Tencent Holdings posted second-quarter earnings that beat forecasts, lifted by gaming and artificial intelligence. Net profit rose 16% from a year earlier to 56 billion yuan ($7.81 billion), while revenue grew 15% to 184.5 billion yuan ($25.8 billion), marking Tencent’s 11th consecutive quarter of growth.
Tencent at a glance
Founded in 1998 and headquartered in Shenzhen, Tencent has grown into one of China’s most powerful technology companies and the world’s largest video game publisher. At the core of its business is WeChat, known domestically as Weixin, a super-app with more than 1.3 billion monthly active users that combines messaging, payments, social media, and advertising. Beyond its all-in-one platform, Tencent’s gaming division remains a major profit engine.
The tech giant has also emerged as a key player in artificial intelligence, holding more generative AI patents than any other Chinese firm. Its in-house foundation model, Hunyuan, powers applications across chatbots, gaming and enterprise services, while new open-source tools are pushing into 3D content generation.
Tencent expands with AI and mobility
Tencent’s revenue mix reflected broad gains, with marketing services up 20% to 35.8 billion yuan ($5 billion) on stronger ad conversions and WeChat video traffic, while fintech and business services rose by 10% to 55.5 billion yuan ($7.76 billion), supported by GPU rentals and API usage. Gaming continued to be a key driver, advancing by 17% in China and by 35% overseas on the strength of Honor of Kings, Peacekeeper Elite, PUBG Mobile, and new releases like Delta Force and Dune: Awakening.
Capital expenditures surged by 119% to 19.1 billion yuan as Tencent doubled down on artificial intelligence, unveiling an open-source 3D model alongside its Hunyuan foundation model. The company also introduced the Hunyuan Turbo S, which it says surpasses DeepSeek, intensifying a domestic AI race. Since its emergence in January, DeepSeek’s swift rise has reshaped China’s tech landscape, pushing industry leaders to fast-track AI initiatives to remain competitive.
Alongside earnings and AI investments, Tencent broadened its ecosystem through a partnership with SoundHound AI to integrate conversational voice technology into in-car platforms. The collaboration aims to deliver hands-free interaction with apps, entertainment, and vehicle controls, signaling Tencent’s ambition to extend its AI capabilities from digital platforms into intelligent mobility solutions (See related article here: The patents behind SoundHound AI’s in-car voice platform).
This article takes a closer look at Tencent’s patenting activity to show how the company is positioning itself in the global competition for generative AI.
Tencent: Patenting Activity
Tencent reached a record 11,338 global patent filings in 2020. Its strongest year coincided with the company’s most aggressive investment push, participating in more than 170 funding rounds worth 249.5 million yuan ($38 million), and accelerating its B2B pivot into cloud computing, fintech, and enterprise software. By the end of 2020, Tencent had built a portfolio of over 800 investee companies, with 70 listed on public markets and more than 160 valued at over $100 million, according to company reports.
In 2021, Tencent led domestic patenting China and ranked among the top 50 European applicants, focusing on blockchain, digital communications, medical technology, and computer technology. In 2022, the company deepened this innovation drive, launching three self-developed chips for AI reasoning, video processing, and high-performance networks. With nearly three-quarters of its workforce in R&D and 7,000 new projects launched, Tencent signaled its long-term push into semiconductors, cloud infrastructure, and generative AI.
Between 2015 and 2025, AI-related patents made up 31.5% of Tencent’s global filings. Activity accelerated after 2018, peaking at over 7,000 applications in 2022.
Beyond China, Tencent has also extended its IP strategy to the U.S. through Tencent America, where patent filings surged alongside efforts to expand its presence in esports, cloud services, and digital platforms. In 2019, Tencent laid groundwork for the 2020 Global Esports Summit, launched Tencent Meeting (VooV Meeting) to enter the cloud conferencing market, and advanced cloud gaming through a partnership with Razer.
Tencent: Top Law Firms
Tencent relies on a concentrated group of firms to manage its patent portfolio, with Scihead IP leading. Beijing Sangao Yongxin follows closely, while China PAT and Shenzhen Shenjian each handled over 5,000 applications. Advance China IP and Essen round out the group with around 4,000 and 3,200 filings, respectively.
The list is dominated by China-based firms, emphasizing Tencent’s strategy of consolidating IP strength domestically while building a platform for global expansion.
Meanwhile, Tencent America relies on a concentrated group of firms to manage its patent portfolio, with Deqi IP taking the lead. Sughrue Mion follows, while ArentFox Schiff and Itoh Patent Attorney Corporation also play key roles. Shiga International Patent Office, China PAT, and Wuesthoff & Wuesthoff provide additional support, with Shenzhen Shenjian, You Me, and Kim & Chang IP rounding out the group.
The list reflects a blend of U.S., Japanese, and Chinese firms, emphasizing Tencent’s strategy of combining domestic and international expertise to strengthen its IP presence in the U.S. market.
