Stay up to date with free updates
Just sign up artificial intelligence myFT Digest – delivered straight to your inbox.
Chinese tech giants are investing heavily in artificial intelligence (AI) infrastructure, doubling their capital expenditure this year despite U.S. sanctions aimed at limiting the country’s progress in key technologies.
Alibaba, Tencent and Baidu spent a combined 50 billion yuan ($7 billion) on capital expenditures in the first half of the year, up from 23 billion yuan in the same period last year. The companies said they were focusing on buying processors and infrastructure needed to train large-scale language models for AI, both their own and those of others.
TikTok’s parent company, ByteDance, is also ramping up its AI spending, backed by more than $50 billion in cash and the advantage of being a privately held company with less investor scrutiny, according to two people familiar with the matter.
“We will continue to invest in research and development and AI capex to ensure the growth of our AI-driven cloud business,” Alibaba Chief Executive Eddie Wu told investors this month. “That’s simply because we see a lot of unmet demand from a lot of customers.”
Alibaba is buying processors to train its “Tongyi” series of AI models and rent out its computing power to other companies. The Chinese tech giant’s capital expenditures totaled 23 billion yuan in the first half of the year, up 123% from the same period last year.
“When you make this kind of capital investment, you spin up a server and it’s basically up and running instantly,” Wu says. “You get a really high ROI.” [return on investment] Over the next few quarters.”
Revenue from the group’s cloud business accelerated in the second quarter, growing 6% from the same period a year ago, and Alibaba said sales of AI-related products more than doubled year-over-year.
The surge is partly driven by investments in Chinese AI startups to acquire customers: Of the $800 million invested in AI startup Moonshot in February, just under half was in the form of vouchers to buy the company’s cloud services.
While U.S. export controls block access to Nvidia’s flagship AI processors, such as the H100 and the upcoming Blackwell series, Chinese tech giants can buy less powerful processors, such as Nvidia’s H20, which are designed not to exceed computing power thresholds set by Washington.
Analysts expect Nvidia to ship more than 1 million processors to Chinese technology companies in the coming months at $12,000 to $13,000 apiece, with ByteDance being a major customer, according to two people familiar with the matter.
Dylan Patel of semiconductor research group Semianalysis estimates that TikTok’s parent company has bought hundreds of thousands of H20s for data centers in China and is also spending heavily to build out computing infrastructure with partners in the Malaysian state of Johor.
“ByteDance is China’s largest AI buyer because it’s investing heavily in China and Malaysia and buying from the U.S. cloud,” Patel said.
Social media and gaming giant Tencent announced that its capital expenditures in the first half of the year reached RMB 23 billion, up 176% year-on-year, partly due to “investment in GPUs and CPU servers.”
Chief Strategy Officer James Mitchell said the company’s cloud business is benefiting from rising demand for renting graphics processing units, but not on the same scale as the boom being experienced by its U.S. peers.
“China doesn’t have a ton of well-funded startups trying to bootstrap large-scale language models. There are a lot of smaller companies, but they have $1 billion, $2 billion in capital,” he said. “They don’t have $10 billion, $90 billion in capital like in the U.S.”
A person familiar with Tencent’s investment strategy said the company is writing smaller checks to the AI group as concerns remain about Beijing’s regulatory stance.
Baidu, China’s longtime AI leader, has been the most restrained on capital expenditures, spending RMB4.2 billion in the first half, up 4% from the same period last year.
Overall, capital expenditures by China’s big tech companies still lag far behind their U.S. peers, with Alphabet, Amazon, Meta and Microsoft spending $106 billion in the first half and pledging more in the coming months.
