News on November 3: at the 2026 World Internet Conference held in Wuzhen, the Huawei Ascend 950 SuperPod officially won the Leading Science and Technology Award. As a top international technology honor in the global internet field, the award is widely solicited worldwide. This year's edition received more than 430 applications from 43 countries and regions, and in the end only 20 winning achievements stood out, marking that this domestic computing power system has received highly authoritative industry recognition.
Core Architecture: Self-Developed Lingqu Interconnect Breaks Through the Trillion-Parameter Bottleneck
As the largest commercially deployed SuperPod equipment to date, the core advantage of the Ascend 950 SuperPod lies in using the self-developed Lingqu Interconnect protocol to tightly weave together a large number of computing power chips, logically achieving perfectly coordinated operation as a single super-large computer.
In terms of hardware scale, its fully configured version can hold up to 1024 NPUs, along with more than 1300 interconnect chips and over 4000 optical modules, built specifically to support pre-training and inference of ultra-large AI models at the scale of trillions or even ten trillions of parameters.
Efficiency Breakthrough: Industry-Leading Computing Power Utilization, Latency Reduced to 5 Milliseconds
Addressing common pain points in traditional large model training, such as large communication losses between chips and long waiting times, the Ascend 950 SuperPod innovatively adopts unified memory addressing technology, so that thousands of cards do not need repeated relays when retrieving data from each other, effectively reducing wasted waiting.
In terms of actual operational efficiency, the system's computing power utilization can reach 45%, about 1.8 times the industry average, which means the same hardware configuration can release more real computing power. Meanwhile, its minimum inference latency can be compressed to 5 milliseconds, perfectly fitting AI application scenarios such as agents that demand extremely high real-time responsiveness. Currently, more than 40 large models have completed native pre-training directly on this hardware platform.
Industrial Deployment and Future Iteration: Accelerating the Building of a Domestic Computing Power Foundation
As the scale of large models continues to expand and computing power demand keeps rising, the SuperPod architecture is gradually becoming the mainstream choice for new AI infrastructure. Official data shows that the Ascend series SuperPods have already been deployed in more than 1000 sets nationwide, widely covering multiple core industries including the internet, telecom operators, finance, education, and energy.
Huawei officially stated that it will continue to iterate and upgrade along the "SuperPod plus SuperCluster" technology route, building a solid, autonomous and controllable computing power foundation for the rapid development of China's artificial intelligence industry.