Over the past three years, the boundaries of computing power have expanded continuously alongside advances in AI. It now encompasses not only hardware such as chips, but also algorithms and models, cloud infrastructure and software. It is further shaped by constraints in energy supply and network capacity, and ultimately only becomes meaningful when deployed in real-world applications.
China and the United States are the only two countries with the most complete computing power industrial chains globally—covering chip-related equipment, algorithms and models, cloud services and software. Nearly all major breakthroughs in global AI innovation over the past three years have emerged from these two economies. Three years ago, comparisons between China and the US focused largely on server capacity. As the definition of computing power has broadened, the comparison has evolved into a contest between two innovation ecosystems.
With a clear disadvantage in investment scale and restricted access to advanced AI chips, where does China’s computing power competitiveness come from? The answer lies in system efficiency: integrating chips, algorithms and energy into a more efficient, lower-cost system, deploying computing power across a wider range of real demand, and generating sustainable commercial returns.
