BEIJING/HONG KONG: Global investors are increasingly betting on Chinese artificial intelligence (AI) companies, seeking the next DeepSeek and aiming to diversify their portfolios as concerns grow over a speculative bubble in the US AI sector.
Demand for China’s AI firms is being boosted by Beijing’s push for technological self-reliance.
China has fast-tracked high-profile listings of chipmakers, including Moore Threads, dubbed “China’s Nvidia,” and MetaX, both debuting this month.
Foreign investors see China narrowing the technology gap with the United States, encouraged by government support for AI chipmakers.
UK-based asset manager Ruffer, cited by Reuters, said it has “deliberately limited exposure” to US tech giants and is adding positions in Alibaba to gain exposure to China’s AI theme.
“While the US remains the leader in frontier AI, China is rapidly narrowing the gap,” said Gemma Cairns-Smith, investment specialist at Ruffer.
“The moat may not be as wide, or as deep, as many think … The competitive landscape is shifting,” Smith said as quoted by Reuters.
Chinese tech giants and AI startups
Ruffer is investing in Chinese technology companies such as Alibaba, which operates an AI chip unit, owns the large language model Qwen, and is expanding cloud infrastructure.
Global fund managers are increasingly eyeing Chinese AI firms as startups list in mainland China and Hong Kong, capitalising on the surge in investor interest following the success of DeepSeek, China’s answer to ChatGPT.
UBS Global Wealth Management rated China’s tech sector “most attractive,” citing strong policy backing, rapid AI monetisation, and opportunities for geographic diversification.
The Nasdaq currently trades at 31 times earnings, compared with 24 times for Hong Kong’s Hang Seng Tech Index, which includes Alibaba, Baidu, Tencent, and chipmaker SMIC.
US investment adviser Rayliant launched a Nasdaq-listed fund in September offering exposure to Chinese tech equivalents of Google, Meta, Tesla, Apple, and OpenAI.
Brendan Ahern, chief investment officer at KraneShares, said the rise of Chinese AI chipmakers like Cambricon demonstrates the scale and speed of innovation across China’s AI and semiconductor industries.
China-US tech race
“In this race narrative, urgency benefits the companies,” Ahern said. KraneShares’ KWEB ETF, which invests in offshore-listed Chinese stocks including Tencent, Alibaba, and Baidu, has grown nearly two-thirds this year to almost $9 billion.
Another KraneShares ETF focused on onshore Chinese tech stocks, including Cambricon and Montage Technology, has also expanded.
Jason Hsu, founder of Rayliant Global Advisors, said the US leads in AI innovation, while China excels in engineering, manufacturing, and power supply.
He added that US tech export curbs have pushed China to invest heavily in domestic innovation.
“For investors, the prudent strategy is to capture AI opportunities and manage uncertainty through diversification,” Hsu said.
Valuation concerns
Despite investor interest, some fund managers warn valuations remain inflated. Kamil Dimmich, partner at UK-based North of South Capital, said, as cited by Reuters, that Chinese chip listings “have no valuation support and are almost entirely driven by hype.”
His fund holds Alibaba and Baidu, which have invested far less in AI development than their US peers.
Carol Fong, CEO of CGS International Securities, advised selective investment in Chinese companies benefiting from government self-reliance initiatives.
“Investors should balance exposure in the current fragmented, geopolitics-driven chip cycle,” she said.
New AI chips
Beijing-based Moore Threads Technology, hailed as “China’s little Nvidia,” unveiled two new chips, Huashan and Lushan, at a developer conference.
Huashan targets AI training and inference, with performance claimed to exceed Nvidia’s Hopper series, according to founder and CEO James Zhang Jianzhong.
The Lushan chip focuses on high-performance graphics processing, offering, Zhang claimed, 15 times higher gaming performance for triple-A games.
Both chips are scheduled for mass production next year and are built on Moore Threads’ new Huagang GPU architecture, providing 10 per cent better performance than the previous generation.
Other announcements included a KUAE computing cluster designed to interconnect tens of thousands of chips for training large AI models.
