By Hamza Faisal
Tencent's cloud business has passed 2 million deployed servers, senior executive Dowson Tong said on Thursday at the company's Global Digital Ecosystem Summit in Shenzhen, as overseas demand for its artificial intelligence services grew by triple digits over the past year. Tong, who leads Tencent's Cloud and Smart Industries Group, said active customers in international markets doubled over the same period.
The scale-up comes as Tencent pushes its AI products beyond China. The company has added or expanded data centre capacity in Germany, Malaysia, Mexico, Japan and Saudi Arabia, and Tong said Tencent Cloud now serves clients from more than 150 countries and regions. In Europe, where it courts firms with open-weight models and a private cloud running on customer premises, business has grown at high double-digit rates for several years, said Tencent Cloud Europe general manager Fred Sun.
Tong also used the summit to preview the next step in Tencent's model line-up: new Hunyuan models called Hy4 and Hy4 flash. Compared with the Hy4 preview released in August, a 770-billion-parameter open-weight model, Hy4 strengthens long-horizon and real-world productivity tasks, improves chain-of-thought efficiency and adds visual understanding for the first time, while Hy4 flash is positioned as a lower-cost, practical option.
An enterprise pivot after a 'leaky ship' warning
The overseas push reflects a deliberate change of direction inside Tencent. Founder Pony Ma described the company's AI effort as a "leaky ship" this summer as it trailed DeepSeek and other rivals in the post-ChatGPT race, and Tong's answer was to retool around business customers rather than consumer chatbots. The centrepiece is WorkBuddy, a workplace AI agent that combines Tencent's messaging and office tools with agents that draft reports and handle office tasks.
The bet appears to be paying off at home at least: WorkBuddy drew 85.34 million monthly visits last month, ahead of competing products from Alibaba at 46.55 million and ByteDance at 30.39 million, according to Chinese financial reports cited by the South China Morning Post. Tencent is now pitching the same agent to companies in Southeast Asia and Europe.
Demand is outrunning supply
Behind the expansion is a supply squeeze Tong did not hide. Agents for coding and office work have pushed token consumption up so fast that demand for computing power is outstripping supply, he told reporters, leaving Tencent itself "somewhat constrained" even as it builds. Infrastructure costs have also "increased quite a bit", he added — a rise he said he does "not believe that is very sustainable".
Part of Tencent's answer is domestic silicon: working with the Hunyuan team, it increased the throughput of a single domestically produced chip fivefold in the two weeks before launch, Tong said. It is also cutting costs by letting customers choose among models for each task — Hunyuan, DeepSeek or Kimi — rather than defaulting to the most expensive option.
What comes next
Tencent says it will keep building or expanding data centres across Asia, the Middle East, Europe and the Americas over the next year, betting that overseas companies' appetite for Chinese AI offerings stays strong. The risk is the same one facing every cloud provider: heavy capacity spending today for demand that may cool.
For now, though, the numbers point one way. A year ago many European firms were still only experimenting with AI, Sun said; now, he added, "everyone has moved from experimentation into more adoption". Tencent's task is to turn that adoption into durable cloud revenue before the infrastructure bill catches up with it.



