Tencent has reorganized its Hunyuan AI model development structure, emphasizing computing power, algorithms, and data. The company plans to increase research and development investments in this area.

As part of the restructuring, Tencent has established two new departments: one dedicated to large language models and another focused on multimodal models.

The large language model team will work on advancements in language-based AI. Meanwhile, the multimodal team will integrate various types of data, including text and images, into AI systems.

Tencent has also enhanced its data infrastructure by creating a new department to manage data processes for large models. Additionally, a machine learning platform department has been formed to develop an integrated platform for AI model training and operations.

🔗 Source: Jiemian

🧠 Food for thought

1️⃣ Tencent’s restructuring reflects China’s strategic AI investment surge

Tencent’s reorganization of its Hunyuan AI team comes amid a significant expansion of its AI investments, with planned capital expenditure set to reach 90 billion yuan in 2025, up from 77 billion yuan in 2024 1.

This investment approach is part of a broader trend among Chinese tech giants racing to develop domestic AI capabilities, with Tencent’s spending on AI infrastructure projected to quadruple year-on-year to 36.6 billion yuan by late 2024 1.

While impressive, this investment strategy operates within constraints identified by analysts, who note that Chinese companies face a 6-24 month technology gap behind US competitors in AI model development 2.

The restructuring into specialized departments for language models and multimodal AI aligns with Tencent’s strategy to maximize returns on its substantial investments while navigating these competitive pressures.

2️⃣ Efficiency-focused approach distinguishes Tencent from competitors

Unlike competitors who prioritize massive GPU acquisition, Tencent claims to achieve superior AI productivity through optimizing chip efficiency rather than simply expanding hardware numbers 3.

This approach is reflected in Tencent’s projected AI training expenditure of approximately $13 billion, which remains lower than competitors like Microsoft and Amazon despite ambitious AI goals 3.

The newly formed machine learning platform department likely aims to enhance this efficiency advantage by building integrated training and inference infrastructure that maximizes return on computational investments.

Tencent’s reorganization into specialized teams focusing on specific model types suggests a targeted resource allocation strategy rather than broad-based expansion across all AI domains simultaneously.

Recent Tencent developments