believe that future startups won’t require as many employees. The only positions needing larger teams are those that frequently interact with people, such as in-store personnel or customer service. Internally, this reflects a fundamental change in organizational structure. In the past, Huawei emphasized “building organizations on top of processes,” converting business insights into processes and placing everyone at nodes within those processes, assigning them responsibilities, rights, and obligations. This approach essentially runs on processes. Today, however, we should consider “building organizations on top of AI.” Everyone forms a roundtable with AI at the center, interacting with AI for problem identification and distribution. The organization transforms into a larger project-based structure. Team members agree on divisions of labor and work collaboratively, enhancing efficiency through AI. The primary advantage is a significant reduction in management and communication friction costs. In the book The Infinity Machine, Demis Hassabis of DeepMind emphasizes that large projects, often involving tens of thousands to hundreds of thousands of people, face challenges in information transmission—distortions occur from the top down and deviations from the bottom up. This is why some extensive projects struggle to progress, taking decades with slow advancements, rendering various project management tools ineffective. By placing AI at the core, we can handle information exchanges, greatly reducing friction costs between individuals. Each member’s capabilities, both in depth and breadth, are significantly expanded. Some may think this requires purely expert talent, but I don’t entirely agree. The essence is the individual’s learning ability. Experts grow through numerous projects, learning from successes and failures. Therefore, our recruitment criteria focus on whether individuals possess strong self-growth capabilities. If they do, they can surpass expectations within three months, regardless of age. This is our organizational philosophy.

Phoenix Technology: Has your mindset changed since you started your entrepreneurial journey a year ago?

Yu Yinan: I feel quite happy now. Today, what we’re doing is being recognized by the industry. Regardless of how many units we sell, we have essentially become a leading player in this track. Anyone wanting to enter this space—investors considering funding others or users looking to purchase—will first ask, “What differentiates you from Vbot?” At the start of our journey, when I wrote our business plan and spoke with investors, they primarily inquired about our differentiation from established companies. Today, no one asks that anymore; instead, the question is now, “If you want to enter the consumer market, how do you differentiate yourself from Vbot?” The momentum has shifted. I believe one of the most gratifying aspects of running a business is whether you can continue to lead the industry. This isn’t contingent upon how much funding you secure.

Phoenix Technology: If you achieve success, would a few more milestones make you feel that Vbot has reached significant success?

Yu Yinan: There’s no endpoint. Building a business is like surfing; each wave must be higher than the last—it’s an infinite game. You can never reach a point where you think, “This has been done well enough,” and transition to a purely operational mode. The quadruped line needs to enter a cycle of iteration with users, akin to nurturing a child to grow healthy. Meanwhile, we must also push boundaries to make new products the industry standard.

Phoenix Technology: What are the specific plans for humanoid products?

Yu Yinan: The humanoid design will debut in August, with a relatively mature engineering prototype expected by the end of the year. There are still gaps between engineering and mass production. Recently, I’ve sensed a pressing urgency—based on insights from various channels, Tesla and Figure are progressing more rapidly than what is publicly visible.

Phoenix Technology: Why have so many robotic companies chosen to reuse established supply chains, resulting in similar product outputs? Why haven’t they adopted a mindset for entirely new product definitions?

Zhao Zhelun: Because VCs demand “speed.” Many of these companies follow VC logic. I’ve also wavered, asking Yu Yinan if we should go to Suzhou to partner with a factory to quickly produce a humanoid solution, merely altering the appearance. That could have a prototype ready in a month or two. Yu responded with two points: first, it would look terrible, and second, it doesn’t align with our values. Such products may serve as demos or research tools, but they lack the comprehensive design and engineering structure that we emphasize. Hence, there are not many companies that can develop a complete product from the ground up.