Dawning Information Industry Co Ltd

Sugon released its 2026 half-year report, with first-half net profit attributable to the parent of 971 million yuan, up 33.3% year on year. Operating revenue was 7.47 billion yuan, up 27.6% year on year. Net profit attributable to the parent after deducting non-recurring items was 844 million yuan, up 48.5% year on year. Net operating cash flow was negative 187 million yuan, an improvement of 86.5% year on year. Second-quarter operating revenue was 4.27 billion yuan, up 30.7% year on year, and net profit attributable to the parent was 744 million yuan, up 37.1% year on year. The company said that amid the opportunity of deep integration between artificial intelligence and computing infrastructure, it successfully achieved a historic leap from ten-thousand-card scale to one-hundred-thousand-card scale, further consolidating its industry-leading position in high-end computing power.

On the morning of August 14, the liquid cooling server concept sector was active. Jones Tech hit the 20 percent daily limit up, Aihua Group and Kangsheng Co. also hit limit up, while Jintian Copper, Siquan New Materials, and Yidong Electronics followed higher. On the news front, China's first national standard for data center liquid cooling, GB/T48023-2026 Technical Specification for Cold Plate Liquid Cooling Systems in Data Centers, led by Sugon Data Energy under Sugon, was officially approved and released, and will take effect on February 1, 2027. The standard systematically regulates technical requirements and testing methods for cold plate liquid cooling systems and key components for the first time, covering the entire process from system design, component requirements, hydraulic calculations, testing and verification, to deployment and operation and maintenance. Guotai Haitong Securities analysis pointed out that generative AI and large model training are driving server rack power density steadily higher, traditional air cooling is gradually hitting physical limits, and liquid cooling has shifted from an alternative for high-performance computing to a core support for next-generation computing infrastructure. China continues to tighten PUE control indicators for data centers, with new large and ultra-large data centers required to gradually reduce PUE to below 1.3, and energy-saving policies are shifting from advocacy to mandatory access constraints, forcing intelligent computing centers to increase liquid cooling deployment. Multiple institutions view the third quarter of this year as a watershed for industry performance. Zheshang Securities judges that large-scale shipments of Nvidia VR200 cabinets are the core driver, and upstream component makers are expected to receive concentrated batch orders in the third quarter. Overseas leaders' earnings also released positive signals. Vertiv reported second-quarter revenue of 3.274 billion US dollars, up 24 percent year on year, adjusted operating profit jumped 51 percent year on year, and backlog reached 15 billion US dollars. The company raised its full-year guidance and said liquid cooling order demand is strong.

Sugon released its 2026 interim results flash, with net profit attributable to the parent up 34.19% year on year to 978 million yuan. Total operating revenue was 7.429 billion yuan, up 26.99% year on year. Operating profit was 1.119 billion yuan, up 37.19%. Total profit was 1.12 billion yuan, up 38.4%. Net profit attributable to the parent after deducting non-recurring items was 853 million yuan, up 50.01%. Total assets at the end of the reporting period were 43.286 billion yuan, up 5.7% from the beginning of the period. Equity attributable to the parent was 23.059 billion yuan, up 3.7% from the beginning of the period. The company said the profit growth was mainly driven by focusing on its core business, deepening product deployment in intelligent computing, and higher investment income from improved operating performance at its investee subsidiaries.

MediTech and Sugon have signed a long-term strategic cooperation agreement to jointly promote the localization of AI healthcare and build a closed loop covering data, models, computing power, scenarios, and hardware. The two sides will collaborate on adapting large models for healthcare and elderly care, system integration, and product development, creating standardized and replicable integrated solutions that merge medical expertise with intelligence. MediTech has been deeply engaged in healthcare informatization for over two decades, serving more than 2,400 medical institutions nationwide across core clinical scenarios such as emergency, anesthesia, operating rooms, and ICUs, and has accumulated a high-quality clinical data resource pool. This cooperation combines Sugon's domestically developed computing power foundation with MediTech's clinical scenarios and data barriers, forming a full-stack delivery capability from underlying computing power to upper-layer applications, with plans to expand into diverse businesses such as smart education and smart cities.

Alibaba's Zhenwu M890 supernode has successfully adapted to Qwen3.8 and gone live on the Alibaba Cloud Bailian platform to provide model inference services, becoming the first supernode in China to successfully run a large model with over 2 trillion parameters. Dawning Information Industry said its supercluster and supernode products mainly include scaleX640 and scaleX40, among which scaleX640 adopts a high-density architecture with one unit driving two, featuring 640 cards per cabinet and a 20-fold increase in computing density, capable of forming a thousand-card computing unit through two machines, while scaleX40 uses an orthogonal cable-free first-level interconnect architecture with 40 GPUs integrated per node. Digital China said it released its Digital China KunTai supernode server product in February 2026, mainly targeting government, enterprise, and industry customers, and progress is currently smooth. Vantone Development said whether the 950 supernode uses its products depends on the downstream server manufacturers' own needs. EmbedWay Technologies said its intelligent computing supernode research and development is based on orthogonal architecture technology accumulated over years in network visualization products, independently mastering core technologies such as structure, high-speed signals, heat dissipation, power supply, and management, with high technological maturity, and is currently engaged in product customization development and early market cooperation with partners. Huatai Securities' latest research report believes that the World Artificial Intelligence Conference's concentrated display of various supernode solutions indicates that supernodes have become an important development direction for domestic computing infrastructure, and system-level competition will drive up the value of high-speed interconnects and complete cabinet supporting components, suggesting attention to segments such as switch chips, switches, copper connections, optical modules, and complete cabinets.

At WAIC, multiple domestic manufacturers showcased their next-generation supernode products. The Bosera STAR 100 ETF recorded a single-day net inflow of 116 million yuan. Huawei exhibited the Ascend 950 supernode offline for the first time, featuring a single cabinet with 64 cards as the basic unit, enabling high-speed interconnection of 1,024 Ascend NPU cards. Alibaba Cloud officially released the Lingjun Zhenwu M890 supernode instance, now open for trial invitations in the Ulanqab region. Sugon showcased China's first fully domestic 100,000-card AI supercluster, the Sugon 8000. Kaiyuan Securities believes that as domestic chips begin mass production, 2026 could become the inaugural year for domestic supernode volume shipments, potentially driving iterative upgrades across the entire AI industry chain. As of 1:58 PM on July 20, 2026, the SSE STAR 100 Index fell 5.65%, while the Bosera STAR 100 ETF dropped 6.00% to 1.58 yuan. The tracked index's latest price-to-earnings ratio stands at the 3.85th percentile over the past year, indicating historically low valuations.