Tencent Holdings' second-quarter financial report shows that advertising and domestic games continue to support growth and profits, but AI investment has shifted from strategic layout to a stage of heavy capital expenditure. The consensus on Wall Street is that Tencent's core business has not lost momentum; the issue is that AI costs have already materialized, while commercial returns still require a wait.
After the Hong Kong stock market closed on August 12, Tencent released its second-quarter 2026 results. The company reported quarterly revenue of RMB 204.785 billion, up 11% year-on-year, and Non-IFRS operating profit of RMB 75.64 billion, up 9% year-on-year. Both revenue and profit slightly exceeded market expectations, but profit growth lagged behind revenue growth, reflecting that AI-related spending is eroding some operational leverage.
Following the earnings release, Tencent shares opened lower on August 13, falling over 3%. Market focus has shifted from whether single-quarter results beat expectations to the upward trend in capital expenditure, negative free cash flow, and when AI products will generate verifiable revenue.
According to Zhuifeng Trading Desk, Goldman Sachs, UBS, Jefferies, and JPMorgan all maintained positive ratings on Tencent, but generally lowered or reassessed short-term earnings forecasts and price targets. The divergence among the four institutions is not about Tencent's AI competitiveness, but rather whether the commercialization speed of the WeChat ecosystem, Hunyuan model, productivity tools, and cloud services can cover the continuously rising costs of training, inference, and infrastructure.
Core Business Remains Stable, Advertising and Domestic Games Support Performance
Tencent's core business performance in the second quarter was better than market concerns.
Marketing services revenue reached RMB 43.565 billion, up 21.8% year-on-year, accelerating from the 20% growth in the first quarter. Goldman Sachs, Jefferies, and JPMorgan all identified advertising as one of the most important growth engines for the quarter. AI-enhanced recommendations, the automated targeting tool AIM+, higher eCPM ad formats, and an increased ad load rate on Channels collectively drove improvements in advertising monetization efficiency.
Channels remains the primary incremental space for Tencent's advertising business. Goldman Sachs pointed out that Channels' usage time grew over 20% year-on-year, but the ad load rate remains below industry levels, indicating Tencent has room to increase monetization density without significantly impacting user experience.
Highlights in the gaming business came mainly from the domestic market. Domestic game revenue in the second quarter was RMB 47.314 billion, up 17.2% year-on-year, significantly faster than the 6% growth in the first quarter. Products such as Delta Force, VALORANT, and Roco Kingdom: World provided incremental contributions, while operations and content updates for existing games also supported revenue recovery.
International games were relatively weak. International game revenue for the quarter declined approximately 1% year-on-year in RMB terms and grew 4% on a constant currency basis. Weak performance from some Supercell games offset growth from Wuthering Waves and the VALORANT PC version. Consequently, Goldman Sachs lowered valuation multiples for international games, noting that overseas business is unlikely to continue being priced with higher growth expectations in the short term.
Fintech and business services revenue was RMB 60.286 billion, up 8.6% year-on-year, showing relatively stable performance. Cloud business growth has entered the low-twenty percentage range, but it is not yet significant enough to change the group's overall profit structure.
AI Investment Accelerates, Cash Flow and Profit Margins Face Initial Pressure
Compared to the slight revenue beat, the market is more focused on Tencent's rapid increase in capital expenditure.
Tencent's capital expenditure in the second quarter was approximately RMB 52.8 billion, higher than the RMB 31.9 billion in the first quarter, accounting for about 26% of quarterly revenue, up 176% year-on-year. JPMorgan noted that Tencent's free cash flow turned negative to RMB 13.8 billion, and net cash fell from RMB 146.9 billion at the end of the first quarter to RMB 58.2 billion. AI investment has shifted from a long-term strategic narrative to a real cost on the cash flow statement and balance sheet.
UBS estimates show Tencent's capital expenditure rose to about RMB 53 billion in the second quarter, with the full-year capex forecast raised from RMB 170 billion to RMB 250 billion. Goldman Sachs also significantly increased its capital expenditure forecasts for Tencent in the coming years, with estimates of RMB 210 billion, RMB 245 billion, and RMB 247 billion for 2026, 2027, and 2028, respectively.
While specific figures vary among institutions, the direction is consistent: Tencent's AI investment intensity is higher than previously expected, and it may continue to rise in the second half of the year.
In the second quarter, investment in new AI products was approximately RMB 10.5 billion, up from RMB 8.8 billion in the first quarter. UBS believes that excluding AI spending, Tencent's adjusted operating profit would have grown 19% year-on-year, compared to the reported Non-IFRS operating profit growth of just 9%. JPMorgan similarly noted that excluding new AI products, Tencent's Non-IFRS operating profit would have reached RMB 86.1 billion, up 19% year-on-year.
This indicates that Tencent's traditional business is still generating profit elasticity, but AI is absorbing this incremental gain. The core issue currently is not whether Tencent has the capacity to continue investing, but how long the cycle of pressure on profit margins will last.
From Model to WeChat Agent, Monetization Paths Still Need Verification
Tencent's current round of AI investment is not primarily aimed at renting out computing power externally, but rather prioritizing internal model training, AI-native applications, and the WeChat ecosystem.
