Today’s CTR: China tech’s mood over the past day is disciplined ambition with a faint whiff of spreadsheet anxiety. Huawei used its developer conference to push HarmonyOS deeper into artificial intelligence [AI] agents and open models, trying to turn an operating system into a national software stack. BYD, meanwhile, reminded the market that China’s electric-vehicle [EV] story is also an energy-storage export story, not just a car-price war. The common thread is clear: Beijing’s champions are moving from products to platforms, and from platforms to standards. The awkward bit is that scale still has to prove it can pay.
Huawei turns HarmonyOS 7 into an AI-agent operating system.
Huawei used its HDC 2026 developer conference to introduce HarmonyOS 7, positioning the system around deeper AI integration, spatial computing and an upgraded Xiaoyi assistant. The company said Xiaoyi now uses an “Agentic” self-evolving architecture, draws on cloud and on-device models, and can access 2,100 HarmonyOS system capabilities plus more than 200 categories of system-level user data.
Impact: This is not merely a feature release; it is Huawei’s attempt to make the operating system the orchestration layer for AI agents. The strategy is especially important because China’s consumer AI race is shifting from chatbots toward task completion, where control of apps, devices and user permissions matters more than demo-room wit.
Reach: Huawei’s pitch also helps explain why HarmonyOS is strategically different from another Android skin. If Xiaoyi can coordinate phones, personal computers, cars and wearables, Huawei gets a distribution advantage that model-only companies will envy and regulators will watch.
The closing thought: Huawei is trying to make the phone less like a screen and more like a dispatcher. Source
OpenPangu 2.0 puts Huawei’s model strategy in the open-source lane.
Huawei also announced openPangu 2.0, with a 512K context window and two versions: a Pro model with 505 billion total parameters and 18 billion active parameters, and a Flash model with 92 billion total parameters and 6 billion active parameters. The company said the model is optimized for Ascend computing and HarmonyOS agent tasks, and plans to open-source seven components from June 30, including pre-training code, post-training code and training operators.
Impact: The useful signal is not the parameter count, which is now the AI industry’s least charming form of peacocking. The signal is Huawei’s decision to connect model architecture, domestic chips and operating-system agents into one stack, giving Chinese developers an increasingly complete alternative to the Nvidia-CUDA-Western-cloud triangle.
Reach: Yu Chengdong’s admission that Huawei kept limited compute for itself because much of it supports other domestic companies is notable. It suggests Huawei is acting less like a single AI vendor and more like a national infrastructure supplier with a consumer brand attached.
The closing thought: Open-source, in Huawei’s hands, is also an ecosystem acquisition tool. Source
Huawei’s Hongtu plan pushes OpenHarmony across chips, devices and industries.
Huawei launched the Hongtu plan at HDC 2026, saying it will accelerate the scale development of the OpenHarmony ecosystem. The plan covers more than 20 industries, 200 chip categories and 1,200 device categories, while Huawei said OpenHarmony has more than 13,000 code contributors, over 140 million jointly built lines of code, more than 3,200 ecosystem partners and over 1.3 billion ecosystem devices.
Impact: This is the standards war in its least glamorous but most consequential form. Huawei is not only trying to win handset users; it is trying to make OpenHarmony a default software substrate for industrial equipment, consumer electronics and embedded devices.
Reach: The breadth matters because China’s technology self-reliance campaign increasingly depends on boring interoperability. A domestic operating layer that touches hundreds of chips and thousands of device types could reduce dependence on foreign software stacks in the places where consumers rarely look but supply chains do.
The closing thought: The future of China tech may be written as much in embedded firmware as in foundation-model benchmarks. Source
BYD powers Hungary’s largest battery-storage project.
BYD’s energy-storage unit said Hungary’s largest battery-storage project, built by Greenvolt Power, has gone into operation with BYD supplying a 288.6 megawatt-hour [MWh] MC Cube battery-storage system. The project has a capacity of 99.8 megawatts [MW] / 288.6 MWh and is designed for grid-level applications such as frequency regulation and peak shaving.
Impact: BYD’s story abroad is no longer just about EVs landing on European roads. Battery storage gives Chinese manufacturers a second route into Europe’s energy transition, one that is tied to grid reliability rather than consumer taste or automotive tariffs.
Reach: BYD says it has delivered large-scale storage projects across more than 110 countries and regions. That footprint gives the company a broader strategic hedge: even when car exports face political resistance, grid storage can keep Chinese battery technology inside overseas infrastructure.
The closing thought: BYD may sell cars, but batteries are the passport. Source
Beijing’s Zhiyuan AI conference turns from digital intelligence to physical-world deployment.
The 2026 Beijing Academy of Artificial Intelligence conference closed on June 13 after bringing together Turing Award winners, academics and executives from Chinese AI companies including Alibaba, Tencent, Xiaomi, Shengshu Technology and ModelBest. The event’s framing emphasized AI moving from the digital world into the physical world, a theme that neatly captures China’s current obsession with agents, robotics, embodied intelligence and industrial deployment.
Impact: The conference circuit can be noisy, but this one matters because it reflects where Chinese AI policy and capital are converging. Model capability is still important, but the prized frontier is now getting AI into factories, vehicles, devices, health care and public services.
Reach: For China, this plays to existing strengths: manufacturing density, hardware supply chains and local-government willingness to pilot new systems. The challenge is that “physical AI” turns software bugs into operational problems, which is a polite way of saying that demos will have to survive forklifts, hospital workflows and municipal procurement.
The closing thought: China’s AI race is becoming less about who talks best and more about who moves the machine. Source
