Tencent Holdings (00700.HK) released its financial report for the third quarter of 2024 on November 13. Tencent achieved revenue of 167.193 billion yuan in the quarter, and gross profit and operating profit (Non-IFRS) were 88.828 billion yuan and 61.274 billion yuan respectively, a year-on-year increase 16% and 19%, net profit (Non-IFRS) was 59.813 billion yuan, a year-on-year increase of 33%, still continuing the growth pattern of net profit being better than operating profit, gross profit and revenue growth.
In the past two years, driven by "emerging" businesses such as large models and SaaS, the proportion of high-quality revenue in Tencent's main business segments has gradually increased, and the quality of earnings has become higher.
In the latest third quarter report, Ma Huateng also named the contribution of "AI" again. He said, "Continue to deploy AI in products and operations, including marketing services and cloud services, and the tangible benefits brought by it are becoming increasingly apparent. We will continue to invest. Focus on AI technology, tools and solutions to serve users and partners.”
Data show that Tencent’s R&D investment reached 17.9 billion yuan in the third quarter, a year-on-year increase of nearly 9%. Since 2018, R&D investment has exceeded 320.4 billion yuan.
Large models continue to be upgraded, and enterprise service business grows healthily
This year, Tencent’s large model technology shows signs of accelerated development and implementation.
With the blessing of AI technology, the third quarter financial report shows that Tencent's enterprise service business revenue increased year-on-year, mainly due to the increase in cloud service revenue and merchant technology service fees.
Tencent launched a new generation of large model "Tencent Hunyuan Turbo" in the third quarter. Compared with the previous generation of large models, the training and inference efficiency doubled, while the inference cost was halved. It performed well in multiple benchmark tests. Relying on self-developed technology, Tencent has built a full-link AI large-model product matrix from computing infrastructure to top-level multiple intelligence applications.
At the same time, large models are also continuing to create greater demand for high-performance computing. Based on this, Tencent Cloud launched the AI infra brand "Tencent Cloud Intelligent Computing" during the Tencent Global Digital Ecology Conference in 2024. In terms of stability and performance, the number of single-day failures of Tencent Cloud's cluster kilocards is one-third of the industry level. , the data reading and writing efficiency is 10 times that of the industry, and the communication time of the kilo-card cluster is half that of the industry. Judging from the application situation, currently 90% of leading large model companies have chosen Tencent Cloud to provide them with intelligent computing services.
"Creating value for scenarios is the meaning of large model development." Wu Yunsheng, vice president of Tencent Cloud, head of Tencent Cloud Intelligence, and head of Youtu Labs, said publicly.
AI is developing rapidly, and new needs such as AI development are constantly emerging. Enterprises and developers can use Tencent's Hunyuan large model related capabilities through APIs, proprietary models, fine-tuned models, etc.
For example, after the upgrade of Tencent Cloud AI Code Assistant, it can provide enterprise customers with enterprise knowledge base Q&A, enterprise research and effect version viewing capabilities, and cloud hosting services. In a certain medical project, the R&D team applied the technical dialogue, code generation and other capabilities of Tencent Cloud AI Code Assistant to significantly shorten the development time of solution design, requirements development, unit testing, code review, and problem repair. A single project R&D improvement More than 20% effective.
Internally, Tencent currently has more than 700 products connected to Tencent's hybrid model, and Tencent Conference, Tencent Documents, and Enterprise WeChat continue to be intelligently upgraded. Externally, Tencent's full-matrix large-model products have been implemented in knowledge management, intelligent customer service, R&D efficiency improvement, intelligent marketing, content generation, office collaboration, risk management and control, etc.
In helping corporate customers implement AI technology into industrial scenarios, AI large model technology has become a new driver of revenue growth for Tencent's enterprise services segment.
The online advertising sector was renamed "Marketing Services", with AI drive as an important trump card
In Tencent’s third quarter financial report, the “Marketing Services” segment made its debut, which was upgraded and renamed from the original online advertising segment.
Data show that Tencent’s marketing services revenue increased by 17% in the third quarter, reaching 30 billion yuan, mainly due to the strong demand from advertisers for video accounts, mini programs and WeChat search advertising inventory, as well as the smaller demand for brand advertising related to the Paris Olympics. magnitude contribution.
Although China's online advertising spending sentiment was weak in the third quarter, Tencent's marketing services revenue growth rate still exceeded the market level, driven by budding businesses such as video account advertising and AI-driven efficiency improvements.
Advertising is the best training ground for large model technology.
Judging from Tencent Advertising's actions in recent years, the "technology-driven" path is clear. As early as 2021, Tencent Advertising began to upgrade the "advertising large model." One of the trump cards of Tencent's advertising business is the large model. After several years of iteration, the AI-driven advertising business has gradually seen results in revenue performance since 2023.
With the underlying capabilities of Tencent's Hunyuan large model, Tencent's advertising system was upgraded to 3.0 in June this year, making brand advertising more precise and improving delivery efficiency. Previously, Tencent also exited the "Miaosi" system to help advertisers more conveniently create creative ideas and generate advertising materials. The traditional process of advertising production and placement is greatly simplified through AIGC's technical capabilities.
It is understood that the new advertising delivery (3.0) upgrades the underlying large model on the Taiji machine learning platform, introduces new advertising IDs, reduces duplication of infrastructure, improves the delivery stability and achievement rate of the advertising system, and promotes the establishment of an industry asset library. The overall production volume rate of the industry increased by 207%. This new system has achieved varying degrees of breakthroughs in key industries such as the gaming industry, education industry, online services industry, and consumer goods industry, helping customers improve marketing efficiency.
Tianfeng Securities pointed out, “We are firmly optimistic that Tencent’s advantages such as portals, scenarios, and data will provide solid support for the long-term development of AI business. This year, Tencent’s AI business B-side and C-segment products have continued to make progress. We believe that the foundation of Tencent’s large model The combination of capabilities and applications is in a state of acceleration.”