Tencent Cloud

Tencent Cloud is the public cloud and AI infrastructure arm of Tencent Holdings (HKEX: 0700), the Shenzhen-based internet conglomerate behind WeChat, QQ, and Tencent Games. Launched in September 2013 and run inside Tencent's Cloud & Smart Industries Group (CSIG) under president Dowson Tong, it operates a global footprint of roughly 24 regions and 68 availability zones, with the bulk of capacity concentrated in mainland China and secondary hubs across Hong Kong, Singapore, Tokyo, Seoul, Frankfurt, Silicon Valley, and Northern Virginia. It sits third in the China cloud market behind Alibaba Cloud and Huawei Cloud, with roughly 15% domestic share, and roughly 2% of global cloud infrastructure spend.

Tencent's AI infrastructure narrative is dominated by two constraints: US export controls limiting access to top-tier Nvidia silicon, and internal demand from the Hunyuan model family (Hunyuan-Large MoE, Hunyuan-T1 reasoning, Hunyuan3D). Tencent has publicly acknowledged stockpiling Nvidia H20 accelerators, the China-compliant Hopper variant, and rerouting inference workloads onto them, while also qualifying domestic silicon. The company has not disclosed a specific GPU count. Parent Tencent guided 2025 capex to a low-teens percentage of revenue, a step up from prior years and explicitly earmarked for AI infrastructure serving both first-party (Hunyuan, ad ranking, WeChat features) and third-party cloud customers.

Compared with the Western neocloud pack (CoreWeave, Nebius, Crusoe), Tencent Cloud is not a merchant GPU renter chasing model labs; it is a diversified hyperscaler whose AI capacity is captive-first, with external GPU-as-a-service and Hunyuan model APIs sold through its TokenHub product. Its distinctive edge is distribution into the WeChat/QQ ecosystem, gaming and video workloads, and Southeast Asian enterprises where US clouds face regulatory friction.