After more than three years of intense competition in large AI models, the first half of the global AI race has concluded. At this critical juncture, Tencent, which dominated the internet era, is facing external skepticism about falling behind due to its failure to launch a groundbreaking large model or AI application. Determined to avoid marginalization, Tencent has quietly accelerated its efforts since late last year, recruiting top talent such as former OpenAI researcher Yao Shunyu while undergoing a series of AI team reorganizations. Amid the worldwide "Lobster" frenzy, Tencent is attempting a major AI counteroffensive through a dense product rollout. Investors, observing Tencent during a crucial period of valuation repair, are waiting for management to provide a vision that can trigger a market reassessment.

On March 18, Tencent Holdings (00700.HK) released its financial results for the fourth quarter and full year of 2025. In Q4 2025, Tencent's revenue grew 13% year-over-year to RMB 194.37 billion, while Non-IFRS operating profit increased 17% to RMB 69.52 billion. For the full year, revenue reached RMB 751.77 billion, up 14% year-over-year, and Non-IFRS operating profit grew 18% to RMB 280.66 billion. These figures exceeded market expectations. However, investors believe that sustaining growth is not Tencent's primary challenge; the real test is whether it can maintain its industry standing in the AI era.

During the earnings call, Tencent President Martin Lau revealed that the company's 2025 capital expenditure of RMB 79.2 billion was indeed below the original target, primarily due to GPU supply constraints. "For 2026, we hope to significantly increase capital expenditure while also boosting investment in flexible leasing options. Through these two avenues, we aim to enhance our overall computing power acquisition," he stated. Lau noted that Tencent has a clear strategic plan for AI and emphasized that executing on important, correct initiatives over the next two to three quarters is crucial. He expressed confidence that quantifiable progress would naturally emerge from these efforts. It is understood that the Hunyuan 3.0 model is currently in internal testing and will gradually become available to the public in early April.

Following aggressive reforms in the preceding two years, Tencent resumed double-digit revenue growth in Q4 2024 and maintained this level throughout 2025. Management attributed the 2025 growth to AI technology improving ad targeting and player engagement in games, coupled with accelerated revenue growth and scaled profitability in the cloud business. As Tencent's pillar business, gaming was the most critical driver of stable revenue growth. Financial reports show that Tencent's gaming revenue reached RMB 241.6 billion in 2025, a 22% increase year-over-year, accounting for 32.1% of total revenue. Domestic game revenue grew 18% annually to RMB 164.2 billion, while international game revenue hit a record high of RMB 77.4 billion, up 33% year-over-year. In the second half of the AI race, focused on application deployment, Tencent's gaming segment has become a key battleground. The earnings report specifically highlighted that game AI is effectively accelerating content production, enhancing user experience, and improving marketing efficiency.

Despite its strong core businesses, Tencent remains cautious about expansion, having stated that AI is likely the only area currently worthy of advanced investment. Compared to the more than twofold increase to RMB 76.7 billion in capital expenditure for 2024, Tencent's 2025 capital expenditure of approximately RMB 79.2 billion shows relatively modest growth. This cautious approach contrasts sharply with the high-profile announcements of multi-hundred-billion infrastructure investments by Alibaba and ByteDance. Martin Lau explained that last year, Tencent allocated more funds to share buybacks due to an inability to acquire sufficient GPUs ("cards"). If GPU supply improves this year, the company might moderately reduce buybacks to increase capital expenditure. Lau also pointed out that Tencent invested RMB 18 billion in new AI products last year and plans to at least double that investment this year. "The solid and sustained growth of our core businesses enables us to increase investment in AI," he said.

Tencent was an absolute winner in the mobile internet era, firmly holding the title of China's "king of tech stocks" thanks to its strongholds in social networking and gaming. However, perceived lagging in the current AI competition has led to a continuous decline in Tencent's stock price since last October. As of March 18, the closing price was approximately HK$550.5 per share, with the total market capitalization falling to around HK$5 trillion.

