Competition Among China’s AI Models Is Intensifying as the Chip Sector Advances

China’s AI industry is shifting from a race for model performance toward competition in cost efficiency, infrastructure and distribution, while its semiconductor ecosystem is advancing faster than expected.

• Discussions at Goldman Sachs’ Asia Leaders Conference brought together 400 companies and more than 1,400 investors across the AI supply chain.
• The performance gap between leading Chinese AI models, particularly in coding, is narrowing, increasing the importance of price-to-performance ratios and the harness layer that connects models to data and workflows.
• Chinese semiconductor progress has surprised the market: leading domestic foundries are planning more capacity expansion than analysts expected, while semiconductor-equipment companies have guided to 40% year-on-year revenue growth over the next few years.
• Bottlenecks remain in GPU availability, local foundry capacity and semiconductor-manufacturing equipment, even as more specialized AI-chip makers list publicly and disclose production volumes and technology migration.
• Optical networking is becoming a key AI infrastructure market because it enables low-latency, high-speed data exchange; Goldman Sachs expects its total addressable market to grow ninefold to $154 billion by 2028.
• Chinese open-weight models are being deployed globally, including on US hyperscalers, while AI labs pursue either frontier performance or lower costs to build durable competitive advantages.

Takeaway: Investors should look beyond model rankings toward the harness layer, price-to-performance, semiconductor bottlenecks and optical-networking infrastructure, with lasting success likely to favor companies combining cost efficiency and financial strength.