JPM Alibaba Group 1

Asia Pacific Equity Research 21 August 2026

Alibaba Group Holding (BABA US & 9988 HK) AI Capex: From Spend to Return

Our July preview argued Alibaba was entering a positive earnings-revision cycle China and we recommended owning into the print. We believe Jun Q 26 validates that Head of China Equity Research call and moves our thesis from anticipating revisions to measuring the returns that and Co-Head of Asia TMT drive them. We raise FY28E adj EPS by 4% (GAAP), our second consecutive Research increase after July's first upgrade since May 2025, and set our Dec-26 PTs at Alex Yao AC US$210/HK$205. Alibaba remains China's largest listed AI and cloud franchise: ( we forecast external cloud revenue growth above 50% with expanding margins SAC Registration Number: S and estimate ~22% after-tax project IRR on incremental AI capex versus a ~10% cost of capital. We suggest investors add ahead of the Sep-quarter print. Daniel Chen ( • The key pushback from our August marketing was that record capex and FCF outflows make the cloud asset difficult to underwrite until returns are proven. SAC Registration Number: S We believe Jun Q 26 provided enough disclosure to reconstruct those returns Olivia Xu independently. Our vintage model reconciles mgmt’s disclosed payback and ( post-breakeven cash-generation framework and uses those inputs to derive an implied ROIC profile, implying ~Rmb36 of NPV per Rmb100 invested, a ~2.9- SAC Registration Number: S year after-tax payback and ~22% project IRR. We believe the reported 12% Nancy Liu cloud margin understates mature-vintage economics near 20%, as record capex ( has left the asset base unusually young. As vintages mature, blended returns SAC Registration Number: S should rise and FCF inflect around FY29 at stable capex. The debate should J.P. Morgan Securities (China) Company shift from how much Alibaba spends to what the marginal RMB invested earns. Limited • Investors we met framed Alibaba as either an AI/cloud rerating or a legacy- commerce earnings-cut story, with resolution requiring sustained cloud acceleration, margin expansion and bounded investment losses. We count Jun Q as quarter one and forecast quarter two: September cloud growth above 50%, continued margin progression and a sequential decline in AI Labs and Applications losses, which we estimate peaked at Rmb13.9bn. We also expect quick-commerce losses to halve this fiscal year and trend toward FY29 profitability. Meanwhile, we estimate the core ecommerce business generates Rmb180–190bn of annual EBITDA, providing the internal funding capacity to absorb the investment cycle and narrowing the plausible earnings-downside case. • Our Dec-26 PTs are based on 16x FY28E P/E and cross-checked against the SOTP now supported by Alibaba's new segment disclosure. For investors debating profit-taking after the rally, we think risk/reward remains attractive into specific September proof points: >50% cloud growth, further margin expansion and declining AI/application losses. Key risks are renewed loss expansion around major model launches, capex stepping up rather than plateauing, compute-price deflation as supply tightness eases, weaker consumption constraining CMR recovery, and limited near-term capital returns. Equity Ratings and Price Targets Mkt Cap Price Rating Price Target Company Ticker ($ mn) CCY Price Cur Prev Cur End Prev End Date Date Alibaba Group Holding Limited (BABA) BABA US 312,889 USD 130.53 OW n/c 210.00 Dec-26 205.00 n/c Alibaba Group Holding Limited 9988 HK 308,632 HKD 126.20 OW n/c 205.00 Dec-26 200.00 n/c Source: Company data, Bloomberg Finance L.P., J.P. Morgan estimates. n/c = no change. All prices as of 20 Aug 26.

See page 12 for analyst certification and important disclosures, including non-US analyst disclosures.

Alex Yao Asia Pacific Equity Research ( August 2026

AI+Cloud: infrastructure returns are now measurable - and improving Alibaba's June quarter gives investors the first auditable return framework for its AI infrastructure investment. Until now, the market broadly accepted that cloud demand was real but remained reluctant to capitalize the growth because neither the return on AI capex nor the trajectory of…