China’s Humanoid-Robot Boom Faces Its Hardest Test: Useful Work

China’s humanoid-robot companies are trying to turn spectacular demonstrations into a business by narrowing the problem. Instead of starting with fully general-purpose home robots, they are targeting repetitive, measurable work in factories, warehouses and logistics sites—places where workflows can be structured and performance can be tracked.

The 2026 World Robot Conference in Beijing offered evidence of that shift. It also exposed the gap that still separates a convincing booth demonstration from dependable commercial automation: most deployments remain pilots or tightly controlled use cases, while broad, unattended operation is still relatively uncommon. 1211

The strategy: start with jobs that can be measured

The most credible applications shown at WRC involved constrained tasks, including:

• sorting parcels;
• packing mobile phones and other components;
• lifting and stacking boxes;
• moving crates and production materials;
• replenishing inventory; and
• welding or supporting production-line work.

These tasks are attractive because companies can define success in operational terms: completed movements, correct picks, cycle time, error rate and labor or equipment cost. Logistics is especially easy to evaluate because the return on investment can be calculated against throughput and staffing requirements. 211

Household chores were also demonstrated, but homes are a considerably harder environment. Objects, layouts and instructions vary widely, and a robot must handle unexpected situations rather than repeat a carefully prepared sequence.

What the 2026 World Robot Conference revealed

WRC ran from August 19 to 23 in Beijing E-Town. Organizers said the event brought together more than 300 exhibitors, over 2,000 exhibits and more than 150 product debuts. 219

The exhibition’s message was not that humanoids had suddenly become general-purpose workers. Rather, it showed a broadening of the commercial pitch. Alongside the dancing, boxing and sports demonstrations that helped attract public attention, companies presented robots for logistics, production lines, inspection, retail and other services. 1211

One UBTECH display used about 10 humanoid robots in a coordinated workflow involving lifting, stacking, sorting and replenishment. That kind of demonstration is useful because it shows how multiple machines might be integrated into a process. But it does not, by itself, prove that the same system can operate without supervision through full shifts in a live warehouse. 12

Early deployments are real—but scale is the question

Companies cited deployments and pilots across logistics, automotive, electronics, aerospace and energy. RobotEra, for example, was reported to have more than 100 parcel-sorting robots operating across 15 China Post warehouses, while other companies demonstrated crate handling and precision packing in industrial settings. 34

Those examples matter because they move the discussion beyond a robot performing once on a stage. Yet the evidence remains uneven. A reported deployment may involve a limited number of sites, a narrow task, human oversight or a workflow adapted specifically for the machine. Industry reporting from WRC noted that robots had entered real posts in logistics, production lines, inspection and commercial services, while large-scale deployment was still relatively rare. 11

The key distinction for buyers is therefore deployment versus production-grade operation. A pilot establishes that a robot can work in a particular environment. Commercial scale requires it to keep working as conditions change, maintenance is needed and the cost of human supervision becomes visible.

The performance metrics that will decide adoption

Walking ability and theatrical dexterity are easy for audiences to understand, but they are weak measures of industrial value. Customers will need evidence on:

• Uptime: How much of a scheduled shift does the robot perform useful work?
• Throughput: How many completed tasks does it deliver per hour?
• Reliability: How often does it drop, misidentify or mishandle an object?
• Recovery: Can it detect and correct mistakes without an engineer?
• Safety: Can it work near people and stop predictably when conditions change?
• Maintenance: How frequently do components, batteries or grippers require attention?
• Total cost: Does the cost per completed task beat the available human or conventional-automation alternative?

This is the sector’s central transition: from proving that a robot can perform a task once to proving that it can perform the task repeatedly, economically and with limited intervention. 1

Investment is running ahead of commercial proof

Investor enthusiasm was highly visible around WRC. Unitree’s Shanghai debut coincided with the conference, and its shares rose more than fivefold on the first trading day. The company’s IPO sought to raise 6.1 billion yuan, or roughly $904 million.

That capital can help manufacturers build supply chains, collect operating data and reduce unit costs. It can also increase pressure to demonstrate growth before customers have adopted robots at scale.

Unitree’s prospectus reportedly showed that most of its customers were universities and research institutions. That detail is an important market reality check: research and education sales can support hardware distribution and development, but they are not the same as sustained demand from factories, warehouses or households.

The development timetable also remains long for general-purpose capability. Galbot founder Wang He described a possible “ChatGPT moment” in 2028, defining it as the point when robots could perform roughly 70% to 80% of everyday tasks without specialized training. That is an industry expectation, not a demonstrated result—and it reinforces how far the sector still has to go beyond task-specific deployments.

Government support is helping build an ecosystem

China’s robotics push is supported by a strong institutional and industrial ecosystem. WRC included a dedicated area for 49 centrally administered state-owned enterprises and added procurement and supply-chain matchmaking elements to the event. 6

Official figures cited at the conference said China’s robotics-sector revenue exceeded 300 billion yuan in 2025, while revenue from major robotics companies reached 165.5 billion yuan in the first half of 2026, up 24.5% year on year. These figures cover the broader robotics sector rather than humanoid robots alone, so they should not be treated as proof of humanoid profitability. 21

The policy and manufacturing environment can accelerate experimentation, component production and customer access. It cannot remove the basic commercial test: whether individual machines deliver enough value in real operating conditions.

Foreign partnerships offer validation—and constraints

International relationships remain part of the commercialization strategy. UBTECH announced a partnership with Airbus involving robotics for aviation manufacturing, linking humanoid-robot development to a demanding industrial customer and a defined production environment.

Foreign suppliers also appeared at WRC. RealSense, a U.S.-based maker of robotic vision systems, was among the event’s relatively few foreign exhibitors, illustrating how Chinese robot makers still depend on international components and technology relationships even as domestic production expands. 2

At the same time, geopolitical restrictions create uncertainty for companies seeking overseas markets and advanced hardware. That makes domestic scale, local supply chains and scenario-specific products more important—but it also means international commercial validation will remain valuable.

What to watch next

The strongest evidence of progress will come from customer-site data rather than larger demonstrations. Useful disclosures would include:

1. the number of robots operating at each site;
2. the percentage of tasks completed without human intervention;
3. uptime and mean time between failures;
4. cycle time and error rates;
5. safety incidents and intervention frequency; and
6. the payback period compared with existing labor or automation.

China has a large supplier base, substantial capital formation and visible momentum in industrial scenarios. WRC 2026 showed that companies understand the need to demonstrate useful work, not just athletic ability. But the sector’s commercial promise remains plausible rather than proven until independent, repeatable performance is demonstrated across many customer sites.

The next phase of the humanoid-robot race will therefore be less theatrical. The winners will be the companies that can turn a carefully staged task into a dependable operating system for work.