Alibaba (BABA / 9988.HK) — SOTP-DCF With the AI Cloud and AI Lab Priced Separately, Sep 2026

20 September 2026 · Price $113.24 (NYSE close 18 Sep) · Market cap $281.5B on 19.885B shares post-placement · Method: Damodaran three-scenario DCF plus narrative-shift case, cross-checked to peers

1. Documented facts (June-quarter 2026 6-K)

Alibaba re-segmented this quarter. T-Head was combined with Cloud Intelligence to form AI Cloud and Compute Services. The model labs, Qwen consumer business and QwenWork were pulled out of All Others into a new AI Labs and Applications segment, so the lab burn is visible for the first time.

• AI revenue: AI-related product revenue was RMB 12.4B. CEO Eddie Wu: "AI-related product revenue delivering triple-digit growth for the twelfth consecutive quarter." On the call, management said MaaS ARR passed RMB 16B in August and kept the RMB 30B year-end target.
• Core commerce stalled: customer management revenue fell 7%, or +1% like-for-like excluding the contra-revenue program. The company attributes the slowdown to weaker transaction activity.
• Cash burn: capex RMB 67.7B (+75%); free cash flow was an outflow of RMB 44.7B. Management reiterated the three-year RMB 380B plan with RMB 190B spent through June. The lab loss of RMB 13.9B narrowed sequentially as Qwen app marketing fell.
• Dilution: on 26 Aug the company completed a placing of 710M new shares at HK$112.70 (HK$80B gross), 3.70% of the 19.175B shares then in issue. HK$112.70 equals $114.92 per ADS, so the ADS now trades 1.5% below the placing price. Buybacks were US$162M in the quarter.
• T-Head: Zhenwu chips serve more than 650 external cloud customers. A T-Head IPO was reported by Bloomberg and Reuters in January 2026 with no timetable; the company has not confirmed it.

2. Discount rate

Cash flows are in RMB, so the USD WACC is converted with an inflation differential of 3% US vs 1% China. The 3% US figure is my assumption, not a verified breakeven. A pure RMB risk-free build gives 7.3%; I did not use it because the suppressed Chinese yield curve makes it too generous. The result sits inside the 7–12% sanity band.

3. Scenario DCF (RMB bn)

Ten explicit years, FY27 to FY36. SBC is expensed. Convertible notes are treated as debt. Net cash includes the placement proceeds and the July dividend. Cloud net reinvestment is hard-coded at RMB 165B and 140B for FY27–28, then driven by sales-to-capital. Quick-commerce losses get no tax shield against Taobao profits.

• Base cloud: growth fades from 45% to 4%; FY36 revenue RMB 823B. Margin dips to 11.5% in FY28 as depreciation catches up with capex, then climbs to 22%. Sales-to-capital 0.8.
• Base labs: losses of RMB 47B, 38B, 28B, 18B and 8B, breakeven in FY32, FY36 revenue RMB 194B at a 15% margin.
• Base e-commerce: growth 3.5% fading to 1.7%; margin recovers from 16.5% to 19% as quick-commerce losses narrow.
• Bear: e-commerce margin slides to 14%, cloud terminal margin 15% at sales-to-capital 0.65, lab burns through FY33.
• Narrative shift (my interpretation, not documented): triggered by the order-book test, since MaaS ARR is targeted to nearly double by year-end. Bull operations, lab floored at 0.75x Zhipu market cap, plus RMB 140B for a listed T-Head.

Base bridge per ADS: e-commerce $99, cloud $30, labs −$3, others and corporate −$2, net cash $16, investments $21, NCI and fine −$4.

PWFV sensitivity, RMB WACC by terminal growth

4. Relative valuation vs peers

Multiples from stockanalysis.com, pulled 20 Sep 2026.

On headline multiples BABA looks expensive against Chinese peers. The cause is roughly RMB 100B a year of lab and quick-commerce losses sitting inside its earnings, which PDD and Tencent do not carry.

5. Consistency gate

Group level passes: DCF $157 vs peer-multiple base $181. Cloud alone fails the 2x test: RMB 499B on DCF vs RMB 1,145B on peer multiples, a 2.3x gap.

Explanation: peer multiples capitalise growth without charging the front-loaded capex. Oracle and CoreWeave are also free-cash-flow negative. The DCF charges RMB 305B of net reinvestment in FY27–28 before the margin ramp arrives. I rely on the DCF side, which is the more conservative of the two.

6. Reverse DCF

• The price implies an RMB WACC of 11.1%, roughly 13% in USD terms, against my 8.6%.
• Market-implied operating EV is RMB 1,348B. With e-commerce at 9x EBITA the residual is RMB 115B, about 0.5x sales for a cloud business growing 45%.
• At my WACC, e-commerce plus net assets alone exceed the share price. The market is valuing the AI businesses below zero.

7. Pricing section (the trade call)

• Supportive: Trump-Xi summit on 24 Sep with chips on the agenda; September-quarter results in November; any T-Head listing news; progress toward the RMB 30B MaaS ARR target.
• Against: US 10Y touched 5.04%, the highest since 2007, and the Fed is hiking; the trade truce expires in November; free cash flow runs near −RMB 45B a quarter, so a second equity raise is plausible; placement holders are slightly underwater; the buyback has all but stopped.
• Momentum: $192.67 high in Oct 2025, $91.99 low in June 2026, $113.24 now.
• Not verified: short interest and float turnover.

Thesis-break triggers

• Cloud revenue growth below 30%.
• Lab loss re-widening past RMB 15B a quarter.
• Like-for-like CMR negative for two consecutive quarters.
• Another equity raise within 12 months.

8. Caveats and cadence

• The FY26 e-commerce revenue base under the new segment definition (RMB 800B) is my estimate; the company has recast only the June quarters.
• Roughly 29% of cloud segment revenue is inter-segment. Segment EBITA still sums to the group, but external-sales multiples would be lower than the 5x shown.
• Ant Group is carried inside equity-method book value with a 20% haircut in the base case; I have not built a standalone Ant valuation.
• VIE structure and ADR delisting risk are not in the discount rate; they belong in the bear weight.
• This DCF will be re-run at the next two earnings (November 2026 and February 2027). A price divergence above 50% triggers a versioned V2 with attribution.

Sources: Alibaba 6-K filings for the March and June 2026 quarters and the August 2026 placing; earnings-call coverage (Investing.com, Yahoo Finance); Damodaran industry betas and country risk premiums, January 2026; Trading Economics for bond yields; stockanalysis.com for peer multiples; Polygon.io for price.

This is independent research for information only, not investment advice. The author may hold positions in securities mentioned.