BABA · Forward model · AI Labs and Applications · Bear case
What has to happen in AI Labs and Applications
Model as of
This page changes AI Labs and Applications 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 Labs and Applications
The burn, finally given its own reporting line. AI model labs, the Qwen Consumer Business Group and QwenWork were pulled out of All others this quarter. The segment lost RMB13,861 million of adjusted EBITA on RMB3,338 million of revenue - a margin of -415% - against a RMB3,224 million loss a year earlier. The release attributes the widening to increased investment in AI capabilities and higher inference cost related to the Qwen app.
Latest: $1.06B (2031Q2E)
| Period | Value |
|---|---|
| 2026Q2 | $492M |
| 2026Q3E | $506M |
| 2026Q4E | $522M |
| 2027Q1E | $539M |
| 2027Q2E | $557M |
| 2027Q3E | $576M |
| 2027Q4E | $597M |
| 2028Q1E | $619M |
| 2028Q2E | $643M |
| 2028Q3E | $668M |
| 2028Q4E | $694M |
| 2029Q1E | $722M |
| 2029Q2E | $752M |
| 2029Q3E | $783M |
| 2029Q4E | $816M |
| 2030Q1E | $851M |
| 2030Q2E | $888M |
| 2030Q3E | $927M |
| 2030Q4E | $968M |
| 2031Q1E | $1.01B |
| 2031Q2E | $1.06B |
Assumptions & reasoning
- This segment did not exist before the June 2026 recut. Its only comparative anywhere is the restated June 2025 quarter, RMB2,882 million of revenue and a RMB3,224 million adjusted-EBITA loss.
- The loss, not the revenue, is what this vertical contributes: -RMB13,861 million of quarterly adjusted EBITA against group adjusted EBITA of RMB27,329 million. At 1.2% of group revenue it takes half the group's operating profit.
- This is the only vertical whose terminal growth is set ABOVE its current rate. That is an assumption about Qwen app monetisation, not a disclosure: the release says 250 million users have had their first AI-driven shopping experience through the app's agentic features, but gives no revenue per user and no monetisation metric of any kind.
- Neither the release nor either transcript sets a date or a path for narrowing this loss. The glide from -398% to -20% of revenue is entirely judgement, and the segment is still loss-making in the terminal year under it.
- Qwen weights are open - Qwen3.8-Max at 2.4 trillion parameters - so adopters can run the model without paying Alibaba anything, while inference cost for the consumer app scales with usage before revenue does.