Tencent: Top Technology Areas
Tencent’s patent filings from 2015 to 2025 are heavily concentrated in electric digital data processing (G06F), which alone accounts for nearly 29% of its portfolio. Pictorial communication (H04N) and transmission of digital information (H04L) follow with 14.4% and 11.3%, respectively, highlighting Tencent’s strong emphasis on core computing, multimedia, and data transmission technologies that underpin its digital platforms. These top categories reflect the company’s strategy of reinforcing its dominance in software infrastructure and content delivery.
Beyond these, Tencent’s portfolio spans a wide mix of emerging areas. Image data processing (G06T) and computing models (G06N) each represent around 9%, while gaming-related patents (A63F) take up 7.4%, highlighting Tencent’s leadership in AI, computer vision, and entertainment. Smaller shares in information and communication technology (G06Q), image recognition (G06V), climate change mitigation in ICT (Y02D), and speech technologies (G10L) signal diversification into applied AI, green tech, and natural language processing. This mix shows Tencent balancing its core strengths with forward-looking investments in next-generation digital technologies.
Tencent’s AI patents
The following sections highlight three Tencent patents that apply AI to enhance digital media. These cover adaptive noise reduction, realistic speech-driven avatar animation, and precise human pose transfer, advancing clarity, interaction, and visual realism in online experiences.
AI-based audio scene classification for adaptive noise reduction
As digital media expands, audio quality has become critical for applications like streaming, conferencing, and gaming. Traditional noise reduction methods often apply generic filters, which may suppress useful sound along with unwanted noise. This results in distorted voices, muffled music, or inconsistent listening experiences across different environments. A smarter way to adapt audio processing to specific noise conditions can help preserve clarity while reducing interference.
U.S. Patent No. 12,308,041 describes an artificial intelligence-based method for audio processing that adapts to real-world noise environments. The system obtains an audio clip, classifies the audio scene to identify noise type, and then determines a corresponding processing mode based on the level of interference. By applying scene-specific noise reduction instead of generic filtering, the method preserves useful audio details while improving clarity and overall quality.
The patent, titled “Artificial intelligence-based audio processing method, apparatus, electronic device, computer-readable storage medium, and computer program product,” was filed on October 20, 2022, and granted on May 20, 2025. The patent lists Wen Wu and Xianjun Xia as inventors. Legal representation was provided by Anova Law Group, with attorneys Xiaoqun Wu, Wenye Tan, Cynthia Wu et al. named on the application.
AI-based speech-to-animation for realistic avatar expressions
As virtual avatars and digital characters become more common in gaming, social platforms, and the metaverse, creating realistic animations that match spoken speech is increasingly important. Traditional lip-sync methods often rely on fixed models tied to a specific speaker’s voice, leading to unnatural expressions or mismatches when applied to diverse speakers. A more flexible approach allows avatars to respond accurately to any speech input, improving immersion and interactive experiences.
U.S. Patent No. 12,002,138 describes an artificial intelligence-based method for speech-driven animation. The system analyzes incoming speech by breaking it into frames, identifying phoneme distributions, and determining expression parameters such as mouth shapes or facial movements. These parameters then drive an animated character to mirror the speech in real time. Because the method uses linguistic probability distributions rather than speaker-specific traits, it works across different voices and accents, delivering smoother and more natural avatar interactions.
The patent, titled “Speech-driven animation method and apparatus based on artificial intelligence,” was filed on October 8, 2021, and granted on June 4, 2024. The patent lists Shiyin Kang, Deyi Tuo, Kuongchi Lei, Tianxiao Fu, Huirong Huang, and Dan Su as inventors. Legal representation was provided by Anova Law Group, with attorneys Xiaoqun Wu, Wenye Tan, Cynthia Wu et al. named on the application.
AI-based image generation for precise human pose transfer
Realistic image generation is central to applications in virtual reality, gaming, and digital content creation. Traditional pose transfer techniques often rely on pose heatmaps, which can produce distortions or mismatches when adapting a source subject into a new pose. A more precise approach leverages model-based pose representations, ensuring that generated images remain visually consistent and lifelike.
U.S. Patent No. 12,333,627 describes an artificial intelligence-based image generation method designed to improve human pose transfer. The system acquires a source image of a person and a target image with a reference pose, then calculates a pose transition matrix to align the two. It extracts appearance features from the source image, applies the pose transition matrix to restructure those features, and generates a synthetic image of the subject in the target pose. By using model-based pose representations instead of heatmaps, the method produces more accurate and natural-looking transformations.
The patent, titled “Artificial intelligence-based image generation method, device and apparatus, and storage medium,” was filed on April 11, 2022, and granted on June 17, 2025. The patent lists Xiaohang Ren, Yang Ding, Xiao Zhou, Youcheng Ben, Yuxuan Yan, Pei Cheng, and Gang Yu as inventors. Legal representation was provided by Morgan, Lewis & Bockius LLP, with attorneys Gary Williams, Michael Lyons, Soyeon Laub et al. named on the application.
