Home Bots on stageWAIC 2026: China’s Robots Take The Stage in Shanghai

WAIC 2026: China’s Robots Take The Stage in Shanghai

by Pieter Werner

The 2026 World Artificial Intelligence Conference (WAIC) opened in Shanghai on Friday under the banner “AI Partnership for a Brighter Future,” and if this year’s exhibition floor is any indication, the partnership China has in mind increasingly has arms, legs, and, in one case, a direct line to the human brain.

Now in its eighth year, WAIC has grown into one of the world’s largest gatherings of AI companies, researchers, and investors, and a key barometer of China’s “AI Plus” industrial strategy. This year’s edition leaned heavily into embodied intelligence, with humanoid and service robotics companies using the conference to show that robots are moving out of research labs and into hotels, restaurants, warehouses and factory floors.

On the policy side, the event also marked a milestone for global AI governance: on July 16, the day before the conference opened, representatives gathered in Shanghai to sign the agreement establishing the World Artificial Intelligence Cooperation Organization, a body intended to promote international cooperation on AI safety, fairness and shared benefit, building on the Global AI Governance Initiative China first proposed in 2023. Organisers also pointed to the “Beinao-1” brain-computer interface, now in large-scale clinical trials and expected to reach the market in 2027, as a sign of how fast frontier neurotechnology is advancing alongside AI.

KEENON: Humanoids and Specialists Learn to Share the Job

KEENON Robotics arrived at WAIC 2026 with a simple pitch: humanoid robots and purpose-built service robots aren’t rivals, they’re colleagues. According to IDC, KEENON ranked first worldwide in commercial service robot shipments in 2025 and continues to lead the global delivery robot market, and the company used its booth to argue that the industry’s future lies in mixing robot forms rather than betting everything on humanoids alone.

The clearest demonstration was a hotel laundry workflow: humanoid robots load and operate washing machines, retrieve clean laundry and fold garments, while KEENON’s wheeled DINERBOT T9 handles the wider delivery run between floors, every task performed autonomously, without teleoperation. In food and retail settings, the company’s XMAN-R1 humanoid takes on front-of-house work that blends conversation with object handling, from preparing drinks alongside NOWWA Coffee staff to serving customers in dessert and retail shops.

The strategy reflects lessons from KEENON’s existing deployments, more than 100,000 service robots operating in over 70 countries for brands including Burger King, Hilton, BMW and Lego. At Shangri-La’s Trader Hotel, a mixed fleet of eight robots across six different types is already running side by side, which KEENON describes as the world’s first hotel to combine humanoid and conventional service robots in daily operations.

BrainCo: Controlling Robots With Thought Alone

Few booths drew a bigger crowd, or more raised eyebrows, than BrainCo’s. The Chinese brain-computer interface (BCI) company demonstrated a person wearing a lightweight EEG headset simply thinking about grabbing a cup, and watching a robotic arm reach out and do exactly that, no button, no spoken command, no physical movement.

BrainCo’s Brain-Controlled Robot AI Platform reads a wearer’s brain signals, uses AI to decode motor intent, and converts that intent into robot commands in under 200 milliseconds. During the demo, the arm handled tasks requiring real precision, such as picking up an apple, and the company says the platform is designed to work across humanoids, robotic arms and four-legged robot dogs, rather than being tied to proprietary hardware. BrainCo calls the underlying concept “Neuro-Embodied-AI”, the BCI decodes intent, an AI layer breaks that intent into actionable steps, and the robot executes.

The company also unveiled an Embodied AI Data Collection Solution aimed at a problem that has quietly throttled progress across the robotics industry: the shortage of high-quality training data for dexterous, real-world tasks. Built around a dual-arm wheeled data-collection platform and a high-precision data glove, the system captures robot execution data, human demonstrations and simulation data, and, notably, EEG readings from the human operator, recording not just what someone’s hands are doing but what their brain told those hands to do.

“A decade of BCI research has given us the ability to decode what a person intends to do and translate that into machine action,” said Nyx He, Partner and Senior Vice President of BrainCo. “By integrating brain-computer interfaces, AI, and embodied AI, we believe it will define the next chapter of human-machine collaboration.”

BrainCo, founded in 2015, also showed off its Revo 3 Dexterous Hand, a 21-degree-of-freedom robotic end-effector with full-palm tactile sensing and 70N grip force, alongside its Intelligent Bionic Hand and Intelligent Bionic Leg prosthetics. The booth drew a visit from Alexander De Croo, Administrator of the United Nations Development Programme.

Pudu Robotics: One Brain, Many Bodies

Pudu Robotics, which Frost & Sullivan ranks first globally in both revenue and shipment volume for commercial service robots, used its WAIC technology briefing to detail “One Brain, Multiple Embodiments”, a philosophy of letting different robot bodies share the same underlying intelligence rather than developing each form from scratch.

The architecture has three layers. At the base sits the robot body itself, specialised robots for high-frequency, standardised jobs; semi-humanoid robots for complex industrial settings; and humanoid robots for open-ended collaboration. Above that sits PuduAgent OS, a system layer using vision-language models, memory modules and safety guardrails to let different robot bodies share skills and experience through one unified “brain.” At the top is PuduFM, Pudu’s embodied intelligence foundation model, which combines a physical vision-language module with a “Physics Intuition Model” and action-expert components so robots can understand their surroundings, plan tasks and adapt from feedback rather than follow pre-programmed rules.

“By building a unified intelligence foundation, we enable different robot forms to share the same ‘brain,’ allowing robotics to evolve from standalone intelligence to a platform-based intelligence ecosystem,” said Shawn Wu, General Manager of the General Embodied AI Product Line at Pudu Robotics.

The architecture’s most tangible expression at WAIC was the PUDU D7, an industrial-grade semi-humanoid robot making its first public offline appearance. Powered by PuduFM, the D7 can carry payloads up to 14 kg and reach heights of 2 metres, targeting material handling, high-reach picking and warehouse logistics. Pudu says it has shipped more than 130,000 robots to over 85 countries and regions to date, and holds the top global position in commercial cleaning robotics revenue.

A Common Thread

Different as their technologies are: neural interfaces, mixed robot fleets, unified AI architectures, the three companies converged on a similar message at WAIC 2026: the next phase of robotics isn’t about building the single perfect humanoid, but about making intelligence portable across many kinds of machines, and proving it can hold up in real commercial settings rather than staged demos. With China’s humanoid robot industry now entering what organisers describe as a phase of more standardised, regulated growth, WAIC’s exhibition floor offered an early look at what that maturing industry actually looks like in practice.

 

Misschien vind je deze berichten ook interessant