Tencent reported third-quarter results on Wednesday that captured the cost of China's AI build-out in a single set of numbers: revenue growth that missed expectations, profit that barely moved, and capital expenditure that ballooned to nearly 40 percent of sales.
The Chinese internet giant posted quarterly revenue of 207.34 billion yuan, up 8 percent from a year earlier but below the 210.5 billion yuan analysts had expected, according to figures compiled by Bloomberg. The miss was driven by the company's accelerating AI spending, which is compressing margins across the business even as it fuels rapid growth in cloud services. For more context on this story, see our ongoing latest AI developments.
Capex Reaches 39 Percent of Revenue
The standout figure was capital expenditure. Tencent spent 81.4 billion yuan on capex in the quarter — far above the 55.4 billion yuan analysts had anticipated, according to Bloomberg data summarized by AI Weekly. That equals 39.3 percent of revenue, up from just 6.7 percent a year earlier, a roughly sixfold increase in AI infrastructure spending that the South China Morning Post described as coming at the expense of margins.
The spending is going into the servers, chips and data centers that power Tencent's Hy and Yuanbao model families, as well as the cloud infrastructure the company rents to enterprise customers. The build-out has turned Tencent's free cash flow sharply negative: the company reported free cash flow of minus 69.1 billion yuan for the quarter, driven in part by advance payments for compute capacity.
Profitability tells the same story. Non-IFRS operating profit fell around 6 to 7 percent year on year as the AI investments weighed on results, with market trackers noting that the decline sent shares of Prosus — Tencent's largest listed shareholder — down 8 percent in European trading.
Cloud Is the Bright Spot
The investment is producing results in one segment: cloud. Tencent's cloud revenue grew 35 percent in the quarter on AI demand, as enterprises across China rent compute and deploy the company's models for their own applications.
That growth, however, is being outrun by the spending behind it. The gap between cloud's 35 percent expansion and capex running at nearly 40 percent of total revenue illustrates the duration mismatch at the heart of China's AI race: companies are betting tens of billions of yuan on infrastructure whose returns will arrive over years, while the costs land on quarterly income statements now.
Chip Curbs: Cloud Safe, Model Training Could Slow
Tencent's earnings call also addressed the resumption of US chip restrictions. According to Reuters, the company said new US chip curbs will not hurt its cloud revenue but could slow its model training efforts. Chinese media reports went slightly further, with the company warning that the restrictions could delay the development of its next AI model.
The comments reflect an awkward reality for China's largest AI spenders: even as Beijing works to expand domestic chip production, frontier-scale model training still depends heavily on Nvidia accelerators, and each round of US export policy creates planning uncertainty for multi-quarter training schedules.
A Pattern Across China's Tech Sector
Tencent's quarter mirrors a pattern now visible across Chinese big tech. Rival Alibaba has made similar commitments, and smaller players such as Moonshot AI have scaled up GPU purchases as the competition for AI capability intensifies. The spending has turned China's consumer internet companies into some of the world's largest buyers of AI compute almost overnight.
For investors, the trade-off is becoming harder to ignore. Revenue growth at Tencent's core gaming and advertising businesses remains steady, and AI is already lifting cloud — but the market showed on Wednesday that it is in no rush to underwrite unlimited infrastructure spending. The 8 percent drop in Prosus shares underlined how sensitive Tencent's valuation has become to the pace of the AI build-out.
Whether the strategy pays off depends on execution over the next several quarters: converting Tencent's ballooning compute fleet into paid AI services at a scale that justifies spending nearly 40 percent of revenue on capex. The company, for its part, signaled it intends to keep spending — while cautioning that the next generation of its models may take somewhat longer to arrive if chip supplies tighten further.
Sources
This article is based on Tencent's Q3 2026 results as reported by Bloomberg, Reuters, the South China Morning Post and AI Weekly's market summary, all published November 18, 2026.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →