On August 8, according to the industry publication Diancan (电参), Xiaomi (小米) — the Beijing consumer-electronics giant known globally for smartphones and electric vehicles — completed an internal reshuffle of its robotics division, creating a new "Embodied Intelligence and Applications" department headed by Kong Tao (孔涛), formerly head of Seed Robotics at ByteDance (字节跳动), China's largest content platform. Previously scattered research teams — including Chen Long's VLA (vision-language-action) group and Cai Rui's embodied-robotics unit — were consolidated under the new department. The signal is unmistakable: Xiaomi's embodied-AI track is shifting from competing on hardware specifications to deeply integrating software and model capability.
For readers outside China: "embodied AI" here means robots controlled by large AI models rather than hand-coded routines, and Xiaomi is one of several Chinese tech giants now betting that the winning edge is the "brain," not the "body." The same day, the broader industry produced a cluster of milestones that framed the moment.
Key takeaways
- Xiaomi (小米) consolidated its robotics teams into one "Embodied Intelligence and Applications" department under Kong Tao (孔涛), ex-ByteDance Seed Robotics.
- Unitree (宇树科技) opened its Star Market (科创板) subscription on August 10 at 150.8 yuan per share (about $21 / HK$164), implying a 609.9 billion yuan (about $84.7 billion) market value — China's first "humanoid-robot stock."
- China's embodied-AI financing hit 935 billion yuan (about $130 billion / HK$1.0 trillion) in the first half of 2026, up fivefold year on year.
- A key module that cost tens of yuan in 2025 fell to 200–400 yuan (about $28–56 / HK$217–435) in 2026, pushing "ten-thousand-yuan robots" toward mass use.
- Roland Berger (罗兰贝格) judges China's AI architecture and training data lag hardware by 3 to 5 years.
Why "unify the reporting lines"
Xiaomi's earlier robotics layout was fragmented: a VLA (vision-language-action) model team, an embodied-robotics unit, and hardware R&D each worked separately. When Lei Jun (雷军), Xiaomi's founder, posted on July 14 a video of the company's humanoid "interning" for four months in an auto factory — nut-fitting success rising from 90.2% to 98%, just one point short of the 99% qualifying rate for skilled workers — it also exposed a problem: the robot's work cycle was far slower than a human's, and the clip had been sped up three times to look smooth.
The root cause was not hardware but the "brain." Robot actuators, reducers, and sensors have reached near-commercial maturity, yet AI architecture and training-data readiness lag hardware by what Roland Berger (罗兰贝格) estimates at 3 to 5 years. Pulling the scattered model teams into one department is Xiaomi's way of concentrating firepower on the core contradiction: software behind hardware.
Who Kong Tao is, and what it means
Kong Tao comes from ByteDance's (字节跳动) Seed Robotics team. ByteDance's robotics investment drew little outside attention, but the Seed team had built real strength in VLA models. That Xiaomi hired from ByteDance rather than from a traditional robot maker shows it locates the competitive core in "AI model capability," not "mechanical engineering."
That contrasts with peer strategies. Huawei Cloud positions itself as "silicon black soil," launching the CloudRobo platform to lay foundations for all players. Tencent (腾讯) shifted from its "three don'ts" — no hardware, no mass production, no commercialization — to releasing a full-stack embodied-intelligence solution. Each large firm picks a differentiated slot based on its resources; the all-out mixed brawl across the whole track is largely over.
The real ledger of the "mass-production first year"
A China Business Journal (中国经营报) report on August 8 supplied the industry's calmest ledger. In the first half of 2026, domestic embodied-AI financing reached 935 billion yuan (about $130 billion / HK$1.0 trillion), up fivefold year on year. But Guo Renjie, CEO of Yuandian Robotics, flagged the key change: a critical module priced at tens of yuan in 2025 had fallen to 200–400 yuan (about $28–56 / HK$217–435) by 2026, pulling "ten-thousand-yuan robots" toward普及 (mass adoption).
JAKA Robotics' CEO Li Mingyang calculated customer payback at a "fairly lenient" two years, with aggressive buyers demanding 10 to 12 months. Zhang Guibing, general manager of Chery's (奇瑞) MoJia Robotics, predicted that as auto giants cross into the field and bring scale effects, a single robot's price will eventually match an ordinary family car.
For founders, falling hardware cost means the "use robots to deliver a service" business model is starting to work. But the real threshold is not buying a robot — it is whether you have scenario data to make it "work without error." The data assets of factories, warehouses, and operation sites are what will decide future competitiveness.
Honest limitations
This article synthesizes a Chinese tech-outlet report and a China Business Journal (中国经营报) citation; the organizational change at Xiaomi (小米) is attributed to the publication Diancan (电参) and was not confirmed by a Xiaomi primary announcement at the time of writing. The personnel moves — Kong Tao (孔涛) from ByteDance Seed Robotics, consolidation of Chen Long's and Cai Rui's teams — are reported, not independently verified against corporate filings. Unitree's (宇树科技) 150.8-yuan issue price and 609.9 billion yuan implied valuation are as stated for the August 10 Star Market subscription and fluctuate with market action. The 935 billion yuan first-half financing total and the "up fivefold" claim come from the cited journal and were not re-audited. The 200–400 yuan module-price figure is a single executive's observation, not a market index. Roland Berger's (罗兰贝格) "3 to 5 years" hardware-software lag is a consultancy judgment. Dollar and Hong Kong-dollar conversions use approximate rates (1 USD ≈ 7.2 RMB; 1 HKD ≈ 0.92 RMB) and are indicative only. No investment advice is intended.
Sources
Based on reporting by the Chinese tech outlet Naojiwang (脑机派对), 2026-08-09, citing Diancan (电参) on Xiaomi's reorganization and the China Business Journal (中国经营报) on industry financing and cost trends, alongside statements attributed to Roland Berger (罗兰贝格) and several robotics CEOs.