China Makes Historic Breakthrough in Advanced Process Domestic AI Chip

In mid-April 2026, Shanghai Dishan Technology officially announced that its self-developed 2nm AI GPU chip, Dishan Zhineo (DS-Core), has completed full-process chip design and entered the prototype verification stage. This landmark achievement fills China’s technological gap in high-end 2nm AI computing chips, marking that domestic advanced-process AI chips have officially joined the world’s top tier and fundamentally changed China’s long-term vulnerable position in high-end computing chips.

As the pinnacle of global semiconductor technology, the 2nm advanced process has long been monopolized by a small number of international giants, featuring extremely high technical barriers in chip architecture, transistor technology and high-density packaging. The newly launched Dishan Zhineo innovatively adopts a hybrid process of GAA Gate-All-Around and FinFET technology, together with Chiplet heterogeneous interconnection architecture and advanced 2.5D CoWoS-L packaging. It integrates as many as 170 billion high-density transistors on a silicon chip no larger than a fingernail, with performance matching world-leading flagship AI chips.

In terms of performance parameters, the chip delivers 50 TFLOPS FP32 precision computing power and up to 400 TFLOPS FP4 low-precision AI computing power. It fully supports large model training, cloud intelligent reasoning, supercomputer operations and other high-end computing scenarios. With a typical power consumption below 350W, its energy efficiency is 40% higher than previous generations of domestic AI chips, greatly reducing energy costs and heat dissipation pressure of intelligent computing centers. The team has conquered three critical bottleneck technologies: HBM4 high-bandwidth memory interconnection, ultra-low nanosecond inter-chip communication delay and high-efficiency microchannel thermal management. These improvements ensure stable long-term high-load operation and meet large-scale commercial AI computing demands.

For a long time, the global cloud AI computing market has relied heavily on overseas high-end GPU chips. Technology blockades, tight production capacity and long delivery cycles of advanced-process chips have severely restricted the independent and secure development of China’s artificial intelligence industry. The independent full-design of domestic 2nm AI chips breaks foreign technological monopolies, builds a fully autonomous and controllable underlying computing foundation, and accelerates China’s AI industry to reduce dependence on foreign chips and proprietary computing ecosystems.

This breakthrough reflects coordinated upgrading across China’s entire semiconductor industrial chain. From independent architecture research and advanced process design to high-end packaging and ecological adaptation, Chinese semiconductor enterprises have kept deepening core technologies. Stable mass production of mature processes consolidates industrial foundations, while advanced processes achieve continuous catching-up and overtaking. Dishan Zhineo will proceed with tape-out testing and mass production iteration, and be widely applied in cloud computing, autonomous driving, aerospace intelligence and industrial AI.

Industry experts pointed out that the verification of 2nm AI chips represents a key leap for China’s high-end AI chips from catching up to taking the global lead. Mass application of domestic advanced-process chips will strengthen national digital security, artificial intelligence and digital economy foundations. It will help China seize strategic initiative in global next-generation computing competition and provide powerful long-term impetus for the high-quality development of China’s AI and semiconductor industries.