China’s cheap AI shifts focus to internet stocks – from e-commerce to advertising

[SINGAPORE] China’s rapid progress in low-cost artificial intelligence is strengthening the investment case for the country’s internet giants after years in the shadow of chipmakers, market participants say.

Investors are increasingly betting that falling AI costs will shift more of the industry’s value toward companies that own the applications and services people use every day, from search and e-commerce to advertising and enterprise software. 

Since a low on Jun 25, the KraneShares China Internet exchange-traded fund has gained more than 20 per cent, about twice the Hang Seng Index’s advance, while the Philadelphia Semiconductor Index has fallen roughly 11 per cent.

The shift reflects a growing belief that the biggest winners from AI may be broadening from those making the hardware to the companies delivering AI services to consumers and businesses.

“Chinese hyperscalers will benefit substantially because they have cheaper currency to run on and cheaper models will increase the use of AI significantly by them and their clients,” said Hao Hong, chief investment officer at hedge fund Lotus Asset Management.

“That is why BABA and Tencent are roaring lately. So cheaper tokens are going to spur demand. And cheaper serving costs can be margin positive for the Chinese hyperscalers.”

Chinese developers have become increasingly competitive in open-weight AI models, narrowing the gap with global leaders while keeping prices low.

Alibaba Group’s latest Qwen3.8-Max model, for example, now ranks among the leading coding models on Arena’s benchmark that compares AI systems on performance, speed and cost.

Those advances are making AI cheaper to use. Silicon Data’s blended inference-price index has fallen 36 per cent from its May peak even as customers increasingly adopt more capable reasoning models and AI agents.

SEE ALSO

The AI token takeover in China

China is getting more anxious about Mythos ahead of Trump-Xi meeting

China doesn’t need to build better AI than the US to win the tech race

Is an AI price war about to begin?

More

Lower computing costs improve the economics of services such as search, advertising, e-commerce, customer support and enterprise software, while making new AI applications commercially viable.

The idea is not unique to China.

On Microsoft’s earnings call last week, chief executive officer Satya Nadella said AI models are becoming interchangeable inputs, suggesting that more value could accrue to companies that own customer relationships and software platforms rather than the models themselves.

Goldman Sachs believes China’s accelerating token growth, rising second-half capex, and increasing domestic chip supply reinforce its preference for cloud and data centres.

Ronald Keung, the broker’s head of Asia Internet research said in a Bloomberg TV interview on Wednesday (Aug 5) that Chinese hyperscalers are best positioned given their strong cash flows and ability to fund capex while scaling cloud businesses.

He estimates Alibaba Cloud grew 45 per cent in the June quarter. 

To be sure, there are reasons for caution.

China’s technology sector has a long history of prioritising scale over profitability, and AI may prove no different.

The internet giants that have their own AI models are also under growing pressure to defend their franchises against rivals including DeepSeek, Z AI, Moonshot and ByteDance.

Meanwhile, Alibaba, Meituan and JD.com continue to face a price war in the food delivery business and sluggish domestic consumer demand.

For now, investors believe the rotation may have further to run because China’s cloud and internet stocks are only beginning to be rerated as beneficiaries of falling AI costs after years of underperformance.

Firm graphics processing unit rental rates also point to sustained compute demand even as token prices decline. BLOOMBERG