- Robot vacuum giant is betting the future of home robotics on embodied AI with its first open-source robot
- Deck 2: But its price, positioning and technical limitations raise doubts about whether consumers—or developers—will buy in
China’s red-hot embodied robotics race has a new contender.
When Suzhou-based robot vacuum maker Ecovacs unveiled its first open-source robot, Bajie, on June 18 with a price tag of 49,990 yuan (about $7,373), it was billed as a bold step beyond robot vacuums into embodied AI.
I think it’s a huge mistake.
Bajie is not a humanoid robot. It combines a wheeled base with a robotic arm that extends from 85 cm to 1.29 meters, with an 800 mm reach and a one-kilogram payload. Its advertised tasks are straightforward: pick up toys, organize tabletops and return shoes to the cabinet.
The problem isn’t the engineering. It’s the product definition.
Nearly 50,000 yuan buys a high-end gaming PC, multiple premium robot vacuums or years of professional home cleaning services. Few consumers will spend that much on a robot whose main value proposition is tidying up toys and shoes.
Developers are unlikely to embrace it either. Most open-source robotics researchers prefer inexpensive hardware platforms that cost hundreds rather than tens of thousands of yuan.
Bajie is caught between two audiences — it’s too expensive for consumers and too costly for developers.
Its choice of application is equally problematic.
Ecovacs argues that home organization will be one of the first commercially viable embodied AI scenarios because the demand is frequent and the need is obvious.
In reality, the opposite is true.
Most challenging environment
Homes remain one of the most challenging environments for robots. Toys vary endlessly in shape, size and material.
Shoes range from slippers to boots and high heels. Children run through rooms unpredictably. Pets introduce constant uncertainty.
What appears to be a simple task—putting toys away or returning shoes to a cabinet—is actually a remarkably difficult sequence of perception, reasoning and manipulation.
Households are among the most unstructured environments imaginable, precisely where today’s embodied AI systems struggle the most.
Open-source positioning
That makes Bajie’s open-source positioning feel premature.
Ecovacs has released dozens of underlying capabilities for developers to build upon. But successful open-source ecosystems depend on reliable foundations.
If the robot cannot consistently complete basic household tasks, developers have little incentive to invest time building applications on top of it.
Open source accelerates capable platforms. It does not compensate for immature ones.
The broader technology is not yet ready either.
Robot generalization—the ability to perform reliably in unfamiliar environments—remains one of the biggest unsolved problems in embodied AI.
Industry leaders, including Unitree founder Wang Xingxing, have repeatedly identified generalization as the sector’s primary technical bottleneck.
Zhu Shiqiang, roboticist and president of Zhejiang University Robotics Institute, has offered a similarly sober assessment: the industry has plenty of prototypes but very few products capable of large-scale commercial deployment.
Against that backdrop, promises that embodied robots will enter ordinary households within three years sound more aspirational than evidence-based.
To be fair, Ecovacs itself is performing well. Its 2025 net profit more than doubled to 1.76 billion yuan as premium products and improved operational efficiency boosted margins.
But Bajie exposes another mismatch.
In different directions
Ecovacs has built a successful business selling hardware. If Bajie sells only in small volumes, it will barely affect revenue. If it eventually becomes an open-source platform, the company would need entirely new business models—developer services, cloud platforms or software ecosystems—that fall well outside its traditional strengths.
The product and the business model point in different directions.
Ultimately, Bajie suffers from three fundamental problems: it is too expensive, it targets one of robotics’ hardest application scenarios, and it arrives before the underlying technology is mature enough.
At its core, Bajie feels more like a laboratory prototype packaged as a commercial product than a product designed around a genuine market need.
Investor enthusiasm for embodied AI is understandable. But enthusiasm cannot replace disciplined product thinking.
Every robotics company should be able to answer two simple questions: Who will pay for this? And what problem does it solve that nothing else can?
Until those questions have convincing answers, Bajie is likely to be remembered less as a breakthrough and more as an expensive lesson for the embodied AI industry.
