BABA · Forward model · AI Cloud and Compute Services · Wu case
What has to happen in AI Cloud and Compute Services
Model as of
This page changes AI Cloud and Compute Services inside the complete BABA model. Every other vertical, corporate programme and valuation assumption stays in the selected portfolio case.
Shares this vertical and portfolio case. Slider and horizon edits stay in your browser.
AI Cloud and Compute Services
Basis quarter$7.14B
Final quarter$22.55B
Implied CAGR+26%
Final revenue mix38%
The reason this model exists. Cloud Intelligence Group plus T-Head silicon, growing 45% with revenue from external customers growing 45% too - management called it the fastest in twenty-two quarters - on AI-related product revenue of RMB12,376 million that has now compounded at triple digits for twelve consecutive quarters. It is also where the RMB67,678 million of quarterly capital expenditure and the HK$80 billion equity placement are going.
Last four quarters
2026 Q2
$7.14B
Reported
Public cloud infrastructureModel-as-a-Service (Qwen) and inferenceHybrid and managed cloudT-Head proprietary silicon (Zhenwu GPU, CPU, storage, networking)
Sequential growth
+9.7%/qtr
decaying toward +2.9%
9.7% a quarter compounds to the 45% year-over-year the segment just printed, on total and external revenue alike.
AI Cloud and Compute Services
Latest: $22.55B (2031Q2E)
| Period | Value |
|---|---|
| 2026Q2 | $7.14B |
| 2026Q3E | $7.86B |
| 2026Q4E | $8.60B |
| 2027Q1E | $9.33B |
| 2027Q2E | $10.08B |
| 2027Q3E | $10.82B |
| 2027Q4E | $11.57B |
| 2028Q1E | $12.32B |
| 2028Q2E | $13.07B |
| 2028Q3E | $13.83B |
| 2028Q4E | $14.59B |
| 2029Q1E | $15.35B |
| 2029Q2E | $16.11B |
| 2029Q3E | $16.88B |
| 2029Q4E | $17.66B |
| 2030Q1E | $18.45B |
| 2030Q2E | $19.24B |
| 2030Q3E | $20.05B |
| 2030Q4E | $20.87B |
| 2031Q1E | $21.70B |
| 2031Q2E | $22.55B |
Assumptions & reasoning
- There is no capacity driver available for this line. Alibaba discloses no megawatts, no accelerator count, no utilisation and no contracted backlog, so a capacity model would have to invent the earning base. Growth anchored on disclosed monetisation - AI-related product revenue and MaaS ARR - is the honest choice.
- The segment's disclosed adjusted EBITA margin is 11.62% (RMB5,628 million on RMB48,437 million), which is the '12%' the CFO quotes. It is an EBITA margin, not an EBITDA margin and not a gross margin. The 28.7% EBITDA margin used here adds an assumed 70% of group depreciation of RMB11,814 million. A widely repeated claim that cloud EBITDA margin was 12% against 7.6% a year ago is wrong: no segment EBITDA is disclosed anywhere.
- Capex intensity of 112% of segment revenue is 80% of the disclosed group capital expenditure line. That single assumption moves the fair value more than the growth rate does, and it is the control to move first. The glide to 35% is judgement, not guidance: management has given no capital expenditure ceiling in any filing, and the CEO has said the RMB380 billion 2026-29 programme is likely to be overshot.
- Disclosed monetisation anchors, all for the June 2026 quarter or the August call: AI-related product revenue RMB12,376 million, which is 25.6% of segment revenue; an annualised AI product run rate above RMB49.5 billion, which management expects to approach US$10 billion next quarter; MaaS ARR past RMB16 billion in August against a RMB30 billion year-end target; Omdia ranking Alibaba Cloud first in China's AI cloud market at 38.1% share in 2025.
- A claim that AI product revenue is 35% of EXTERNAL cloud revenue cannot be checked: Alibaba publishes external cloud revenue only as a growth rate for this quarter, never as an amount. 25.6% of total segment revenue is the ratio that is actually computable from the filing.
- Zhenwu silicon, including the M890, is described in the release as having broad commercial adoption through Alibaba Cloud services with more than 650 external customers across over twenty industries. The chip is Zhenwu, not the 'Chunwu' some third-party transcripts render.