At the WAIC 2026 exhibition, computing clusters featuring domestic AI supernodes are being demonstrated, utilizing Sugon Dawn and Hygon processors and systems scaled up to 100 thousand cards. This indicates a shift in focus within AI computing from competing individual chips to the system level.
The World Artificial Intelligence Conference 2026 signals a fundamental change in the narrative surrounding Chinese AI computing. Instead of comparing individual chips against NVIDIA GPUs, the focus has shifted to systemic computing using domestic supernode clusters capable of competing with the world's best AI infrastructure.
A central exhibit illustrating the transition of Chinese AI computing into a systemic era is Sugon's Dawn 8000 super-AI cluster. This 100,000-card cluster is built on Hygon processors and connected to the National Supercomputer Internet.
Numerous local suppliers are presenting supernode-scale solutions at WAIC. Sugon showcases its super-AI architecture, which integrates computational precision from FP64 to INT8 within a single cluster, eliminating the traditional separation between scientific computing infrastructure and AI training. Hygon displays its latest high-performance processors, specifically designed for large-scale AI tasks, featuring improved memory bandwidth and inter-chip connectivity capabilities.
Other chip manufacturers are also presenting latest-generation products optimized for supernode deployment. The emphasis of all participants is no longer on matching or exceeding the performance of a single NVIDIA H100 or B200 chip, but on demonstrating that domestic chip clusters can achieve competitive training throughput at the system level for models up to one trillion parameters when paired with the correct architecture featuring optimized network, storage, and software stacks.
WAIC also features several next-generation AI computing innovations, such as liquid-cooled supernode racks, disaggregated memory architectures for large model training, and native domestic software stacks ensuring high utilization of heterogeneous accelerators. Some exhibitors are demonstrating training sessions on clusters exceeding 10,000 cards, using exclusively domestic hardware and software, while showing strong scaling efficiency.
This systemic focus reflects a mature understanding that leadership in AI computing is determined by the integrated performance of the entire infrastructure chain, not by the characteristics of individual chips. Network bandwidth, memory architecture, cooling efficiency, power delivery, supported software toolkits, cluster management software, and application optimization contribute more to actual training throughput than the raw performance of the chip itself.
This structural view directs the development of Chinese AI computing toward holistic system optimization rather than component-level competition. Industry observers note that this mirrors the trajectory of global AI computing, where even NVIDIA transformed from a GPU manufacturer into a company creating data center systems with DGX and HGX platforms. The supernode demonstrations at WAIC show that China's domestic AI computing ecosystem has reached a stage where systemic competition is feasible, even if the performance of individual chips still lags behind cutting-edge foreign processors in peak specifications.
Sugon Company has introduced Dawn 8000, which is the first fully domestic artificial intelligence supercluster in China consisting of 100,000 cards. This cluster integrates high-performance computing and AI computation on a unified architecture connected to the national computing network.
At the Photosynthetic Organization 2026 Intelligent Computing Application Conference, held on July 10th, Sugon announced the completion of Dawn 8000 in Dafen. This cluster has been integrated into the national unified computing network from its first day of operation and is connected to the National Supercomputer Internet.
Dawn 8000 marks the first production-level implementation of a native super-AI fusion architecture. This architecture differs from the traditional segregated approach, which resulted in two types of computing power: clusters for high-precision scientific work that struggled with core AI tasks, and AI-oriented clusters that lacked the precision for scientific calculations. By unifying FP64 precision through INT8 on a single platform, Dawn 8000 can handle tasks ranging from high-precision scientific modeling to low-precision AI training and inference.
The entire cluster is built using domestic components. The basic computing power is provided by Hygon processors, while high-speed internal communication among 100 thousand cards is guaranteed by the scaleFabric IB-like native RDMA networking solution. The massive volume of data is processed by the ParaStor distributed storage, which received two awards in the Global IO500 production list for 2026. Furthermore, the system supports megawatt-level power density per rack thanks to advanced liquid cooling technology.
Unlike vertically integrated solutions that restrict users to a single vendor's ecosystem, Sugon chose an open computing approach. This allows for the use of various brands of AI accelerators and core AI software ecosystems, giving users freedom in choosing hardware and migrating workloads. Despite the greater complexity of integration, this method proved more convenient for developers at scale.
Dawn 8000 has already enabled the optimization of over 300 super-AI fusion applications across more than 20 fields, including robotics, automotive industry, new drug development, new materials, and quantum computing, as well as large language models. More than 70 applications have been successfully scaled up to 100 thousand cards. The cluster has demonstrated breakthrough results in solving global research problems, such as protein folding simulation, dynamics of water molecules in trillions of atoms, and simulation of turbulent flow on a grid of 100 trillion cells.
Sugon also announced the start of developing a second fully domestic super-AI fusion system with 100,000 cards in collaboration with the Beijing Academy of Science and Intelligence. This indicates that the super-AI fusion architecture is becoming the standard template for future AI infrastructure in China.
China's AI computing industry is undergoing a historic transformation as major internet giants begin actively adopting domestic chips. According to an in-depth analysis by TMTPost, the share of domestic Graphics Processing Units (GPUs) in new deployments has reached 41%.
In mid-June, reports emerged that ByteDance was negotiating to purchase at least 50,000 AI chips from Tianshu Zhixin, primarily for inference tasks, utilizing the Zhikai and Tianai GPU series. Shortly thereafter, it became known that Alibaba was also considering Tianshu Zhixin as a potential client.
The market reacted very strongly. On the day of ByteDance's procurement announcement, Tianshu Zhixin shares listed in Hong Kong jumped nearly 15%, reaching a record HK$887.50 within three days. Meanwhile, Baidu's subsidiary, Kunlun Chip, plans an IPO in Hong Kong with a target valuation of approximately $50 billion USD, which is more than 15 times its previous valuation of 21 billion yuan after the Series D round in July 2025.
This rapid transition to domestic chips is driven by export controls imposed by the United States. NVIDIA CEO Jensen Huang publicly acknowledged that NVIDIA's share in the Chinese market has dropped from 95% to zero, as even heavily restricted H20 chips stopped being supplied in 2025. This supply shortage coincided with explosive growth in demand for AI computing. Daily token requests in China increased from 100 billion at the beginning of 2024 to over 100 trillion by the end of 2025, reaching 140 trillion by March 2026.
According to IDC, out of 4 million AI GPUs delivered to China's AI server market in 2025, domestic chips accounted for 41%. Only Huawei Ascend supplied about 812,000 units, while Baidu Kunlun Chip and Cambricon each provided approximately 116,000 units. In May 2026, China's information security authorities first included AI chips in the security verification system, assigning the highest Level I rating to nine domestic chips for training and inference.
Internet giants are employing various approaches. ByteDance, whose assistant Doubao reached 368 million monthly active users and 120 trillion daily token requests, adheres to a pragmatic strategy: using Huawei Ascend for training and adopting a dual-sourcing strategy, including Cambricon and Tianshu Zhixin, for inference. ByteDance's budget for AI infrastructure in 2026 was increased by approximately 25%, totaling 200 billion yuan, with domestic chip purchases expected to exceed 40 billion yuan.
Baidu has chosen the most vertically integrated path. Its subsidiary, Kunlun Chip, which originated from an FPGA accelerator project launched in 2011, released its third-generation P800 chip in 2024 and built the first AI cluster in China with 32,000 cards, entirely assembled from domestic components, in 2025. Kunlun Chip won a centralized tender worth one billion yuan from China Mobile, marking its first major external contract. Baidu's strategy represents the most comprehensive model of self-sufficiency, covering the entire cycle from design to deployment.
