Real Commercial Evidence
Tencent reported second-quarter revenue of 204.8 billion yuan, up 11% year over year, while its marketing-services revenue climbed 22% to 43.6 billion yuan. Tencent specifically attributed the advertising improvement to AI upgrades boosting ad performance and pricing inside its Weixin ecosystem.
This is stronger than another vendor saying its AI ad system increases productivity. Tencent is reporting material advertising revenue growth in a huge operating ecosystem where AI is improving targeting, recommendation, relevance, pricing and inventory monetisation.
The Cost Side Matters Too
The counterpoint is equally important: quarterly capital expenditure surged to 52.8 billion yuan from 31.9 billion yuan in Q1, and net profit rose only 0.7%, below expectations. So AI is producing observable advertising upside, but that upside currently comes with enormous infrastructure cost.
Better ROAS for advertisers does not automatically mean better AI economics for the platform.
The Two-Auction Problem
The eventual winners need to optimise two simultaneous auctions:
- The ad auction — which allocates advertiser money to impressions
- The compute auction — which determines how much intelligence can economically be applied to that impression
This makes AI efficiency a competitive advertising capability.
AI-Adjusted Ad Economics
| Metric | What It Shows | Tencent Evidence |
|---|---|---|
| Ad revenue growth | AI improves monetisation | +22% YoY |
| Capex growth | Intelligence requires infrastructure | +66% QoQ |
| Profit growth | Net economics still unclear | +0.7% (below expectations) |
| Revenue per yuan of capex | Efficiency of AI investment | Declining short-term |
For two years, almost every discussion about AI-powered advertising focused on better targeting + better creative + better bidding = better performance. Tencent exposes the other side: better performance requires inference, infrastructure and enormous capital investment.
What This Means for Advertisers
Tencent's results provide unusually concrete evidence that AI can materially increase advertising monetisation, but they also show why the future of AI advertising will be determined by compute economics as much as targeting intelligence.
For advertisers, the practical question is: which platforms can afford to apply the most intelligence per impression while maintaining sustainable economics? The platforms that solve this compute-efficiency challenge will offer the best performance at scale.
