WRC 2026: Humanoid Robots Shift from Stage Performances to Real Work
At the World Robot Conference (WRC 2026), which opened August 19 at the Beiren Yichuang International Convention and Exhibition Center in Beijing's Yizhuang district, the keyword changed from "can dance" to "can work." Over 300 domestic and international companies exhibited more than 3,000 products, including 311 new releases — but what attracted capital and media was not another backflipping robot. It was Robot Era (星动纪元), which after running for months in SF Express and China Post logistics centers, was ready to sit down with customers and discuss ROI.
Authoritative reports released at the conference show China shipped over 40,000 humanoid robots in the first half of 2026, raising its share of global shipments to 97%. Morgan Stanley forecasts 50,000 units for the full year in China and more than 440,000 by 2030; the Chinese Institute of Electronics estimates the 2030 market at roughly 870 billion yuan. Both point to the same conclusion — the mass-production inflection point has arrived.
Capital Markets Did the Math First
Unitree Robotics (宇树科技) listed on the STAR Market on August 19, opening at 1,100 yuan (up 629.44% from an IPO price of 150.80 yuan) and closing at 845 yuan, up 460.34% intraday, with a market cap exceeding 340 billion yuan. The retail subscription abandonment rate was just 0.09% — the lowest since China's registration-based IPO system began — and institutional investors had zero abandoned subscriptions.
But the first-day fade was striking: the close was 255 yuan below the open, the largest intraday retreat among recent IPO debuts. Fundamentals: Q1 2026 revenue of 423 million yuan (+68.49% YoY), H1 revenue of roughly 1.152 billion yuan (+48.54% YoY), and net profit attributable to shareholders of 274 million yuan (versus a loss of 32.02 million yuan a year earlier). The swing from losses to profit coincides exactly with embodied AI moving from prototype demos to mass production and delivery.
It's not just Unitree. Leju Robotics had its IPO application accepted by the Shenzhen Stock Exchange in May under the ChiNext's fourth set of criteria (for unprofitable companies); Dobot passed its listing review on July 22, planning to raise about 1.2 billion yuan for multi-legged and humanoid robots. The embodied-AI IPO window has shifted from proof-of-concept to revenue and shipment volume.
Where the Money Comes From
In H1 2026, embodied-AI funding in China totaled 93.5 billion yuan, five times higher than the same period last year. With the money come questions: can robots actually work, and where?
Reporters from TMTPost's STAR Market Daily surveyed the WRC floor by scenario. Logistics, production lines, inspection, and commercial services all had working robots on display, but large-scale deployments remain rare. ROI that actually adds up exists today mainly in logistics.
Robot Era co-founder Xi Yue put it bluntly: "More and more customers are willing to genuinely discuss ROI." Logistics scenarios have clear requirements for tempo, accuracy, and working hours; meet them with acceptable ROI, and clients pay. Deployment time for its solution at a new site has shrunk from about two months to under a week. On display were the full-size bipedal humanoid Robot Era L7 and the fully direct-drive dexterous hand XHAND 1 — the latter with 12 fully active degrees of freedom, 80N maximum single-hand grip force, the ability to lift 25 kg, and more than 100 operations such as peeling express delivery labels and driving screws.
MagicAtom (魔法原子) recreated a real warehouse workflow at its booth: wheeled humanoid D1 handled machine loading/unloading and cross-zone transfer; XG Z1 used 3D vision and AI multimodal perception to grasp and sort cluttered, stacked materials; AGV sorting robot HyperBot handled final sorting, achieving a 99% real-time multi-SKU sorting success rate. XG Z1's 6-second end-effector quick-change system aims to break the traditional industrial-robot rigidity of "one machine, one use."
Scenario Events Covered Humanoids' "Work Capability Checklist"
From August 22–26, the second World Humanoid Robot Games were held at the National Speed Skating Oval ("Ice Ribbon"): 666 teams from 16 countries across six continents and 2,056 robots competed — team count up 138% from the first edition, robot count quadrupled. Events grew from 26 to 51: 30 competitive and 21 scenario-based.
Forbes noted that this event schedule is the real highlight of this year's WRC — it clearly marks which capabilities are mature enough for full autonomy and which still face high technical barriers requiring human teleoperation. The scenario events sketch out China's actual demand for humanoid deployment, closer to the real battlefield than any consulting report.
The Real Breakout Is Still Ahead
Unitree founder Wang Xingxing offered a judgment at the conference: the "ChatGPT moment" for embodied AI will arrive in two to three years at the fastest, five to ten years at the slowest. The trigger: robots deployable into arbitrary unfamiliar environments, such as homes, completing about 80% of tasks.
Between now and that threshold stand three hurdles: generalization capability of embodied-AI foundation models, the fine durability of key hardware (reducers, torque sensors, dexterous hands), and the pace of scale-up in vertical industry scenarios. Analysts widely agree that foundation-model capability and hardware durability remain the key unsolved problems, and vertical-scenario volume still needs a one-to-two-year observation period.
But the financial value has already run ahead — Unitree's 460% first-day gain and 93.5 billion yuan in half-year embodied-AI funding are pricing in "things two to ten years out." The key observation point for the next 12 months is simple: whether the "working" robots shown at WRC can leave a line on customers' actual invoices that clearly reads "labor cost saved."
Sources: Xinhua/Economic Information Daily, CCTV, STAR Market Daily, Securities Times, China Economic Net, Forbes (The European edition), and republished Wall Street Journal reporting.