"Before we launched, the most common application for robots was commercial entertainment. We didn't see long-term value in that—it wasn't sustainable."
Speaking at the 2026 Intelligent Industry Development Conference, Zhao Zhelun, co-founder and product president at Vita Dynamics, made his stance clear. The event was co-hosted by the Intelligent Industry Innovation Center and Hubei Science & Technology Investment Group.
So while the rest of the industry rushes into humanoid robots, Vita Dynamics bet on four-legged machine dogs.
"We believe a machine dog is a viable commercial product. Before us, no one had turned this into a consumer business; everything on the market was geared toward research or industrial use."
It is a path that looks somewhat contrarian.
But the numbers are in: since mass deliveries of its first-generation quadruped, DaTou BoBo (hereafter "DaTou"), began in May, the dog has climbed to the top of JD.com's consumer robot category in just four months, amassing over 2,000 individual users.
How did Vita Dynamics pull it off?
Zhao Zhelun, Co-founder and Product President of Vita Dynamics
Turning a 100,000-Yuan Dog Into a 10,000-Yuan Product
Vita Dynamics's choice of quadrupeds wasn't a snap decision.
The reality is this: before Vita Dynamics, there was no quadruped robot truly designed for the consumer market.
Globally, quadruped robots aren't new. Over the past few decades, many have tried: Sony with AIBO back in 1999, followed by Boston Dynamics' SpotMini, Xiaomi's CyberDog, and Unitree's Go2.
Everyone tried to make a consumer play. "But every one of them seemed to be missing something," Zhao says.
Sony's AIBO had great interaction but lacked mobility. SpotMini proved quadrupeds could execute tasks, but mostly for industrial inspection. Unitree's Go2 pushed athletic performance to new heights, yet remained essentially a research-grade device focused on remote control and development.
As a latecomer, Vita Dynamics aims to change that dynamic and unlock a broader market.
Specifically, Vita Dynamics hopes to leverage its automotive background—packing edge AI and autonomous driving tech into a consumer-grade, practical machine dog capable of full environmental understanding, perception, planning, decision-making, and execution.
But technical capability is just the entry ticket. To truly crack the consumer market, there is a harder barrier: price.
Historically, quadrupeds failed to go mainstream largely because they were too expensive. High computing power, advanced sensors, and large batteries pushed configurations into the 100,000-yuan range—hardly a mass-market price point.
To control costs, Vita Dynamics developed core components like motors in-house, ultimately bringing DaTou's price down to just over 10,000 yuan.
The strategy is working.
According to Zhao, more than 2,000 consumers are actively using DaTou, feeding back massive amounts of data daily. The two-week retention rate sits at 86%, and even after three months, roughly 80% of users are still engaging with the device.
Cumulatively, DaTou has completed 1.26 million large-model interactions, accompanied users on walks covering over 23,000 kilometers, and logged more than 130,000 hours of companionship.
Zhao offered a comparison: DaTou averages over 100 interaction rounds per day—more than ten times the volume of typical smart vehicle interactions.
This means a high-frequency connection has formed between DaTou and its users; it is no longer just a cold piece of hardware.
To expand further, Vita Dynamics is rapidly scaling its channels. It has already established its first manufacturing and service center, opened 8 branded experience stores and 22 joint outlets, covering users in 107 cities.
The business model is already working at some locations. On that basis, Zhao says Vita Dynamics will aggressively expand, aiming to grow its offline sales and experience network to 100 locations by year-end.
Vita Dynamics's stance is clear: to drive volume, the offline channel is critical. This kind of product demands hands-on experience.
Image Source: Vita Dynamics
Four Months, Four OTAs: Following the EV Playbook
There is no doubt Vita Dynamics has made its mark in the quadruped space.
But this isn't accidental.
Vita Dynamics's core team mostly hails from Li Auto and Horizon Robotics. They know how to refine consumer products and are well-versed in data-driven iteration—forming the company's solid foundation.
