Tencent Just Showed What AI Really Costs. America’s Cloud Giants Are Next.

The numbers Tencent reported this morning are not complicated. Revenue rose 8% year over year to RMB207.336 billion in the quarter ended September 30, while operating profit fell 7% to RMB59.147 billion. Revenue fell short of Bloomberg analyst estimates of RMB210.5 billion. Cloud, the business most directly tied to AI demand, was the standout: FinTech and Business Services revenue was RMB63.655 billion, up 9% year on year, with cloud services revenue growing 35%.

What makes this quarter genuinely instructive for professional investors is the gap between those two numbers, the 35% and the minus 7%, and what sits between them.

The Bill Arrives

Capital expenditure payments jumped 225% to RMB65.1 billion during the quarter, and the company recorded negative free cash flow of RMB69.1 billion, partly reflecting advance payments for computing capacity to support its AI models and cloud services. To put that trajectory in context: capex as a share of revenue stood at 6.7% in Q3 2025, then climbed to 16.3% in Q1 2026, 25.8% in Q2 2026, and 39.3% in Q3 2026. That is not a rounding error. Tencent is funding the AI build from the balance sheet: capex rose from 6.7% to 39.3% of revenue in a year and net cash of RMB102.4 billion a year ago became net debt of RMB13.7 billion.

The operating margin fell to 29% from 33%. Tencent’s new AI products, Hy, Yuanbao, CodeBuddy, WorkBuddy and Xiaowei, weighed on margins: excluding them, non-IFRS operating profit rose 8% to RMB83.3 billion and the margin held at 40%. The core business is not broken. The AI products are simply expensive.

What the Bull Case Requires

WorkBuddy has become the most-used AI product in China, measured by large language model token consumption, with more than 10 million users per day consuming a combined total of over 100 trillion tokens, and enterprise customers are growing rapidly. Marketing services revenue rose 19% to RMB43.049 billion, and Weixin and WeChat combined monthly active users reached 1.449 billion. Those figures suggest the demand side is real.

Bulls argue that the capex is front-loaded: much of the Q3 outflow reflects prepayments for compute procurement rather than recurring operating cost. Excluding those prepayments, free cash flow would have been positive RMB16.9 billion. The bear case is that prepayments become the new normal as Tencent competes for GPU capacity in a supply-constrained market.

What Investors Are Missing

The question most commentary has not addressed directly is what Tencent’s P&L implies for Microsoft, Amazon, and Alphabet. Those three companies have each guided to rising AI infrastructure spend through 2026 and 2027, and each has so far managed to absorb the cost without a visible margin hit at the operating line. Tencent’s quarter illustrates the mechanism by which that changes: when capex as a share of revenue crosses 25% to 30% and shifts from balance-sheet-funded to income-statement-visible, operating leverage reverses. What shareholders should expect to see is that the incremental revenue generated by AI offsets ongoing investment and progressively translates into higher earnings per share and sustainable cash returns. At 39.3% of revenue, Tencent has not yet shown that crossover. Neither has anyone else.

Stocks to Watch

Tencent (0700.HK / TCEHY) is the clearest read-through. Cloud growth of 35% is a decent answer, but it comes from a base that is still small next to the capex. The stock’s valuation hinges entirely on whether WorkBuddy and Model-as-a-Service revenues compound fast enough to justify the balance-sheet deterioration.

Alibaba (BABA) reports next. Its cloud unit has been growing at comparable rates, and investors will be applying Tencent’s template to Alibaba’s capex line before the ink is dry on this morning’s release.

Baidu (BIDU) entered the AI cycle earlier and at smaller scale, which makes its margin path a useful point of comparison. If Baidu’s operating leverage holds while Tencent’s compresses, the difference in timing and product mix becomes a live debate.

Microsoft (MSFT) and Alphabet (GOOGL) are the ultimate stress test. Both carry higher base margins than Tencent and both have argued that AI demand will fill the capacity they are building. Tencent’s quarter does not prove them wrong. It does, however, show investors exactly what the downside scenario looks like when the filling takes longer than the building.