Pony Ma and Tencent's 'AI Anxiety.'

As 2026 unfolds, how will Tencent, this colossal ship, navigate the intensifying competition in AI?

Tencent Holdings (0700.HK) reported total revenue of RMB 196.458 billion in the first quarter, up 9% year-on-year; adjusted net profit reached RMB 67.905 billion, up 11% year-on-year.

As of market close on May 15, Tencent's stock price stood at HK$456 per share, down over 26% from its January 2026 peak; its market cap evaporated by nearly RMB 1.5 trillion in the first quarter.

01. Stock Price Under Pressure in Q1

Judging solely by the first-quarter financials, Tencent's fundamentals appear solid: all three core businesses achieved year-on-year growth. Marketing services revenue reached RMB 38.2 billion, up 20% year-on-year, accelerating from the previous quarter's 17%. Gaming-related services revenue hit RMB 96.1 billion, up 4% year-on-year, though growth slowed slightly. Tencent attributed this to the later timing of the 2026 Chinese New Year holiday, which delayed recognition of domestic gaming revenue into subsequent quarters.

However, profit changes warrant closer attention. In Q1, Tencent's non-IFRS operating profit reached RMB 75.6 billion, up 9% year-on-year. But excluding revenue, costs, and expenses from new AI products like Hunyuan Large Model, Yuanbao, CodeBuddy, WorkBuddy, and QClaw, operating profit would have been RMB 84.4 billion, up 17% year-on-year. In other words, new AI products dragged down operating profit by approximately RMB 8.8 billion in the quarter.

In fact, Tencent's heavy AI investments have significantly impacted profits, with this aggressive spending continuing from last year to the present.

In Q1 2026, Tencent allocated nearly one-third of its profits to R&D: R&D costs reached RMB 22.54 billion, up 19% year-on-year, primarily driven by employee benefits (AI talent acquisition), which hit RMB 16.255 billion, up 9% year-on-year.

Tencent stated it would continue investing in AI development. 'Capital expenditures in the second half will increase compared to last year, with more domestic chips being deployed. We expect continuous improvements year after year.' Based on this, securities firms have sharply downgraded their full-year profit forecasts for Tencent.

While AI investments erode profits, Tencent's share buybacks in Q1 2026 plummeted year-on-year.

Bo Wenxi, Chief Economist of the China Enterprise Capital Alliance, believes that beyond business fundamentals, Tencent's stock price pressure is also tied to tight overall liquidity in Hong Kong stocks. The U.S. Federal Reserve maintaining high interest rates has led to capital repatriation, slowing inflows via Stock Connect. The Hang Seng Tech Index's overall valuation is under pressure, dragging down Tencent as a heavyweight stock.

In response to investor concerns, Tencent provided a more detailed AI growth plan and noted that its stock price has diverged to some extent from its actual fundamentals, making this a suitable window for share buybacks.

02. A Late Start Leads to Falling Behind

The stock price pressure stems from the market shifting from AI hype to focusing on tangible returns. However, Tencent's current AI output scale and level fall short of capital expectations.

In its Q1 financials, Tencent repeatedly highlighted breakthroughs in its AI business: WorkBuddy, launched in March, is now China's most widely used AI productivity agent; Hy3 preview ranked high in OpenRouter's token consumption rankings shortly after its late-April launch. Nevertheless, these achievements have failed to alleviate investor anxiety over Tencent's heavy AI investments and slow results.

This anxiety even spilled over into Tencent's recent shareholders' meeting. An investor directly asked founder Pony Ma whether Tencent's AI had fallen behind. Ma responded, 'A year ago, we thought we had boarded the ship, only to find it was leaking. Now we feel like we're standing on it but can't sit comfortably. We hope the ship can speed up.' This indirectly acknowledged the gap between Tencent's AI and industry leaders, serving as both a realistic self-assessment and a pressure signal to internal and external stakeholders.

This gap originates from hesitation at the starting line. Tencent has spent the past year making up for its early sluggishness.

During the 'Hundred-Model Battle' in 2023, Tencent was grappling with fading pandemic dividends, gaming license restrictions, and sluggish growth in its two cash cows—advertising and gaming. A Tencent employee revealed to Yicai Global that in 2023, the company focused on cost reduction and efficiency gains across the board. AI Lab discouraged researchers from publishing papers, and business units needed to apply for budgets and project approvals from top management to train models, with the overall strategy emphasizing WeChat Channels more.

Changes in leadership of Tencent's large model team also reflect shifts in its AI strategy. When Tencent launched the Hunyuan Large Model in September 2023, Vice President Jiang Jie concurrently oversaw the general-purpose large model team. Jiang, lacking a professional background in LLMs (Large Language Models), previously managed product technology for advertising platforms. Under his leadership, Tencent's AI excelled in advertising applications but lacked sensitivity to cutting-edge technological shifts.

During Jiang's tenure, the Hunyuan Large Model remained a 'follower' rather than a 'leader.' It lacked the technical sharpness of DeepSeek or the product explosiveness of Doubao.

