Analysis of the Differentiation Between China and US AI Development Paths and the Long-Term Value of RMB Assets

The divergence in AI development paths between China and the U.S. is obvious: the U.S. takes a closed-source route with heavy investment, while China relies on cost advantages and an open-source ecosystem to reshape the industry’s cost structure; the heavy-asset model is subverting traditional high-margin software valuation logic, and current AI investments face bubble risks; China’s core barriers lie in low computing power costs, rich industrial application scenarios, and sound infrastructure; cost control and application implementation capabilities will be the decisive factors, which may lay the foundation for the long-term value of RMB assets.

• Path Differentiation
  : By 2025, the divergence in AI paths between China and the U.S. will be significant. The U.S., represented by OpenAI and others, adopts a closed-source model with heavy investment, relying on huge capital and super computing clusters; China, represented by DeepSeek, Qwen, etc., uses an open-source low-cost strategy, reducing computing power requirements through engineering optimization.
• Investment Scale
  : In 2024, U.S. private AI investment reached $109.1 billion, while China’s was only $9.3 billion (a gap of about 12 times); U.S. tech giants’ 2025 AI infrastructure investment is approaching $400 billion, and OpenAI plans to invest $500 billion in the StarGate project.
• Cost Advantage
  : The training cost of China’s open-source models has been significantly reduced (e.g., DeepSeek only costs $4.6 million, with performance comparable to GPT-5); China’s total intelligent computing scale is 788 EFLOPS, ranking second globally, and its configuration is optimized through the “Eastern Data, Western Computing” project.
• Bubble Risk
  : Multiple Wall Street executives have warned of overheating AI investments; the five major tech giants’ AI capital expenditures are expected to reach $371 billion, but returns are questionable, and the heavy-asset model subverts traditional software valuation logic; Hong Kong stock valuations are at a historical low.

The differences in AI paths between China and the U.S. form a complementary competitive pattern: the U.S. leads in technology but faces high costs and large bubble risks, while China’s cost advantages and rich application scenarios may become long-term advantages; RMB assets are influenced by AI concepts but have low valuations, and their long-term value depends on cost control and implementation capabilities.

• Risks
  : The bursting of the U.S. AI bubble may spread globally, and the progress of technology commercialization is slow; rapid growth in capital expenditures does not match returns.
• Opportunities
  : China’s AI enterprises have prominent cost advantages and broad industrial application scenarios; RMB assets have great potential for valuation repair.