For RMB9,998, Noetix Robotics offered buyers something that had usually belonged to research laboratories, corporate demonstrations and expensive technology showcases: a humanoid robot they could consider buying for themselves.
Bumi, the small robot the Chinese company introduced in October 2025, was aimed at programming education and companionship. Its appeal rested on an immediate proposition. Children could programme its movements, watch it dance and interact with a machine that looked and moved like a small person.
That proposition also expressed a larger choice about how to build a robotics company. Noetix would first make the body affordable and capable enough to sell. More sophisticated intelligence would follow.
Three years after its founding, the company is attempting to turn that sequence into a much larger business. It is pursuing higher deliveries, investing in its own embodied-intelligence models and seeking customers overseas. In October, media reports said it had confidentially submitted a Hong Kong listing application, potentially targeting a 2027 flotation.
The question connecting those ambitions is whether the hardware business Noetix built first can provide enough customers, experience and cash to support what comes next.
A Price Decided Before Production
Founder Jiang Zheyuan left his doctoral studies at Tsinghua University in 2023 and established Noetix that September. His early technical priority was the robot’s body and motion control: the machinery and software that allow it to balance, move and respond reliably.
The distinction matters. A robot can execute demanding movements while relying on external models for conversation or higher-level instructions. Building an athletic machine and building one that can work out what to do in an unfamiliar room involve different capabilities.
Jiang’s decision was to address the physical machine first. He explained that, when the company began, he saw unreliable bodies and limited manufacturing readiness as the immediate bottlenecks. AI approaches were also changing rapidly. Noetix concentrated its resources on hardware while waiting for a clearer direction in the intelligence stack.
Low cost was part of that engineering brief. In a March interview, Jiang described using composite materials in Bumi’s body to reduce its dependence on metal structures. The price was being worked into the design, rather than left for a sales team to negotiate after production.
That approach gave the company a product with a readily understood purpose. Parents and schools could buy a programmable humanoid for education and interaction. They did not have to wait for general-purpose domestic labour to become commercially viable.
The narrower use case helped Noetix reach buyers sooner. It also established the next commercial task: giving those buyers enough continuing value to make the robot more than an exciting first purchase.
Attention Becomes a Manufacturing Problem
Noetix’s machines found a powerful route to public attention through performance. Its N2 robot finished second in the 2025 Beijing humanoid half-marathon. In 2026, the company’s robots appeared alongside comedian Cai Ming in a sketch on China’s Spring Festival Gala.
The exposure connected technical demonstrations with a mass audience. A robot that families had seen on television was also available through retail channels.
At the AWE exhibition in March, Jiang said orders had surged after the gala and set a target of more than 10,000 Bumi sales during 2026. The company was moving from proving that people would notice its robots to showing that it could deliver them consistently.
Jiang’s account of the obstacles was strikingly practical. He identified component availability and production-line management as central challenges. A missing part can stop an entire machine from being completed. As monthly output rises, the company must train and manage a larger workforce while maintaining assembly quality.
These are consequential problems for a company whose market entry depends on affordability. Every reworked unit, delayed delivery and service visit consumes part of the margin that a low selling price leaves available.
The manufacturing target therefore carries more meaning than a shipment milestone. It tests whether Noetix can preserve the economics designed into Bumi while producing and supporting it in much larger numbers.
The Brain Becomes the Expensive Part
By September, Noetix was making its next technical commitment explicit. It launched the embodied-intelligence sub-brand Scalabot and the HERON world-model architecture, extending its ambitions into how robots understand environments and learn to act within them.
The company had begun assembling a dedicated intelligence team in July 2025, according to Jiang’s September interview with Caijing. The work added a substantial research programme to an organisation already scaling physical production.
His logic connects the two activities. Cheaper, reliable machines can be deployed more widely; those deployments can create opportunities to collect experience for training better models.
Making that connection productive requires another layer of engineering. A robot in a classroom generates experience, but useful training data must still be selected, recorded, organised and connected to tasks the company wants its models to learn. A large installed base creates access to potential data. The collection and training system determines its value.
Jiang told 21st Century Business Herald that intelligence research had become a major area of expenditure. He also described hardware as a potential source of funds for that work.
This is the financial hinge in Noetix’s strategy. The body-first approach opened an earlier route to revenue. The company now wants that operating business to help support a more expensive effort whose commercial returns will take longer to develop.
External capital helps bridge the timing. Its backers include Chendao Capital, an investment platform associated with CATL. A prospective Hong Kong listing would take that financing journey into public markets, where investors would be able to examine more closely how hardware sales and research spending fit together.
Taking a Chinese Starting Point Overseas
Noetix’s domestic route has combined educational demand, accessible hardware and unusually broad public exposure. Its overseas expansion will test how much of that combination travels.
At CES in January, co-founder Zhang Shipu outlined five target regions: North America, Europe, the Middle East, Southeast Asia, and Japan and South Korea. He described an ambition to reach thousand-unit scale in overseas expansion during the second quarter, supported by local deployments, customer service and repeatable reference projects.
That programme involves more than shipping robots abroad. Schools need teaching materials and support. Distributors need spare parts and repair arrangements. Families need a reason to keep using the product after its novelty fades.
Noetix has some of those relationships to build from in China. An investor, Beijing Industrial Development Investment Management, reported partnerships with programming-education provider Codemao and retailer Kidswant, alongside deployments and donations involving schools. These connections show how a robot can become part of an existing educational or retail system.
Overseas growth will require comparable connections in markets with different curricula, languages and purchasing habits. The product’s price can win attention, but local usefulness and support will determine whether that attention becomes a durable business.
Noetix’s first move was to make a humanoid affordable enough to enter more classrooms and homes. That choice has given it a product to sell while the broader intelligence problem remains under development.
It is now trying to make that early foothold carry three ambitions at once: larger production volumes, more capable intelligence and international growth. Building the body first won Noetix a place in the market. The next stage is to make that place valuable enough to finance everything it still wants the robot to become.