That expertise shows most directly in product updates.
In the four months since DaTou's delivery, Vita Dynamics has pushed four over-the-air (OTA) updates, adding 31 new features and optimizing 66 others. That pace mirrors the early days of the electric vehicle industry.
The evolution of outdoor mobility is a prime example. When DaTou first launched (SOP), it only had a "pathfinding" mode where users controlled direction via a leash while the dog avoided obstacles autonomously.
"Engineers thought it was great, but the friction for actual users was high," Zhao explains.
So Vita Dynamics quickly rolled out an outdoor "follow" function, allowing users to put a small beacon in their pocket for the dog to trail behind. But a new issue emerged: with the dog constantly trailing, users had to look back every half minute—not exactly user-friendly.
In the latest OTA, Vita Dynamics introduced an outdoor "accompany" mode. The dog now dynamically adjusts its position relative to the user, actively staying within their field of view.
Pathfinding, following, accompanying—three iterations rooted in the same loop: real user feedback identifies issues, product teams iterate via OTA, and the updated device returns to the user to generate more data.
"Without user feedback and this data loop, we couldn't have dreamed this up just sitting in a lab or office," Zhao says.
This leads to another of Zhao's core beliefs: commercial deployment provides the training infrastructure for embodied intelligence.
After all, that "feedback—data—iteration—delivery" loop requires enough devices running in the real world.
"Running 1 to 5 devices in a lab creates a fundamentally different loop than running 1,000 or 2,000 devices in the real world," Zhao notes.
Vita Dynamics's approach is virtually identical to the playbook used in autonomous driving over the past few years.
In fact, Zhao views the auto industry as the "parent ecosystem" for embodied intelligence. Smart EVs and smart robots share a highly similar core architecture—sensors, compute platforms, motors, and batteries—and the supply chain for these in China is already mature.
But he emphasizes that migration doesn't mean direct copying.
"What can truly be reused is industrial capacity. Excellent motor, chip, and lidar companies are re-optimizing for robots. Without the accumulation of the auto industry, trying to build robot-specific chips, battery cells, or motors from scratch would leave a massive gap."
Image Source: Vita Dynamics
Deepening the Quadruped Play, Exploring the Humanoid Frontier
With the commercial loop initially validated, the question is: where to go next?
Vita Dynamics's answer is clear: advance both quadruped and humanoid lines, but with different rhythms and positioning.
On the quadruped side, the product matrix is expanding.
Beyond the consumer DaTou, Vita Dynamics has launched the DaTou EDU-W, a wheeled version with a top speed of 2.5 m/s (roughly 9 km/h). In standardized testing, its energy efficiency is 50% higher than the standard EDU version.
There is also the developer-focused DaTou EDU, aimed at industries like retail, hospitality, healthcare, tourism, and education. It supports secondary definition and development, providing a base for entering more vertical scenarios.
Zhao reveals that the DaTou platform will now focus on three directions: lowering prices and enhancing interaction for family companionship; boosting payload and range with point-to-point navigation for outdoor delivery; and leveraging navigation and obstacle avoidance for guiding the visually impaired.
On the humanoid front, Vita Dynamics launched its first robot, the Vita Dynamics ATOM, last month. Pre-orders are open, with deliveries expected by year-end.
According to released specs, ATOM stands 1.6 meters tall with 31 degrees of freedom. It features extended arm and leg designs to expand its operational reach and movement efficiency, targeting four living spaces: homes, retail, offices, and public services.
The humanoid market is already crowded with players tackling both general and vertical scenarios. Why launch now?
Zhao's explanation: because humanoid robots are finally ready for productization.
More specifically, DaTou's success over the past year provides a solid foundation. The infrastructure for integrated software-hardware definition and user-feedback-driven model iteration is already running and can be directly reused to quickly close the data loop for humanoids. Engineering know-how—motor selection, battery specs, structural design, materials—also transfers over.
