Tencent Capex Triples As It Invests In AI

Chinese social media giant Tencent said its capital expenditure on AI and other areas more than tripled from a year earlier to 12 percent of its total revenue as it joins other Chinese and US companies in building up data centre capacity for deployments of generative AI.

The company said in an earnings announcement that it spent $10.7 billion (£8.3bn) on capital expenditure for the year, up from $3.4bn in 2023.

Tencent said it spent 39bn yuan ($5.4bn, £4.2bn) on AI initiatives in the fourth quarter.

Data centre outlays

The company’s president Martin Lau said in a post-earnings call that capital spending would rise to the “low teens” as a percentage of revenue for this year, with AI a key strategic focus.

“We will continue to increase our AI investments, increasing investment in our proprietary Hunyuan model while expanding our contributions in multimodal and open-source capabilities,” Lau said.

The description of spending as being in the “low teens” suggests it will be largely flat on 2024 as a percentage of revenue.

Chief strategy officer James Mitchell said separately that the large increase in 2024 spending should adequately address the company’s needs for this year, and said Tencent’s fourth-quarter capital expenditure was larger than that of any other listed Chinese firm.

Chinese companies have been joining their Western counterparts in making large outlays on AI infrastructure, with Alibaba saying in February it planned to spend at least 380bn yuan on cloud and AI infrastructure over the next three years.

Meanwhile, TikTok parent ByteDance is expected to double its capital spending to 160bn yuan this year, with 90bn yuan spent on computing power and the rest spent on infrastructure including data centres, said a December report from Zheshang Securities.

AI spending

Baidu, Alibaba and Tencent spent a combined 100bn yuan on capital outlays in 2024, the report said.

Similar initiatives have been announced in the West, including the $500bn Stargate plan to build US AI infrastructure and similar investments in France announced at a Paris AI summit in February.

Companies’ large outlays on data centres were called into question in January after Chinese start-up DeepSeek developed a high-performance AI model that required less computing power, but Nvidia chief executive Jensen Huang told developers this week that new AI agents would see companies’ AI computing needs multiplied by a factor of 100 this year.