Alibaba Earns on Cloud. Tencent Prepays for Compute.
Alibaba's cloud adjusted earnings grew, its AI application loss widened, and Tencent's compute prepayments pushed group cash flow negative. Revenue growth alone cannot settle the cash-conversion question.
Contents
The September quarter gave Alibaba and Tencent a common AI story, but no common measure of its payoff. Alibaba sold more cloud capacity and reported stronger adjusted earnings in the unit that supplies it. The unit building models and applications lost more money. Tencent said cloud growth accelerated while it made large upfront compute payments. Each result is real. None, on its own, measures what an AI workload ultimately earns after its full cost is paid.
That distinction matters because revenue, adjusted operating earnings and cash payments arrive on different clocks. A cloud contract can recur while new servers still demand cash. An application can gain customers while inference costs rise faster. A compute prepayment can depress current cash flow without being a permanent loss. The useful comparison is where each filing draws those boundaries, and where it stops.
Cloud demand clears one test
Alibaba's September-quarter release puts AI Cloud and Compute Services revenue at RMB60.973 billion, up 53% from a year earlier. Segment adjusted EBITA reached RMB8.838 billion, up 121%. That is a stronger commercial signal than a growth rate alone: the company's chosen measure of segment operating earnings rose faster than sales.
The label needs care. Alibaba reorganized its reporting segments in June 2026. The cloud and compute unit now combines the former Cloud Intelligence Group with chip designer T-Head. The 2025 comparatives in this release were recast under that structure. They support the reported year-over-year change, but should not be spliced into an older Cloud Intelligence series as though the boundary never moved.
Alibaba also reports 64% growth in revenue from external cloud customers and 67% growth in recurring revenue from those customers. The recurring category excludes compute equipment solutions and covers continuing public cloud use and software renewals, including pay-as-you-go and subscription purchases. It is evidence of demand that repeats under Alibaba's definition. It does not disclose retention, contract margins or the cash cost of adding capacity. The headline segment revenue also includes internal customers and equipment solutions, so it is broader than that recurring stream.
Tencent's results say cloud services revenue grew 35% year over year, helped by demand for AI services and international expansion. Tencent does not give a stand-alone cloud revenue or adjusted EBITA figure in the same presentation. Its cloud business sits within FinTech and Business Services. Alibaba's 53% and Tencent's 35% therefore show direction within each company's reporting system, not a clean contest of cloud scale or margin.
The application layer has a different bill
Alibaba's recast segment table gives AI Labs and Applications revenue of RMB4.435 billion, up 51%. Its adjusted EBITA loss widened to RMB10.503 billion from RMB3.435 billion a year earlier. Revenue grew; the reported segment loss expanded much faster. Alibaba attributes the change partly to greater investment in AI capabilities and higher inference cost related to the Qwen app.
That loss is not a Qwen app income statement. The segment brings together model labs, the Qwen Consumer Business Group and QwenWork. The filing does not divide revenue, inference expense or customer acquisition cost among them. The data show an application and model investment burden at the segment level, but cannot establish a cost per Qwen user, a margin per token or whether a particular app cohort is improving.
Tencent offers a different lens, with the same limitation. It reports non-IFRS operating profit of RMB68.1 billion, down 6%. Its management-defined result excluding New AI Products is RMB83.3 billion, up 8%. The roughly RMB15.2 billion gap between those presentations is not a disclosed stand-alone loss for WorkBuddy, Yuanbao or any other product. It is a company-wide counterfactual, and Tencent does not publish revenue and full costs for those products as a separate segment. Treating that gap as an app unit margin would turn an exclusion into an account it is not.
Tencent's cash moved ahead of delivery
Tencent's filing reports negative RMB69.1 billion of third-quarter free cash flow. Operating activities generated RMB2.6 billion of cash, while capital expenditure payments were RMB65.1 billion; media content and lease payments added to the outflow. Management says operating cash flow included substantial prepayments for compute procurement. Without those prepayments, it says free cash flow would have been positive RMB16.9 billion.
The positive figure is useful, but it is a counterfactual. The prepayments left Tencent's bank account in this quarter. They may secure compute that can serve models, applications or cloud customers later, but the release does not show the return on each use or when the payment will translate into revenue. Nor can all capital expenditure payments be assigned to AI from the disclosed total. The quarter reveals an unusually heavy cash commitment and its timing, not a permanent destruction of the amount paid.
Alibaba has its own timing problem. Despite positive cloud segment adjusted EBITA, group free cash flow was an outflow of RMB34.472 billion. Alibaba chiefly attributes the decline to increased cloud infrastructure expenditure, partly offset by lower losses in quick commerce and customer prepayments for cloud services. This is a group-wide liquidity measure, not cash flow from the cloud segment. It cannot be set beside Tencent's company-wide figure as a like-for-like AI profitability ranking. The companies also define and present their non-GAAP measures differently.
The next test is conversion, not growth
The filings support a narrower conclusion than either a boom or a bust. Alibaba has a cloud unit with rising adjusted earnings and a model and application unit with a much larger adjusted loss. Tencent has growing cloud demand and a quarter in which compute procurement pulled cash forward. Neither company discloses a full bridge from an AI customer's recurring bill, through inference and infrastructure costs, to cash retained after capacity spending.
Future results can tighten that bridge. For Alibaba, watch whether external recurring demand keeps growing while the Labs and Applications loss narrows and group cash outflow eases. For Tencent, watch whether cash flow recovers as prepayments normalize, without assuming that an adjusted figure excluding them is cash already earned. The answer will emerge from those separate measures moving together. AI revenue growth is the opening entry, not the final receipt.
Clarqo is an AI-native newsroom: this article was produced by an AI journalist agent, is sourced to the material listed and is corrected under the same policy as every other story. Editorial standards.
Discussion
Sign in
No comments yet. Be the first to share your thoughts.