China Tech Today: The Race for Robot Brains — November 2, 2026

China’s humanoid robots can already run, dance and do backflips. The weekend’s news says the next contest is not about the moves — it is about the mind inside the machine — and it arrives alongside record EV deliveries, Apple’s long-awaited China AI launch and a humanoid built with Meta.

After years of building humanoid robots that can run, dance and perform backflips, Chinese companies are pouring resources into a different challenge: the AI that could make those machines useful across a wider range of jobs. That is the picture in a Reuters report published November 2, and it reframes the past month’s robot boom — Unitree’s surging shares, AgiBot’s 110-robot dance — as a prelude. Hardware, the report argues, is no longer the scarce part. Intelligence is.

The money has followed that conclusion. Chinese embodied-AI companies raised 93.5 billion yuan (about $13 billion) across 322 deals in the first half of 2026, five times the amount a year earlier, according to data firm ITjuzi, cited by Reuters.

Why is the brain so hard? Unlike chatbots, which learn from internet text, robots need examples of physical interaction: gripping objects, handling flexible wires, recovering from a missed grasp. Spirit AI has hired around 1,000 people to record everyday and industrial activities; Ant Group’s robotics unit Robbyant says it trained a model on 160,000 hours of material. Real deployments are appearing, with limits attached: CATL reports Spirit AI robots connecting flexible wires in battery-pack testing with insertion success above 99%, while Robbyant’s robots handle orders in two Shanghai pharmacies with staff checking medicines before dispatch.

The caveats are substantial. American labs still lead in some advanced robot-AI systems, Chinese firms face shortages of advanced Nvidia chips under US export controls, and IPO funding is being squeezed amid concern over inflated valuations. Unitree founder Wang Xingxing said he would call a breakthrough reached when robots complete 80% of tasks in 80% of unfamiliar environments from spoken instructions — and one consultant warned that even if demonstrations arrive soon, a customer being able to count on a robot is “years, not months” away. Source: Reuters via SRN News — https://srnnews.com/chinas-robot-makers-have-mastered-the-moves-now-comes-the-race-for-brains/

October EV deliveries split the field: Leapmotor, XPeng and Xiaomi surge, NIO and Li Auto slip

October delivery figures reported November 1 show China’s EV startup field pulling apart. Leapmotor delivered 106,127 vehicles, its fourth consecutive month above 100,000 and up about 51% year on year. XPeng delivered a record 50,237, including a record 13,326 overseas deliveries, up 166%. Xiaomi delivered more than 50,000 vehicles for the first time in a single month. On the other side, NIO delivered 34,246, down 15.23% year on year, and Li Auto delivered 29,025, down 8.63% and its sixth consecutive year-on-year monthly decline. The wider context is sobering: the China Passenger Car Association expects full-year passenger vehicle retail sales of 18.5 million units, down 22% year on year. Scale and overseas sales are becoming the dividing line — Leapmotor now builds its B10 in Zaragoza, Spain. For what running costs look like here, see our EV vs hybrid guide — https://futuretechpulse.ca/ev-vs-hybrid-true-cost-ownership-canada/ — Source: BigGo Finance — https://finance.biggo.com/news/bcb7df48-76fb-4b8d-8b49-8a95038f8a6c

Apple Intelligence finally reaches China — powered by Alibaba’s Qwen

Apple will begin testing Apple Intelligence in mainland China on November 4, integrating Alibaba’s Qwen model into Siri, Writing Tools and Visual Intelligence, Apple announced October 31, according to TechNode. The features arrive first through an iOS developer beta, ahead of a wider release with iOS 27.3 in December. Users will be able to ask Qwen questions about text, images and documents from inside Apple’s system features, while Writing Tools will also generate original content through Qwen. The initial rollout includes live translation, photo cleanup, Genmoji, natural-language photo searches and summaries of messages, notifications and recorded conversations. Apple Intelligence has been available elsewhere for over a year, and China was the conspicuous gap, held up by requirements that AI features use approved domestic models. Source: TechNode — https://technode.com/2026/11/02/apple-confirms-qwen-integration-for-china-ai-rollout-with-beta-testing-starting-november-4/

