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.
Against the backdrop of global discussions about whether China is experiencing a deficit or surplus of computing power, China's semiconductor center in Zhangjiang, Shanghai, presents a clearer picture: stable and universal computing power remains in acute shortage, with orders for certain chips scheduled for next year and beyond.
Sun Guolian, Product Director at the GPU manufacturer MetaX (MXC series), noted that the shortage is observed regarding universal and reliable computing power. However, there is a surplus in non-universal architectures that can only operate with outdated or highly specialized models. Chips from this company's MXC600 series are supplied in large batches, and next-generation product development is in advanced stages.
Parallel to traditional electronic computing, Zhangjiang positions itself as China's national silicon photonics center. The Shanghai Zhangjiang silicon photonics cluster now unites over twenty companies covering the entire production cycle. Among them stands Lightelligence—the world's first company producing AI silicon photonic chips, which has been listed on the Hong Kong Stock Exchange.
Shen Yichen, founder of Lightelligence, who holds a Ph.D. from MIT and has previously published in the journal Nature, forecasts that silicon photonic chips will account for over 30% of the intelligent computing center market within five years, compared to less than 3% today. The company has already implemented hybrid optoelectronic solutions in three optical interconnect clusters of a thousand cards, enabling the operation of over 5000 GPUs for 44 corporate clients, including research institutions, internet companies, GPU manufacturers, and AI infrastructure operators.
The Zhangjiang 895 incubator has attracted several developing photonic startups, such as Qisuan Guangqi, specializing in optical chip R&D, and Guangjiu Technology, focusing on photonic technologies. The cluster also includes JCET (Changdian Technology) for advanced packaging and Suzhou Yite for silicon photonics testing, ensuring full coverage of the chain from design to testing.
The Zhangjiang model represents a dual approach to solving China's computing power problem: developing traditional electronic computing to ensure universal reliability while simultaneously creating a silicon photonics ecosystem. This ecosystem potentially offers a breakthrough in optical interconnects, allowing thousands of GPUs to function as efficiently as one chip by transmitting data via optical rather than electrical channels.
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.
