Tencent Holdings Ltd (TCEHY) (Q2 2026) Earnings Call Highlights: AI Investments Drive Revenue …

This article first appeared on GuruFocus.

• Total Revenue: RMB204.8 billion, up 11% year-on-year.

• Gross Profit: RMB118.4 billion, up 13% year-on-year, with overall gross margin at 58%.

• Non-IFRS Operating Profit: RMB75.6 billion, up 9% year-on-year; excluding new AI products, RMB86.1 billion, up 19% year-on-year.

• Non-IFRS Net Profit: RMB68.4 billion attributable to equity holders, up 9% year-on-year.

• Diluted EPS: RMB7.433, up 9% year-on-year.

• Value-Added Service (VAS) Revenue: RMB98 billion, up 8% year-on-year, with gross margin up 4 percentage points to 64%.

• Domestic Games Revenue: Up 17% year-on-year, contributing 23% to total revenue growth.

• International Games Revenue: Down 1% year-on-year (up 4% in constant currency), contributing 9% to total revenue growth.

• Marketing Services Revenue: RMB44 billion, up 22% year-on-year, contributing 21% to total revenue growth; gross margin at 57%.

• Fintech and Business Services Revenue: RMB60 billion, up 9% year-on-year, contributing 30% to total revenue growth; gross margin at 52%.

• Social Networks Revenue: RMB32 billion, up 1% year-on-year.

• Operating Profit: RMB67.3 billion, up 12% year-on-year.

• Operating Expenses: Selling and marketing expenses up 26% to RMB11.9 billion; R&D expenses up 35% to RMB27.2 billion.

• Operating CapEx: RMB51.8 billion, up 190% year-on-year.

• Free Cash Flow: Negative RMB13.8 billion; excluding AI-related prepayments, RMB37.6 billion.

• Net Cash Position: RMB58.2 billion, down from RMB146.9 billion as of March 31, 2026.

• Employees: Approximately 116,000, up 4% year-on-year.

Release Date: August 12, 2026

For the complete transcript of the earnings call, please refer to the full earnings call transcript.

Positive Points

• Total revenue grew 11% year-on-year to RMB205 billion, with gross profit up 13% and non-IFRS operating profit up 9%, demonstrating solid financial performance.

• Hunyuan 3 production version achieved a 6x increase in average daily token usage and ranks among the top 3 models globally on OpenRouter, validating its practical utility and cost-performance.

• WorkBuddy and CodeBuddy are achieving breakout user growth and are clear leaders in China’s AI productivity market, with high retention and willingness to pay among users.

• Domestic games revenue grew 17% year-on-year, driven by successful titles like Delta Force, Valorant PC, and Roco Kingdom: World, which ranked first among new games released in China this year.

• Marketing services revenue grew 22% year-on-year, driven by AI-driven ad targeting and increased impressions, with ad loads on Weixin Video Accounts still well below industry average, indicating further growth potential.

• AI infrastructure investments provide downside protection, as compute can be rented out at cost recovery or better via Tencent Cloud, with management noting a 30% profit on recent compute orders.

• Xiaowei prototype within Weixin leverages the ecosystem’s social graph and mini-programs, laying groundwork for agent-to-agent transactions and long-term monetization potential.

Negative Points

• Operating profit excluding new AI products grew 19% year-on-year, but including them, growth was only 9%, indicating significant drag from AI investments.

• Operating CapEx surged 190% year-on-year to RMB51.8 billion, leading to negative free cash flow of RMB13.8 billion in the quarter, despite strong core business cash generation.

• Net cash position dropped sharply from RMB146.9 billion to RMB58.2 billion due to capital expenditure payments and dividend payouts, reducing financial flexibility.

• International games revenue declined 1% year-on-year (up 4% in constant currency), with decreases from two Supercell games partially offsetting growth from other titles.

• Long-form video subscription revenue decreased 6% year-on-year, indicating challenges in that segment despite overall social network revenue growth.

• R&D expenses rose 35% year-on-year, and selling and marketing expenses increased 26%, reflecting higher costs to support AI initiatives and game marketing, pressuring margins.

• Management acknowledged potential headwinds in the China advertising market due to economic consumption softness, which could impact future marketing services growth.

Q & A Highlights

Q: How should we think about the payback cycle for the significant step-up in CapEx, and will it be paid off from incremental AI revenues or eat into earnings?A: James Mitchell (Chief Strategy Officer) explained that given the surge in demand and rental pricing for compute, Tencent could recover depreciation almost immediately by renting compute to third parties. However, the company is executing a larger strategy by allocating most new compute to building state-of-the-art models and market-leading AI applications, believing superior intelligence will convert into superior long-term economic returns. Martin Lau (President) elaborated that CapEx is divided into two parts: one for existing cash-generative businesses and another lump-sum investment for the new AI-native business. He noted that compute orders made just months ago can already be sold at over 30% profit, providing clear downside protection.

