Goldman Sachs reaffirms 'Buy' rating on Tencent, optimistic about the reconstruction of Hunyuan 3.0 and the explosion of intelligent entity ecosystems

Goldman Sachs released the latest research report, pointing out that at the recently held Asia Communacopia+ Technology conference, James Mitchell, Chief Strategy Officer of Tencent Holdings (00700), elaborated in detail on the group's AI strategy map.

Goldman Sachs said it is encouraged by Tencent's restructuring progress in the AI field, maintaining its "buy" rating, and giving a 12-month target price of 700 Hong Kong dollars.

Hunyuan 3.0 fully restructured, the basic model strategy ushers in an inflection point

Goldman Sachs pointed out that Tencent had previously reorganized the large language model (LLM) team and redesigned the pre-training and reinforcement learning stack, expanding and cleaning up the dataset, shifting the evaluation standards from public benchmarks to general intelligence.

After a thorough reshaping of the basic strategy, the newly restructured Hunyuan 3.0 (Hy3) model has become one of the most competitive models in its domestic category. This not only lays a solid foundation for Tencent to launch larger and smarter model iterations in the next 6 to 9 months but also signifies that Tencent is shaking off the label of a "basic model latecomer."

Intelligent agents become a new engine, WeChat ecosystem releases long-term memory

As the AI narrative shifts from chatbots to more commercially valuable agent-based AI (intelligent agents), Tencent's layout is about to enter the harvest period:

• Core application landing: Tencent, with CodeBuddy and WorkBuddy as the core, first completed a clear early verification in the field of code generation and is gradually infiltrating the general computing usage (LAM) and white-collar productivity areas.

• Front-end control surface barriers: WeChat, Enterprise WeChat, QQ, Tencent Yuanbao, and QQ browser constitute a natural user-end control surface. Among them, the WeChat AI agent has a longer memory than ordinary chatbots, capable of perfectly handling complex long texts and multi-party dialogue situations.

• Strong retention and high willingness to pay: Tencent emphasizes that active and paid users who use AI agent functions show strong retention, and users' willingness to upgrade to higher-priced tier tokens (Tokens) is extremely high.

Computing power and investment double-cycle deployment, compatible with domestic ASIC chips

In terms of commercial monetization and underlying computing power infrastructure, Tencent has shown a highly matched double-cycle deployment strategy.

In terms of short-cycle deployment, Tencent focuses on advertising technology and Tencent Cloud. This move can bring immediate measurable returns, with extremely high profit margins in advertising and considerable revenue contribution from Tencent Cloud.

In terms of long-cycle deployment, Tencent focuses on the continuous iteration of the Hunyuan large model. Relying on Tencent's unique differentiated dataset and WeChat super scenario, this move will precipitate long-term ecosystem value.

Goldman Sachs specifically mentioned that Tencent has spent more than a year preparing its technology stack, achieving perfect compatibility with a variety of domestic ASIC (Application-Specific Integrated Circuit) chips, capable of running AI inferences efficiently and on a large scale. As more of its own computing power comes online, Tencent can also flexibly direct incremental computing power to the outside to meet the strong demand from Tencent Cloud customers.

Valuation repair path: AI narrative in 2026 will determine elasticity

Goldman Sachs concluded that while maintaining the resilience of core profitability, Tencent is significantly increasing AI capital expenditure (CapEx) commitments. The token usage of the Hunyuan 3.0 preview version and the retention indicators of early productivity agents have both exceeded expectations.

Goldman Sachs believes that as one of China's three largest hyperscale cloud computing service providers, the subsequent valuation repair elasticity of Tencent will highly depend on the progress of the AI narrative in the remaining time of 2026 — especially the iteration of the new Hunyuan model, the efficiency of capital expenditure turning into revenue, and the official launch of the "WeChat AI agent" that may cause a sensation within the WeChat ecosystem.