UBS analysts outline China’s journey toward AI self-sufficiency. The nation shows strength in applications and models. Nevertheless, it remains dependent on foreign titans like NVIDIA for advanced computing capabilities.

Local Champions in AI Applications

• Leading Platforms & Models: Major internet players, such as Baidu and Alibaba, along with agile startups like DeepSeek, excel in developing chatbots, recommendation systems, and visual technologies.

• Rapid Acceptance: A vast data reservoir and uniform language environment accelerate the adoption of AI across sectors like e-commerce, finance, and public services.

These components are now largely self-sufficient, as homegrown models achieve competitive accuracy and customization suitable for local contexts.

The Semiconductor Challenge

Building Self-Sufficiency

• Projected Growth: UBS predicts China’s overall semiconductor self-sufficiency will rise to 27% by the end of 2025, up from mid-teens during the pandemic. This includes DRAM, analog, and equipment sectors.

• Resilience Against Export Controls: Progress continues despite U.S. export restrictions, supported by government backing and local investments.

NVIDIA’s Competitive Edge

• Performance Disparity: NVIDIA’s GB200 chip achieves approximately 40 image generations per unit time, whereas Huawei’s 910C lags at around 13, revealing a 3x performance gap.

• Market Dominance: NVIDIA commands over 80% of the global AI compute market, while Chinese companies accounted for merely a third of domestic inference hardware in 2024.

UBS anticipates that by 2029, local compute share will grow to 90%, although this transition may involve lower pricing power and soaring production costs.

Valuation Insights

With China’s semiconductor ecosystem evolving, investors should pay attention to the valuation differences between Western leaders like NVIDIA and up-and-coming Chinese chip manufacturers. Monitoring the semiconductor stocks yields insights on how these companies are valued relative to their competitors.

Essential Takeaways

1. Strong Local Applications: Domestic AI services and models offer world-class capabilities, fostering accelerated use in commercial and governmental sectors.

2. Compute Challenges Persist: Premium inference hardware remains dominated by NVIDIA, with years needed to close the performance gap.

3. Trade-Offs Between Cost and Capability: The localization ambition enhances revenue growth but challenges efficiency and global competitiveness in pricing.

By keeping an eye on industry valuation trends and the emergence of new local AI compute chips, investors can assess China’s path toward genuine AI self-sufficiency while weighing related investment opportunities and risks.