Not until 2025 did Tencent transfer leadership of its large model team from Jiang Jie to Yao Shunyu, a recruit from OpenAI. By late 2025, Tencent overhauled its AI R&D team, establishing the AI Infra Department, AI Data Department, and Data Computing Platform Department to strengthen its large model R&D system and core capabilities. Yao Shunyu was appointed Chief AI Scientist under the 'CEO/President's Office,' reporting directly to Tencent President Martin Lau. This move elevated the large model business to a 'top-priority' level.

'Qujie Business' learned that one of Tencent's most critical strategic goals this year is to reclaim a spot in the top tier of large models. However, even after personnel changes and organizational restructuring, closing the technological gap won't happen overnight.

After Yao's arrival, he nearly rebuilt Hunyuan from the ground up. The Hy3 preview released in April 2026 represents the first fruit of this rebuild. This MoE architecture model, with 295B total parameters and 21B activated parameters, innovates in fast-slow thinking integration—supporting three reasoning modes (no think, low, high) that dynamically switch based on task complexity.

'Qujie Business' noted that Hy3 preview focuses on refining long-text parsing, code compilation, and agent-linked operations, significantly improving contextual learning and instruction-following capabilities in real-world work and life scenarios. However, Tencent openly acknowledges on its open-source page that Hy3 preview underperforms GLM5.1 or Kimi K2.5 in some scenarios due to parameter quantity gaps.

Evaluations only reflect 'exam-taking' abilities; today, both industry insiders and outsiders prioritize practical usage effects when judging large models. According to Tencent's internal progress, a larger flagship model is in training, with the next version expected to achieve significant breakthroughs. Whether Hunyuan can truly secure its place in the top tier will then be judged more clearly by users and developers.

03. Can It Defend Its Ecosystem Advantages?

Tencent's early strategic hesitations have slowed more than just model iteration.

First, Tencent Cloud has 'stalled,' nearly pushed to the margins of the cloud services arena. Large models and cloud services typically fuel each other in tech giants' portfolios, but Hunyuan's lack of competitiveness, coupled with macro factors like computing power shortages, has constrained Tencent Cloud's expansion.

According to IDC's 'China Public Cloud Services Market (H2 2025) Tracker,' Volcano Engine, Alibaba Cloud, and Baidu Intelligent Cloud together captured 90% of the public cloud IaaS market share, leaving Tencent Cloud with just 8%, ranking fifth.

Second, Tencent's application ecosystem faces scrutiny. Industry observers point out that no new apps with over 50 million daily active users have emerged since 2020 unrelated to Tencent. 'Yuanbao' briefly surged into the '100-million MAU club' during the Chinese New Year holiday thanks to RMB 1 billion in red envelopes but quickly fell back to 57.35 million, raising concerns about user retention.

Meanwhile, WeChat, a social fortress with 1.4 billion users, is showing cracks under the impact of AI Agents. During the earnings call, an investor directly asked how Tencent views the potential threat to WeChat from OS-level AI Agents (those from iOS and Android phone manufacturers).

Doubao Mobile can directly invoke some WeChat functions via voice commands, such as sending messages or checking Moments. This 'bypassing app entry points, directly accessing services' model is eroding WeChat's exclusivity as a 'super entry point.' Just as WeChat blocked Doubao Mobile's permissions, Tencent did not directly answer this question at the earnings call, instead stating, 'Agents provide services to users but require permissions from each app—otherwise, they would be plundering other apps, not a good way to manage an operating system.'

To prevent apps within its ecosystem from losing 'entry point' traffic, Tencent apps are deeply integrating AI to retain users. For example, Tencent Docs has integrated Hunyuan and DeepSeek to boost PPT generation efficiency.

The most closely watched development is WeChat's AI transformation. Recently, The Information reported that Tencent is developing a WeChat AI agent capable of hailing rides, ordering groceries, and booking flights—akin to a full-service 'life manager.' Internet analyst Ding Daoshi noted that this would likely be implemented via mini-programs, adding an intelligent hub to various office, life, and entertainment mini-programs, enabling users to complete tasks with minimal perception. 'This doesn't affect merchants' mini-programs; users just rarely manually open them anymore, delegating everything to the agent.'

In fact, the mini-program Agent ecosystem WeChat aims to build is also being pursued by major phone manufacturers, as well as Alibaba and ByteDance.

However, as Tencent mentioned, Agent ecosystems require authorization from each app. For WeChat to build its own mini-program Agent, it must also navigate authorization hurdles. Whether high-frequency tool apps like Meituan and Didi can deeply collaborate with Tencent remains to be seen.

From 'falling behind' to 'frantic catch-up,' Tencent is at a critical juncture in transitioning between old and new growth drivers. The AI report card Tencent must deliver isn't as simple as climbing one or two ranks in benchmarks. Catching up requires not just capability but also time. The question is: how much time does Tencent have left to mount its comeback?