Tencent Cloud

Full briefing

Tencent Cloud is the cloud computing division of Tencent Holdings (HKEX: 700), one of the world's largest technology companies by market capitalization. Tencent Cloud is the second-largest cloud provider in China, behind Alibaba Cloud, with approximately $10–11 billion in annual revenue as of fiscal year 2024. Its global infrastructure spans over 70 availability zones across more than 25 countries, with the majority of AI workload capacity concentrated in mainland China data centers in Guangzhou, Shanghai, and Beijing. Tencent's Hunyuan (混元) large language model was announced in September 2023 and subsequently opened to enterprise customers via Tencent Cloud API. The model was trained on a multitrillion-token Chinese and English corpus, with particular depth in Chinese cultural context, coding assistance, and multimodal understanding. Hunyuan-T1, released in early 2025, is Tencent's reasoning-focused model positioned to compete with DeepSeek-R1 in the Chinese enterprise market. The Hunyuan family extends to image generation (Hunyuan-DiT, a diffusion transformer), video generation, and speech models — giving Tencent a multimodal AI portfolio that mirrors the breadth of its consumer product suite. The export control dimension is central to Tencent Cloud's AI hardware trajectory. Prior to the BIS October 2023 rule expansion, Tencent had operated NVIDIA A100 and H800 clusters for AI training and inference. After the rule change designated H100, H800, A100, and A800 as controlled items for China export, Tencent transitioned its training workloads to NVIDIA H20 GPUs — which are export-compliant but deliver approximately 1/3 of the H100's training throughput for large transformer models due to reduced NVLink bandwidth and memory bandwidth. Tencent has also deployed Huawei Ascend 910B clusters as a supplemental training substrate, with Huawei's CANN software stack progressively improving compatibility with standard PyTorch workflows. The WeChat integration angle distinguishes Tencent Cloud's AI deployment from pure infrastructure plays. WeChat (微信), with over 1.3 billion monthly active users, is one of the world's largest messaging and payments super-apps, and Tencent has integrated AI assistance — powered by Hunyuan — directly into the WeChat ecosystem. WeChat Search, Weixin Video, and the WeChat mini-program ecosystem all consume Tencent AI inference at scale. This gives Tencent Cloud a massive captive inference workload with relatively price-inelastic demand, since these AI features are table stakes for maintaining WeChat's dominance in the Chinese market. Tencent Cloud competes with Alibaba Cloud and Huawei Cloud for Chinese enterprise AI cloud market share. Its gaming division — which includes titles like Honor of Kings and PUBG Mobile — also drives significant AI workloads for content moderation, player behavior analytics, and procedural content generation, giving Tencent's AI infrastructure a more diverse demand base than pure-enterprise cloud competitors. The company has also partnered with domestic Chinese AI chip startups including Biren Technology and Cambricon to stress-test their chips on real production workloads, positioning itself as a gateway for domestic AI chip validation.