A New Era in Global AI Competition

The latest AI Index Report from Stanford University paints a revealing picture of the shifting global technological landscape. The analysis indicates that China and the United States have reached a stage of remarkably comparable capability in the development of cutting-edge AI models. In 2025, the U.S. produced 50 notable models, while China contributed 30. Significantly, leading models from institutions across both nations, including Alibaba and DeepSeek, now occupy the same tier within authoritative AI capability ranking systems, signaling a substantial narrowing of the technological gap.

Multidimensional Comparison of U.S. and Chinese Capabilities

  • Leading Model Producers: In the global ranking by number of significant models produced, Alibaba, DeepSeek, Tsinghua University, and ByteDance all place within the top ten, demonstrating robust R&D vitality in both applied and academic spheres within China.
  • Investment Disparity: Private investment in AI within the U.S. reached $285.9 billion, vastly exceeding China's $12.4 billion—a gap of over 23 times—highlighting divergent market-driven approaches.
  • Academic Influence: Regarding the share of citations, a key metric of research impact, Chinese AI papers contributed 20.6% of global citations in 2024, compared to 19.5% for Europe and 12.6% for the U.S., indicating growing international recognition of Chinese research.
  • Intellectual Property Dominance: China holds a commanding lead in patents, accounting for 74.2% of the world's AI patents, with the U.S. second at 12.1%, underscoring China's deep strategic positioning in technological IP.

The Mounting Energy Challenge of AI

A critical warning within the report points to the escalating energy consumption driven by AI's rapid advancement. The aggregate electricity demand of all global AI systems has now risen to a level comparable to the total national power consumption of countries like Switzerland or Austria. A more striking comparison is that this energy footprint is approximately equivalent to 50% of the total electricity demand of global Bitcoin mining operations.

Even excluding cryptocurrency mining, total global data center power demand stands at around 47,000 MW. Within this, the share of power consumed by hardware dedicated specifically to AI computation is continuously and rapidly increasing. This has prompted deep reflection within the industry on the sustainable development path for AI technology. Balancing increased intelligence with controlled energy consumption has emerged as one of the central challenges for the next phase of technological evolution.