By NUPIAO News | Reporter: Song Jianan
On August 20, just one day after Unitree Robotics made its grand debut on the STAR Market, founder and chairman Wang Xingxing stepped onto the main forum stage at the 2026 World Robot Conference. His talk? A no-nonsense deep dive titled “From Exhibit to Product: The Next Decade of Humanoid Robotics.”
Wang’s speech was laser-focused on the company’s most talked-about products from recent years. Not a single word about stock prices or personal feelings post-IPO — it was all robots, all the time, delivered at a blistering pace.
This year happens to mark Unitree’s 10th anniversary. Kicking things off, Wang showcased the G1 humanoid — the robot that put the company on the map. It’s become the industry benchmark, and let’s be honest, a whole wave of copycats has followed.
Wang revealed that Unitree started collaborating with automotive factories back in 2024, putting robots on production lines to handle real-world tasks. But here’s the catch — those factory projects haven’t scaled up broadly, and he’s upfront about why.
“The biggest hurdle is that robots still can’t match human efficiency and versatility in factory settings. Sure, they can do some work, but compared to people, they’re just slower,” Wang said. His stance is clear: only when efficiency and adaptability improve will we see mass deployment of robots in factories.
The industry is largely chasing a data-driven playbook these days. Unitree pulls data from two streams: in-house collection and third-party sourcing. In Wang’s view, high-quality data for robots largely comes from real humans physically demonstrating tasks. Companies also need to nail the alignment between AI model outputs and physical experiments. The unpredictability of the physical world, he argues, is the fundamental difference between robots and pure software-based large models.
Wang also took the audience on a tour of the company’s hardware highlights from the past year or so. Take the GD01, a three-meter-tall pilotable mech unveiled this May. Wang was quick to clarify its positioning — think off-road gear, not a city commuter. It’s built for hauling supplies through complex terrain. Bonus feature: it can switch into a four-legged mode for extra stability, and pre-orders are already open.
Then there’s the “Superman” prototype still in testing — and this thing is fast. It clocks in at 12.65 meters per second, beating the fastest human sprint record, and its jumping height has blown past historical human limits. But hold your horses — it’s just a tech preview, not officially released.
Wheeled-legged robots are another key focus for Unitree. These are fully waterproof, built for both indoor and outdoor use, and keep the weight down. The company has been porting its humanoid tech onto these four-legged platforms, with hopes of one day taking them into consumer territory like outdoor hiking.
While hardware keeps pushing boundaries, Wang let slip that the biggest R&D bucks are actually flowing into AI models, with a major acceleration kicking off in early 2024. During the live demo, Unitree showed off a multimodal, end-to-end system that auto-generates robot actions straight from voice commands.
Wang was candid about the trade-offs: this approach introduces some inference latency, and the same command can produce slightly different movements each time. Since the AI is generating on the fly, motion smoothness falls short of pre-programmed scripts — but it’s way better at handling interference. Unitree has already started stress-testing it with real-world tasks like tidying up meeting rooms. In messy, multi-task environments, the robot gets the job done, just not at lightning speed. The old action generation was all about showmanship; the new version stacks actual work capability on top.
Wang also dropped a timeline for home robots that’s worth paying attention to: “Embodied intelligent robots truly entering households — two to three years if we’re quick, five to ten if we’re not.” In his eyes, the biggest hurdle to reaching that ‘ChatGPT moment’ for embodied AI is aligning AI model outputs with physical robots. Language models are pure digital code — input and output with almost no loss. But every physical action a robot takes introduces real-world error, and those errors pile up, chipping away at the model’s ability to generalize. That’s the core reason robots still struggle in unfamiliar environments. Still, he’s confident the problem will get solved down the road.
Wang also unveiled Unitree’s Physical AI Robot Self-Evolution V1.0 blueprint. The logic goes like this: once this self-evolving loop is fully operational, robot iteration and improvement will hit warp speed. The system reuses diverse data across multiple channels, leans on self-circulation to boost iteration efficiency, and layers in real-world deployment to push data utilization and performance growth even higher. The full closed loop spans large-model driving, task definition, code generation, simulation, real-world deployment, and result evaluation — enabling continuous skill expansion.
One particularly memorable moment: Wang proudly showed off his selection to TIME’s TIME100 AI list, complete with a magazine cover featuring him alongside the GD01 mech. The write-up called him “an atypical prophet of the AI era.”
Reflecting on the journey, Wang recalled that Unitree’s first World Robot Conference appearance was back in 2017. Nearly a decade later, the pace of robotic evolution has visibly accelerated — and he believes the industry is about to shift into an even faster gear, entering a brand-new phase of development.
Just a day earlier, Unitree had officially become the first humanoid robot stock on China’s A-share market. The IPO price was set at 150.8 yuan per share, spiking to a high of 1,100 yuan intraday — a jaw-dropping 629.44% gain. At one point, the company’s market cap blew past 440 billion yuan before closing at 845 yuan, valuing the company at roughly 341.8 billion yuan. With a combined direct and indirect stake of about 31%, Wang’s paper wealth crossed the 100 billion yuan mark, putting him at the top of the domestic post-90s entrepreneur leaderboard.
But as of this writing, Unitree’s stock has pulled back to 720.28 yuan, tumbling 14.76%, with market cap settling at around 291.3 billion yuan.
