A humanoid robot can make an audience applaud by taking three steps. The harder trick is arriving at a factory on Monday, finding the right object, moving it safely, and doing that again after lunch. Deng Taihua has built his current career around the distance between those two moments. As the co-founder, chairman and chief executive of Shanghai’s AgiBot, he talks about robots less like characters in a film and more like colleagues who will eventually need a job description. The performance gets attention. The shift earns a place on the payroll.
At AgiBot’s partner conference in April 2026, Deng gave that distinction a name. He said the industry was moving from development to deployment, from machines that can move to machines that can work. It was a neat line for a conference hall full of robots, customers and partners. It was also a challenge to his own company. A robot that dances is judged for minutes. A robot that works is judged for the length of a contract.
Deng is an unusual protagonist in a field that likes its inventors visible and its executives in the background. His co-founder Peng Zhihui is widely known for hardware projects and a public maker persona. Deng came out of the vast machinery of Huawei, where careers are built around taking difficult systems to very large markets. AgiBot needs both instincts. The machines have to be clever enough to move through a messy physical world, and ordinary enough for a customer to trust them with Tuesday.
A career made for the unglamorous part
Deng studied computer communications at the University of Electronic Science and Technology of China. He joined Huawei after graduation and spent more than 20 years there, holding senior posts that included vice president and president of its wireless and computing product lines. Wireless networks and AI computing may seem distant from humanoid robots. They share a stubborn commercial fact: a system is only useful when its pieces cooperate at scale. Radios, chips, software, installers and customers all have to agree with one another. A robot adds joints, cameras and a world that refuses to sit still.
The career history helps explain why Deng keeps returning to words such as production, deployment and ecosystem. Those terms can sound bloodless beside a robot doing a backflip. Yet the person buying a machine is likely to ask how often it needs attention, what happens when it fails, and how much useful work it can finish. There is comedy in asking a humanoid to be exciting in a ballroom and boring in a warehouse. Deng’s strategy depends on winning both rooms, in that order.
In February 2023, he helped found AgiBot in Shanghai with Peng and other partners. The company introduced its first humanoid, the Expedition A1, that August. Six months is a short interval between incorporation and a machine ready for a public debut. The speed also fixed an expectation: the company would be measured by visible hardware, even as its leadership argued that the intelligence and data behind the body would matter just as much.
A body, then a business
AgiBot opened a Shanghai manufacturing facility in 2024. It broadened its machines from full-size humanoids into smaller humanoids, wheeled systems and quadrupeds. In January 2025, the company said its 1,000th general-purpose embodied robot had rolled out. By March 2026, it said the count had reached 10,000. These are production milestones, a measure of how many machines can be made. They are separate from the question Deng wants customers to answer: what do those machines accomplish once they are delivered?
The distinction matters because robotics has two kinds of spectacle. The first is the familiar one: a machine walks onto a stage and performs a task under friendly conditions. The second happens backstage, where a team must manufacture parts consistently, repair failures, update software and teach the system enough about a workplace to make it useful. There are no standing ovations for reliable inventory handling. There may, however, be a purchase order.
AgiBot’s products have been pitched for manufacturing, logistics and service settings. At its partner conference, Deng described a “one body, three intelligences” approach: the physical platform, plus capabilities for movement, interaction and work. He also said the company devotes a large share of its research staff and budget to AI. In his account, the body is shaped around the intelligence it must carry. That sounds tidy on a slide. In practice, it means hardware design, training data and customer tasks must keep adjusting to one another.
“We hope to leverage robots to solve real-world work problems and advance practical technology deployment in a down-to-earth manner.”Deng Taihua, speaking to TIME in 2026
The three curves
At the 2026 conference in Shanghai, Deng sketched a future as three curves. The first belongs to development, when a robot learns to move, perceive and perform. The second is deployment, when customers begin putting machines into real tasks. The third is a longer-range phase in which data from use could help systems improve and spread. His framework is a forecast, and its later stages remain ambitions. The interesting claim sits in the middle: he believes the industry has reached the point where useful deployments can start to grow.
That claim is more exacting than it sounds. A factory contains changes in lighting, the wrong box in the right place, a person who needs the aisle, and the sort of small interruption that never appears in a launch video. A robot does not need to solve every problem in the building to matter. It does need to solve a defined problem often enough that the people running the building invite it back.
Deng’s language has become sharper as AgiBot has grown. In 2025, his public remarks dealt with shipments, finance and commercial applications. In 2026, he emphasized deployment and productivity. The company had moved from asking whether it could build robots to asking where they could earn their keep. That is a change in audience as much as in engineering: the approving crowd gives way to the person who signs off on operating costs.
Production says the robot can leave the factory. Deployment says it has a destination. Repeatable useful work says whether it should stay.
The quiet executive in a noisy field
Deng has at times been described in coverage as rarely seen in public. That made his first major partner-conference appearance in August 2025 notable in itself. At the next gathering, in April 2026, he spoke for nearly two hours. The image is striking: an executive who had kept a relatively low public profile explaining an industry’s future to a room that included customers and partners from many countries. This is not a story of a sudden conversion to showmanship. The long speech was itself part of the job: make a technical plan legible to everyone expected to help put it into use.
His relationship with Peng offers a similarly practical picture of partnership. Both attended the same university and worked at Huawei, but their public reputations differ. Peng brings the excitement attached to engineering invention. Deng’s experience lies in organizational scale. A humanoid company needs a prototype and a supply chain, a model and a market, a person who can delight an audience and a team that can answer the telephone afterward. The division is never that simple inside a company, but it helps explain why AgiBot’s story cannot be told through a single inventor.
In August 2026, Deng appeared on TIME’s list of influential people in AI. The recognition arrived while AgiBot was talking publicly about factories, customer sites and practical applications. He told the magazine he hoped robots would solve real work problems. It was a less cinematic ambition than the cover photograph might suggest. It was also a useful way to read his career: from networks people depend on without seeing, to machines people will watch closely until they become dependable enough to stop watching.
After the applause
There are reasons to be careful with every number in a young industry. Units produced do not equal units working full shifts. A model release does not guarantee that a machine can understand a new workplace. A presentation about a future curve is a plan, not a timetable written in stone. Deng’s own preferred measure provides the stricter standard: practical deployment that creates value for a customer. That standard keeps the story open and keeps it honest.
For now, AgiBot is moving quickly. It has expanded its product range, built manufacturing capacity and attracted an international audience. Deng, who has spent much of his working life around systems that must operate beyond the lab, is trying to make humanoids pass the same test. The question is almost comically mundane for a technology that can resemble science fiction. Can it show up, do the task, and come back tomorrow? If the answer becomes yes often enough, no one will need to ask a robot to dance.