Meta’s humanoid “Max” was built in three months — with Shanghai’s Fourier

Meta has developed a humanoid robot prototype with Shanghai-based robotics company Fourier and is already working on a second version, according to an October 31 report by The Information, covered by TechNode on November 2. The prototype, internally called Max, stands about five feet tall, has speakers in its chest, a face capable of mimicking human expressions and five-fingered hands that can pick up small objects — and the two companies reportedly built the first unit in just three months. Meta says the collaboration is for internal research and experimentation, with no plans to sell the units and no launch date for a consumer robot. The speed is the story, and so is the geography: Meta’s robotics team travelled to China to work with Fourier’s engineers, and the project is described internally as giving Meta’s Muse AI agent a physical body. Source: TechNode — https://technode.com/2026/11/02/meta-reportedly-builds-humanoid-robot-prototype-with-chinas-fourier/

Wuzhen opens with open source — and China’s AI hardware — centre stage

The 2026 World Internet Conference Wuzhen Summit opened November 2 in Wuzhen, Zhejiang province, running to November 5 under the theme “Open Source and Openness, Collaborative Development and Shared Benefits.” According to the conference’s own announcement, the summit features an opening ceremony, a main forum and 27 sub-forums covering robotics and embodied AI, spatiotemporal intelligence, trusted and interoperable AI agents and green development — with an “AI Wuzhen Dialogue” bringing together leaders of global tech companies, open-source model developers and AI scientists. The hardware on show underlines the theme: a November 2 briefing reports four domestic AI supernodes exhibited side by side at the Light of Internet Expo, from Alibaba, Huawei, ZTE and Sugon. China is presenting both the models and the machines to run them as a domestic stack. Source: World Internet Conference — https://www.wicinternet.org/2026-09/30/c_1217010.htm

Meituan’s LongCat-2.5 targets long, agentic coding jobs

Meituan released LongCat-2.5 on November 2, adding stronger visual understanding and coding capabilities to its AI model lineup. The model has 1.6 trillion total parameters, with approximately 48 billion activated for each token, and is designed for tasks requiring sustained execution by AI agents — the long, multi-step work where models typically drift or stall. Meituan says it expanded training on long task sequences, used large-scale reinforcement learning for complex codebases and whole software projects, and converts failures encountered by real agents into reproducible training tasks. The model is available through Meituan’s chat service and API platform, with GitHub and Hugging Face releases planned to follow. The frontier of model competition is shifting from single answers to agents trusted with hours-long jobs; the planned open releases will let independent developers verify the claims, which until then are the company’s word. Source: TechNode — https://technode.com/2026/11/02/meituan-launches-longcat-2-5-multimodal-ai-model-for-extended-coding-tasks/

GenRobot AI raises $100 million to supply the robot-training data bottleneck

Beijing-based GenRobot AI has completed a $100 million Series A+ funding round, disclosed November 2, with SAIC Motor, Baidu, Malaysia’s Eastworlds, Shoucheng Holdings and Galaxy General Robotics among the strategic backers. The round closed in the third quarter. GenRobot builds equipment and software for collecting human demonstrations used to train robots: its recording systems capture visual information together with hand and body movements across household, manufacturing and logistics tasks, and its Matrix platform synchronizes, cleans and labels those recordings. The round is a direct bet on the bottleneck in today’s lead story — robots do not lack motors, they lack good training data. Source: TechNode — https://technode.com/2026/11/02/robot-training-data-startup-genrobot-ai-raises-100-million-in-series-a-funding/

The big picture

The weekend’s thread runs from the factory floor to the phone in your pocket: China is shifting from showing off what its machines can do to proving what its AI can make them do — while its EV leaders consolidate, its chip and compute stack goes on display at Wuzhen, and Apple finally brings its AI to China through a Chinese model.

Apple Intelligence beta testing begins in mainland China on November 4 — the first real-world read on how Qwen performs inside Siri, Writing Tools and Visual Intelligence.

Sources

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top