Goldman Sachs believes Tencent is directing more resources towards products like the Hunyuan model, WorkBuddy, CodeBuddy, Yuanbao, and WeChat's "Xiaowei". This type of investment differs from directly renting out GPUs to external customers, which typically generates short-term revenue more easily. The value realization of internal models and applications relies more heavily on user scale, retention, payment rates, and ecosystem synergy.
The Hunyuan model is the first observation line. UBS pointed out that HY3 has been online since July, with daily token consumption growing six times compared to the preview version; a larger HY4 model may be launched more quickly. The market will focus on its performance in complex workflows, agent capabilities, inference costs, and practical application effects, rather than just model rankings.
Productivity tools are the second line. WorkBuddy and CodeBuddy have become important vehicles for Tencent's AI commercialization. UBS stated that Tencent's PC-side productivity AI tools reached 27.5 million interactions in June, higher than the second-ranked peer's 12.8 million. JPMorgan believes that WorkBuddy's user growth, retention, and willingness to pay for tokens provide positive initial signals, but are not yet sufficient to constitute group-level revenue proof.
WeChat's "Xiaowei" is the most promising long-term variable but also the hardest to quantify. Tencent is testing a WeChat Agent driven by a customized WeLM model, aiming to embed AI capabilities into WeChat's content discovery, service execution, and transaction closure. Goldman Sachs believes that if WeChat gradually evolves towards an AI-first ecosystem, agent-driven services and transactions could become a new commercial gateway.
However, this concept also implies ongoing inference costs. Goldman Sachs estimates that the incremental inference cost of "Xiaowei" in the fourth quarter of 2026 could be equivalent to 0% to 5% of Tencent's EBIT; by 2027, the impact range could expand to 0% to 14% of EBIT. The wide range is due to significant uncertainty regarding user penetration rates, token prices, and product usage frequency.
Cloud Business Is Benefiting, But Still Struggles to Cover AI Costs
Tencent Cloud is the part of AI investment closest to revenue realization, but its current scale is still insufficient to fully offset capital expenditure pressure.
In the second quarter, Tencent Cloud's revenue growth rate entered the low-twenty percentage range, driven by AI demand, international expansion, and increased usage of general cloud services. Goldman Sachs stated that as of August, the daily token usage on Tencent Cloud's MaaS platform TokenHub had reached 25 trillion, growing five times in two months. AI demand is gradually translating into revenue from GPU leasing, MaaS, and productivity tools.
Management also mentioned that the pricing environment for cloud services has improved since May, with price increases and reduced discounts helping optimize pricing.
However, investment banks generally believe that accelerated cloud revenue does not mean AI investment has achieved a complete return. AI cloud, model services, and productivity tools can provide some profit support, but in the short term, they are still difficult to cover the large-scale costs of server procurement, depreciation, training, and inference.
UBS's assessment is that Tencent's AI business is currently closer to an "income option" than a mature business that can significantly boost group profits. The market has seen the costs, but has not yet seen a sufficiently large, auditable AI revenue line.
Short-Term Earnings Forecasts Under Pressure, Revaluation Depends on Verification Milestones
Goldman Sachs has lowered its net profit forecasts for Tencent for 2026 to 2028 by 1% to 4%, with the 2026 EPS forecast reduced from RMB 30.07 to RMB 29.73, and the 2027 EPS forecast from RMB 32.55 to RMB 31.19. The bank expects Tencent's EPS growth in the third and fourth quarters of 2026 to be only 4% and 0%, respectively.
UBS forecasts that Tencent's revenue growth in the second half of 2026 will be around 7% to 8%, with adjusted operating profit potentially declining about 1% year-on-year; revenue growth in 2027 is expected to be 10%, but adjusted operating profit is expected to be roughly flat. In the short term, the squeeze on the income statement from AI investment will continue.
In terms of valuation, the four institutions still maintain positive ratings, but their price targets reflect caution regarding short-term earnings realization. Goldman Sachs maintains a "Buy" rating with a target price reduced from HKD 700 to HKD 670; UBS maintains a "Buy" rating with a target price adjusted from HKD 780 to HKD 770; Jefferies maintains a "Buy" rating with a target price of HKD 750; and JPMorgan maintains an "Overweight" rating with a target price of HKD 690.
JPMorgan's assessment aligns closely with the core market sentiment: Tencent has a more solid earnings floor and has clearly disclosed the scale and boundaries of its AI investment, making it suitable for a 12-month holding period. However, in the absence of clear catalyst dates and auditable AI revenue, the stock price may remain in a phase of "not overvalued, but difficult to revalue quickly."
Going forward, the market will focus on the progress of subsequent Hunyuan models, retention and payment data for WorkBuddy and CodeBuddy, user adoption of WeChat's "Xiaowei", and whether Tencent Cloud can continuously increase AI-related revenue. Tencent has proven its core business can support the investment; the next step is to prove whether the rising AI costs can ultimately correspond to a sufficiently clear revenue and profit curve.
Comments