阿里云业务势头强劲但资本回报仍不明朗
阿里云业务势头强劲但资本回报仍不明朗
I'll start with the current BABA quote and the latest earnings and cloud disclosures.
Conclusion: Alibaba's cloud momentum is real and accelerating, but the June quarter does not yet show that AI capex is paying back. Revenue growth is strong, while profit and free cash flow have moved the wrong way. The payoff depends on margins and returns that the company has not yet reported. BABA's 4.91% gain on 4 November is a one-day move; I found no catalyst for it in the evidence I gathered.
Data as of: The quote is the NYSE close on 4 November 2026 (16:10 ET). The operating figures are from the June 2026 quarter, reported on 20 August 2026. The September-quarter report was not in the sources I could find.
1. What the cloud momentum implies
- AI Cloud and Compute Services revenue reached RMB 48.44bn, up 45% year on year, the fastest growth in 22 quarters (; ). That is 18.01% of group revenue of RMB 268.95bn (). Calculation: 48.44 ÷ 268.95 ().
- Demand looks broad. AI-related product revenue was RMB 12.4bn, with triple-digit growth for 12 straight quarters (). Reuters reports that the AI model-as-a-service business has passed RMB 16bn in annual recurring revenue ().
- Cloud profitability is improving from a low base. Cloud adjusted EBITA rose 133% to US$830m, a margin of 11.69% on US$7.1bn of revenue (). Calculation: 830 ÷ 7,100 ().
- Pricing helps. CNBC reports that Alibaba raised prices for AI compute and storage by as much as 34% in March ().
Read together, this is a real demand signal. The unresolved question is whether the cloud unit's margin can absorb the capacity build-out.
2. Why AI returns are hard to judge
- Capex is lumpy. Capex was RMB 67.7bn, up 75% year on year () and up 151.67% from RMB 26.9bn in March (). Calculation: 67.7 ÷ 38.7 − 1 (); 67.7 ÷ 26.9 − 1 (). Alibaba attributed the jump partly to the timing of customer purchases. It is therefore hard to tell how much is growth capex and how much is timing.
- Cost inflation muddies the signal. Management cited higher chip-component prices and more CPU capacity (). Higher unit costs can raise spending without adding proportionate capacity or revenue.
- Cash flow has deteriorated. Free cash flow was −RMB 44.7bn, against −RMB 18.8bn a year earlier (). Calculation: a swing of −25.9bn ().
- Headline profit mixes AI effects with other items. Net income fell 75.4% to RMB 10.4bn. Alibaba attributed this to lower investment-disposal gains and lower mark-to-market gains, as well as lower operating income, not AI spending alone (). Adjusted EBITA fell 29.6% to RMB 27.3bn ().
- Losses sit in the model layer. The AI labs and applications segment lost RMB 13.9bn, about 4.3 times the prior year's RMB 3.2bn (; ). Cloud returns and model-layer spending are therefore hard to separate.
- Adjusted EPS missed. Adjusted EPS per ADS was RMB 8.52 against an estimate of RMB 10.53 (). Calculation: −19.09% ().
- The payback target is management's claim. Management said it expects to break even on AI capex within three years, based on current average gross margins. It also has a RMB 380bn AI investment plan for 2026–29 (). The break-even depends on gross margins rising, which management links to its in-house chips replacing purchased ones. That shift has not yet appeared in reported numbers.
3. Disclosures that would show whether spending is paying off
- Cloud adjusted EBITA margin by quarter. This is the clearest test. A margin of about 12% on 45% growth is a good start, but investors need to see it rise as in-house chips scale.
- AI-related revenue and its growth rate. Track whether triple-digit growth continues as the base grows.
- Capex split and depreciation. Compare capex with cloud revenue growth, and track how quickly depreciation rises as servers come into service.
- Share of compute on in-house chips and the gross margin on that compute. This tests management's margin thesis directly.
- AI labs and applications losses. Watch whether they narrow as inference costs fall or as paid Qwen subscriptions scale.
- Free cash flow and capex prepayments. Track whether negative free cash flow narrows as the build-out matures.
- Adjusted and GAAP earnings reconciliation. Investment gains and mark-to-market effects distort the headline figures.
4. What would change this view
- Positive: Cloud margin rising toward double digits, a visible shift to in-house chips, and capex growing more slowly than cloud revenue.
- Negative: Cloud growth slowing while capex keeps rising, or cloud margins falling as depreciation builds.
Analysis: The cloud franchise is growing fast and the margin is improving, which supports the argument that AI demand is real. But the June quarter shows the cost side moving faster than profit. Until the company shows that capex growth is slowing relative to cloud revenue, and that the cloud margin is rising, payback remains a forecast rather than a result.
Sources
- Alibaba Group IR, "Alibaba's Full-Stack AI Accelerates Monetization with 22-Quarter-High Cloud Growth," 20 Aug 2026 (primary): link
- CNBC, "Alibaba shares fall 5% as AI spending drives 75% drop in net income," 20 Aug 2026: link
- Reuters via BNN Bloomberg, "Alibaba's quarterly revenue up 9%," 20 Aug 2026: link
- KrASIA, "Alibaba's cloud growth comes with a rising AI bill," 21 Aug 2026: link
- Driven market quote for BABA (NYSE), 4 Nov 2026 close, and Driven daily price history, 17 Aug–4 Nov 2026 (dividend-adjusted)