"These factors allow our new product to be fast and good," Zhao says, viewing this as Vita Dynamics's competitive edge entering the humanoid space.
For instance, ATOM's head has 2 degrees of freedom, continuing DaTou's natural interaction design. Visually, ATOM adopts DaTou's design language, using a facial screen and ear light effects for intuitive human interaction.
However, Zhao admits that while DaTou has handled 70% to 90% of user education, that process may need to be repeated for humanoids. "Today, most companies buy humanoid robots in batches of two or three; it's not a long-term demand. We need a model that can be replicated at scale."
One potential path he envisions is scalable chain services: "If one Starbucks store can use it, then thousands can. If it works for Starbucks, it works for Luckin."
Bubbles Aren't Bad; The 'Ford Moment' Will Come
In 2026, the financing boom in embodied intelligence has been accompanied by persistent talk of a bubble.
A recent case in point is Unitree Robotics. Since listing on the STAR Market, its stock has steadily declined. As of the close on September 11, Unitree shares sat at 477.12 yuan with a market cap of 193 billion yuan—down more than 56% from its peak of 1,100 yuan and 444.8 billion yuan at listing.
Unitree's continued pullback has thrust debates over embodied intelligence valuations into the spotlight.
When asked about the "bubble," Zhao's take is surprisingly positive.
In his view, emerging industries require pioneers to invest and build infrastructure in advance to pave the way for post-maturity prosperity. The U.S. internet bubble, for instance, saw fiber optics and communication infrastructure laid early, which later incubated a generation of tech giants.
"So a bubble is inevitable." Zhao even believes the hype turns the industry into fertile ground where truly excellent companies will eventually emerge.
For entrepreneurs, rather than worrying about the bubble, it is better to focus on finding a direction with long-term value and investing consistently, rather than chasing short-term gains.
Notably, in this current wave of embodied intelligence, besides startups like Vita Dynamics with auto backgrounds, many automakers are jumping directly into robot manufacturing.
Zhao believes this influx of resources proves the sector's promise. But beyond the noise, it is not easy for OEMs to master embodied intelligence; at least two capabilities are missing.
First is a true Physical AI loop.
Zhao argues that the organizational capacity to link user feedback, data collection, model training, simulation, real-world verification, and OTA updates is still rare in the auto industry. Only a handful of companies that have cracked end-to-end autonomous driving and achieved continuous OTA iteration via user feedback possess this—not everyone.
Second is integrated innovation capability.
"Cars mostly see incremental annual updates or supply-chain improvements. Embodied intelligence requires integrated innovation across software and hardware, product definition, and business models," Zhao says, noting this demands organizational shifts from OEMs.
Take channels: if OEMs rely on dealer inertia and flood thousands of 4S stores with robots, "that could be a resource, but it could also be a shackle."
So, will embodied intelligence eventually see a "Ford Moment" like the auto industry did?
Zhao's answer: you can't simply draw a parallel to cars and wait for one unified tipping point.
"In terms of diversity, the industry's development will look more like the computer sector. Many categories will emerge, and each will experience its own 'Ford Moment'—the moment of mass production, the establishment of industry standards, the drop in costs, and the moment it becomes a standard part of life."
Specifically, he judges that machine dogs will be the first category to enter homes and achieve mass adoption, with rapid momentum between 2027 and 2028—a segment currently undervalued. The humanoid cycle is relatively longer, especially for home entry, which will take more time.
Conclusion
Looking back, Vita Dynamics's strategy is clear: ignore the humanoid hype, start with quadrupeds, slash prices, build retention, close the data loop, then leverage that capability for humanoids and more.
In a sector filled with foam and noise, where most chase the grand narrative of humanoid robots, Vita Dynamics chose to get machine dogs into thousands of homes first.
That path may lack short-term appeal, but in the long run, it might just go the distance.