Q: Could management share preliminary feedback on Xiaowei’s testing phase, and how should we evaluate its net monetization potential given concerns it may shift existing Mini Program volume to higher-cost agents?A: Martin Lau (President) dismissed the risks, stating that AI-enabling the Weixin ecosystem will make it more useful, similar to how Weixin magnified QQ’s value by over 10x in the mobile age. He envisions Xiaowei as making Weixin “AI-first,” where users can execute transactions with simple instructions, creating incredible user experiences and empowering the entire ecosystem. He emphasized that WeLM is designed for privacy, cost efficiency, and executing within the Weixin environment, and that ecosystem expansion would translate into value based on current monetization mechanisms.

Q: How should we think about the margin profile of Tencent AI Cloud given rapid token price commoditization and a price-sensitive China cloud market?A: James Mitchell (Chief Strategy Officer) stated that while domestic token prices are low, manufacturing costs are also extremely low, allowing for positive gross margins. Gross margins for paying WorkBuddy users and Model-as-a-Service are already comparable to Tencent Cloud overall. He noted the China cloud pricing environment has improved, with Tencent increasing prices across the board in May and reducing discounts, partly due to rising input costs like memory.

Q: Where will Hunyuan 4 differentiate itself in a crowded market, and what is the timeline for focusing more CapEx on cloud as a high-ROIC business?A: Martin Lau (President) said Hunyuan 3 is a small model that matches or beats much larger models and focuses on real-world use cases rather than benchmarks. Hunyuan 4 will be bigger, beat models of larger size, and be more useful, with co-designed products becoming more powerful. James Mitchell (Chief Strategy Officer) added that the immediate primary use of CapEx is training bigger models, with a secondary use for inference behind WorkBuddy. He noted WorkBuddy’s subscription model creates a time lag between cash receipts and reported revenue, but a substantial ramp is underway, with GPU ASIC capacity for cloud rental stepping up toward year-end.

Q: Will the agent-to-agent transaction loop lead to a fully autonomous agent ecosystem within Weixin, and what are the benefits and challenges of on-device inference for Xiaowei?A: Martin Lau (President) envisioned a future where users send complex instructions to agents that execute transactions, with merchants also having agents that interact with user agents. On on-device inference, he explained it will happen step-by-step, with more inference moving to devices as compute becomes more powerful and efficient, returning to the normal state of the computer industry where most compute happens on-device. He noted this shift would make models and software more important, with compute CapEx borne across the ecosystem.

Q: Should we expect the accelerated marketing services growth to continue, and what future benefits might come from deeper Hunyuan 3 integration?A: James Mitchell (Chief Strategy Officer) cautioned against straight-line extrapolation, noting in-app advertising games contributed about 2 percentage points to growth this quarter and the China consumer market remains choppy. However, he expressed confidence in outperforming the overall China advertising market by a substantial margin, citing upside from AI ad targeting, increasing engagement on Weixin Video Accounts, and being early in the evolution toward closed-loop advertising that drives higher ad pricing.

Q: How should investors think about Tencent’s capital allocation priorities over the next 12-24 months given increased buybacks and accelerated CapEx?A: James Mitchell (Chief Strategy Officer) said capital allocation will be dynamic, steering more cash toward CapEx if superior returns are identified, potentially reducing buybacks. Martin Lau (President) stressed that the AI CapEx is a lump-sum investment for this year and next, not recurring annually. He noted it should be measured against operating cash flow, balance sheet cash, the investment portfolio, and prudent debt capacity, not just annual operating cash flow.

Q: What specific business value would a truly frontier-level model create that Hunyuan 3 cannot deliver, justifying a major increase in training spend?A: Martin Lau (President) clarified that Weixin’s agent doesn’t depend on Hunyuan’s SOTA status, as it’s designed for privacy and cost efficiency. However, SOTA status would enable a significant token business and empower WorkBuddy to complete more challenging, value-added services, potentially helping users make more money. He added that reaching SOTA would allow creation of multiple models at different cost-efficiency levels, all on the frontier curve, generating margins through control of the model, inference cost, and compute.

Q: How do you manage the AI product investment drag, and what signals would cause you to step up investment or shift from investment to harvesting mode?A: Martin Lau (President) said investment is dynamic and prudent until a breakout opportunity is identified, then investment may step up. He emphasized the business is for the long run, with economics improving over time, and noted the fallback option of renting out compute would be profitable. James Mitchell (Chief Strategy Officer) added that spending is dynamically reprioritized within the envelope, citing a big change between Q1 and Q2 where WorkBuddy was aggressively prioritized while other products were deprioritized.

Q: Where does Xiaowei fit in terms of inferencing capacity, and when should operating profit including AI investment grow faster than excluding it?A: Martin Lau (President) said Xiaowei’s cost envelope would be less than what was invested in Yuanbao over the past year, making costs manageable while returns flow in as the experience improves. He declined to provide specific guidance but reiterated there is an investment envelope adhered to with discipline

For the complete transcript of the earnings call, please refer to the full earnings call transcript.