AI is now a central focus for the market regarding Tencent. In 2025, Tencent adjusted its organizational structure and increased AI-driven restructuring across multiple business scenarios. Whether integrating AI product lines like Yuanbao into the Cloud and Smart Industries Group (CSIG) to accelerate commercialization or deeply embedding large model capabilities into WeChat Search, Tencent's strategic intent is clear: it does not aim for a standalone AI super-app but rather seeks to make AI an integral part of its ecosystem infrastructure. However, this defensive counterstrategy is revealing limitations in the current, rapidly evolving AI landscape. Particularly with the accelerated arrival of the Agent era, market concerns persist about Tencent's AI development and investment pace. Under pressure from both Alibaba and ByteDance, there is a genuine risk of falling behind.

The model side, which underpins core capabilities, remains a critical area of market focus. Martin Lau indicated that over the past few months, Tencent has intensively upgraded the organizational structure and workflows of the Hunyuan large model team. It has also rebuilt its entire pre-training and reinforcement learning infrastructure while further enhancing data quality. Through these adjustments, Tencent aims to develop a more intelligent model and accelerate its overall iteration speed.

Within Tencent's AI initiatives, applications have garnered more attention. During the Spring Festival this year, Yuanbao, positioned as Tencent's pioneering AI-native application, aggressively spent RMB 1 billion to lead the "red envelope" marketing battle. While its daily active users briefly surged past 50 million, they declined noticeably after the holiday. Martin Lau believes Yuanbao has already accumulated a substantial user base, and the Spring Festival promotion met the company's expectations. The next step is to drive user acquisition and improve retention through continuous iteration of features, including dialogue and search capabilities. "We are already seeing that as these capabilities improve, user activity and retention rates continue to rise," he noted. Lau also acknowledged that many users continue to choose Yuanbao despite previous experience shortcomings because of its tight integration with the Tencent ecosystem. Yuanbao will further strengthen this synergy going forward.

From the heavily promoted Yuanbao during the Spring Festival to the current comprehensive layout around "OpenClaw," these represent Tencent's visible moves. Starting in March 2026, Tencent launched a matrix product plan centered around "Lobster," dubbing it the "Lobster Task Force." This product matrix targets not only novice users but also hardcore developers and large enterprises, aiming to drastically lower the barrier to using "Lobster." During the earnings call, Pony Ma, Tencent's Chairman and CEO, pointed out that "Lobster" serves as a decentralized entry point, better leveraging Tencent's resource advantages and enabling coordinated efforts across different business lines. It also provides inspiration for the WeChat Agent currently under development.

Nevertheless, Tencent's biggest AI trump card remains WeChat. Last year, management repeatedly mentioned plans to build an AI Agent embedded within and operating based on WeChat's unique ecosystem. Recent media reports suggest Tencent is secretly developing an AI Agent for WeChat, with a gray-box test potentially starting mid-year and a full user rollout possible as early as the third quarter. Martin Lau confirmed that the next step involves building an AI Agent within WeChat. This Agent can leverage WeChat's close user connections to create a rich ecosystem encompassing mini-programs, content, transactions, social interactions, and payments, enabling it to perform practical tasks beyond just chatting with users. "However, given WeChat's vast user base and concerns regarding privacy and security, we need to further enhance model capabilities to address these issues," Lau stated. He added that development of the WeChat AI Agent is progressing but does not yet have a definitive timeline.

After years of preparation, as the AI competition shifts focus from "competing on model parameters" to "competing on implementation capabilities," Tencent may have found its ideal moment to strike and is acting swiftly. Analysts at China Merchants Securities believe the market continues to undervalue Tencent's historical development and competitive moat. They also suggest that the emergence of "Lobster" could help Tencent find its "WeChat" for the AI era. With a WeChat ecosystem boasting 1.4 billion monthly active users and substantial cash flow, Tencent is far from being out of the game. However, whether this full-scale AI offensive is perfectly timed, and whether the trump cards of "Lobster" and ultimately WeChat can secure its core position in the internet landscape, remains to be validated by the market.