China: An overlooked AI frontier

The tech boom has created a massive valuation gap.

In Asia, global capital has chased artificial intelligence (AI) hardware in Taiwan and South Korea, which has meant China’s domestic value AI plays have been largely left out of the party. 

China’s role in the global AI boom should not be underestimated. The AI supply chain is centered around north Asia – with Taiwan and South Korea leaders in logic chips and memory – but the bulk of AI data center-related infrastructure is manufactured in China, including the racks, power supply, energy storage, cooling systems and optical networking. 

The Beijing government has done much to overhaul the country’s economy, moving away from its reliance on low-end manufacturing – and property-led growth – and towards higher-value industries such as technology and advanced manufacturing. 

China has rapidly built up its own semiconductor industry – and the gap with other countries is narrowing. China’s integrated circuit exports reached $201.9bn in 2025, representing year-on-year growth of 26.8%.

In AI software, China is now second only to the US. Its strongest advantage is cost: some Chinese models can deliver around 70-80% of the performance of leading global AI models, but at less than 10% of the cost.

As a result, Chinese software companies are among the fastest growing in the world. Many leading open-source AI models originate in China and the country’s whole AI ecosystem is developing rapidly and with ever-more self-sufficiency. Increasingly we are seeing Chinese AI models using Chinese chips manufactured by Chinese foundries. 

In light of persistent investor caution towards China, investors need to distinguish between areas where skepticism is justified and those where it has gone too far – and what characteristics to look for to ensure downside risks are properly priced.

In our opinion, there are three types of Chinese AI companies. 

One: Global AI supply chain

The first group is companies that are on the global AI supply chain and have enjoyed rapid revenue growth and margin expansion. Earnings momentum is solid, but the price-to-earnings ratio (P/E) is on par with – or even higher than – those of equivalent global peers. 

Two: China-centric AI 

The second group is China-centric AI companies, which have limited earnings support. These are groups that have not yet found a way to generate sufficient returns from the extensive investments in AI. However, we believe that over time some of them have the potential to become very successful.  

The global market for Chinese AI models is growing, and the gap between China and the US is shrinking. 

Such investment opportunities hold out the prospect of high returns but remain high risk. An approach investors may choose to take is to buy a basket of companies; the return could be justified if just one becomes a winner in the AI race. 

Three: AI optionality

The final group offers AI optionality that the market has not yet priced in. These are typically well-known Chinese tech heavyweights that are still being valued on their legacy businesses – and in some cases de-rated because AI spending is dragging on near-term profits.

Our view 

It is this last group that we believe offers the most attractive risk-reward opportunities, while a cautious position could work for the second group. We are, however, less enthusiastic about the first group because they have already re-rated significantly and earnings upside from here is limited.  

We expect global AI capital expenditure growth to slow from 2027, while supply-chain margins are already elevated. For now, investors are fixated on the risks of microchip shortages. But shortages tend to be solved. When the cycle turns, the greater risk may be excess capacity.