中金：中美AI投资“差”在哪？

Overall Landscape - The United States has a first-mover advantage in AI, while China is rapidly catching up. The U.S. faces a power shortage, while China has a chip shortage, with limited differences in models, especially open-source models [1][17]. Investment Scale - Both the U.S. and China have similar investment intensities, each accounting for approximately 3% of GDP. However, when considering infrastructure, China's investment intensity is stronger at 6% compared to the U.S. at 4.6% [1][40]. Funding Sources - U.S. AI investments are primarily driven by the private sector, amounting to $552 billion, with limited government investment of $11 billion. In contrast, China's private sector investment is only one-sixth of the U.S. at $90 billion, but government investment is significantly higher at $75 billion [1][51]. Investment Focus - The U.S. allocates more funds to data centers and supporting infrastructure, accounting for 83% of investments, with 5% in chips and 12% in models. Conversely, China invests more in models and chips, with 70% in data centers, 8% in chips, and 22% in models [1][40]. Implications of Differences - The U.S. private sector's focus on returns leads to concerns about bubbles, while China's government-driven investments allow for greater resource allocation without immediate financial returns, focusing more on chips and models [1][68]. AI Contribution to Economic Growth - AI has significantly contributed to economic growth, with the U.S. tech hardware and software investments contributing one-third of GDP by 2025. AI's impact on stock markets is also notable, with major tech stocks contributing significantly to market returns [5][90]. Talent Pool - The number of AI researchers in China is increasing rapidly, with Chinese researchers holding three times as many AI patents as their U.S. counterparts by 2022, indicating a strengthening of China's research capabilities [36][121]. Infrastructure and Chip Development - The U.S. leads in infrastructure but faces power supply challenges, while China is accelerating its chip development despite still lagging in advanced processes. The U.S. semiconductor industry is projected to reach $318 billion in sales by 2024, while China's chip output is growing rapidly [23][28]. Model Development - The U.S. leads in the quantity and quality of AI models, but China has gained an advantage in open-source models, with significant downloads and usage surpassing the U.S. by 2025 [31][115]. Future Outlook - The performance of the foundational layer in AI investments is expected to be more certain, while the technical and application layers have greater potential for upward movement, driven by advancements in models and commercial applications [85][82].