Alibaba Group (BABA): Goldman Sachs sees Alibaba's full-stack AI build-out supporting accelerating Cloud growth and EPS recovery

Overview

Goldman Sachs summarizes Alibaba Cloud's APSARA conference announcements and argues that Alibaba is strengthening a differentiated full-stack AI position spanning compute infrastructure, proprietary silicon, Qwen models and enterprise AI applications. The firm sees this as supporting above-consensus Cloud growth and an earnings recovery, while highlighting the funding and execution demands of the 20GW capacity ambition.

Core views

Goldman Sachs views APSARA as reinforcing Alibaba's progress across a full-stack AI strategy: infrastructure, models, and applications/ecosystems. Alibaba Cloud aims to reach 20GW of total data-center power consumption by 2032 through Alibaba Cloud and partners, supported by demand from intelligent driving, embodied AI, world models, frontier AI labs, physical AI and overseas customers. This compares with Goldman Sachs' estimate of roughly 5-6GW at end-2025 and implies a 1-2GW annual increase over the next several years. The report estimates this expansion could require Rmb200-300bn of annual Alibaba capex over the next few years, based on approximately US$20-25bn of capex per GW attributable to Alibaba. Goldman Sachs believes the funding requirement is manageable through stable e-commerce EBITA, growing Cloud profits, balance-sheet cash, future funding, partnerships with data centers and telecommunications providers, and overseas compute rental arrangements. The firm expects a mix of asset-heavy, asset-light and rental capacity. It also argues that T-Head custom silicon provides a cost advantage per GW versus domestic and overseas peers; T-Head could account for about half of Alibaba Cloud compute in the medium term, compared with about 10% currently. Alibaba's model strategy centers on maintaining Qwen among leading models while progressing toward agentic AI and artificial superintelligence. The company indicated that Qwen4 is coming soon and that Qwen4.5/5 could scale toward 5-10 trillion parameters, alongside multimodal capabilities and efforts toward recursive self-improvement. The report highlights Qwen3.8 Max user revenue growth of 8.5x and token consumption growth of 12x within two months versus the prior generation. For Qwen3.8 Flash, activated parameters were reduced to one-third of Qwen3.7 Flash, training cost to one-ninth, input-token pricing to one-quarter, ultra-long-context prefilling throughput improved 8.6x, and the AA score rose by 14 points. At the compute layer, Alibaba announced the Zhenwu V900 chip with three times the performance of the M890. The company is scaling supernode architecture toward a 500,000-PPU cluster and eventually a 1 million-PPU cluster. Goldman Sachs links this infrastructure expansion to the model roadmap and the need to make large-scale AI inference and agent workloads commercially viable. The report also highlights commercialization and enterprise adoption. Bailian, Qwen's MaaS platform, reached Rmb20bn ARR by end-August, up from Rmb16bn in mid-August and Rmb8bn in mid-May 2026, and remains on track for Rmb30bn by end-2026. QwenWork introduced Enterprise Context and DingTalk integration, reflecting Alibaba's focus on combining harnesses, private enterprise context and models to improve AI-agent productivity while preserving enterprise data security, access and control. International expansion is another component of the Cloud thesis. Alibaba Cloud covers 31 countries or regions, with 106 availability zones and more than 3,200 edge nodes. Planned data-center launches over the next 12 months include Malaysia, Turkey, Germany, the UAE, France, the Netherlands, Finland, Hong Kong and Thailand. The company cited roughly 100 days to build a new data center, 5-10x higher rack density than previous deployments, and PUE below 1.15. Goldman Sachs notes accelerating private AI Cloud growth among domestic and overseas enterprises. Goldman Sachs maintains above-Visible Alpha consensus Cloud revenue-growth expectations of 53%, 55% and 55% year-on-year for the Sep-2026, Dec-2026 and Mar-2027 quarters. It also notes Alibaba's expectation of US$100bn in external AI Cloud revenue by calendar 2030 and a medium-term 20% EBITA margin, contingent on capex-to-revenue conversion and the internal-versus-external revenue mix. The firm forecasts Rmb209bn, Rmb243bn and Rmb260bn of capex for FY27E, FY28E and FY29E. For group earnings, Goldman Sachs expects FY27E/FY28E EPS growth of 58%/27%, with a strong inflection from the September quarter. It attributes this to continued AI-and-Cloud leadership in China, further Cloud-growth acceleration, and recovery in e-commerce profits as quick-commerce losses narrow. The firm remains Buy-rated on BABA and 9988.HK, with SOTP-based 12-month targets of US$177 and HK$172, respectively.

Analysis framework

Goldman Sachs combines APSARA management commentary with its estimates of data-center capacity, capex needs, Cloud revenue growth, segment earnings and EPS. Its reasoning links AI compute demand and proprietary silicon economics to capacity expansion, then assesses whether e-commerce profits, Cloud profitability, partnerships and rental capacity can fund the build-out. The valuation conclusion uses a sum-of-the-parts framework.

Methodology notes

• Valuation methodsSOTP (Sum-of-the-Parts) Valuation

  Sum-of-the-parts valuation

  Goldman Sachs states that its 12-month price targets for BABA and 9988.HK are SOTP-based, meaning it values Alibaba through the contributions of its businesses rather than relying on one aggregate valuation measure.

• Industry AnalysisSupply-demand framework

  AI compute supply-demand and capacity build-out analysis

  The report relates expected demand from AI workloads to Alibaba Cloud's planned power capacity, annual capex and financing sources.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Alibaba Group (BABA.US)

  Primary covered ADR; Goldman Sachs links its Buy rating to AI Cloud leadership, Cloud-growth acceleration and e-commerce profit recovery.

  Strengths

  Full-stack AI capabilities spanning infrastructure, T-Head silicon, Qwen models and enterprise applications.

  Weaknesses

  Large planned AI infrastructure investment requires substantial capital and successful capex-to-revenue conversion.

  Comparison

  Goldman Sachs considers its Cloud growth expectations above Visible Alpha consensus and sees T-Head as offering US$/GW advantages versus domestic and overseas peers.

  Risks

  Lower GMV growth, slower China retail monetization, weaker strategic-investment execution and Cloud revenue-growth deceleration.

• Alibaba Group (H) (9988.HK)

  Primary covered Hong Kong listing; carries the same Buy rating and Alibaba operating thesis.

  Strengths

  Same group-level AI Cloud, Qwen and e-commerce recovery drivers as the ADR.

  Weaknesses

  Same capex and execution demands associated with Cloud expansion.

  Comparison

  Goldman Sachs sets a HK$172 SOTP-based 12-month target.

  Risks

  Lower GMV growth, slower China retail monetization, weaker strategic-investment execution and Cloud revenue-growth deceleration.

Key data

• Alibaba Cloud power-consumption target20GW by 2032Goldman Sachs estimates approximately 5-6GW at end-2025, implying a 1-2GW annual increase over the next few years.
• Alibaba annual capex forecastRmb209bn / Rmb243bn / Rmb260bnGoldman Sachs forecasts for FY27E/FY28E/FY29E.
• Alibaba Cloud revenue growth forecast53% / 55% / 55% year-on-yearFor the Sep-2026, Dec-2026 and Mar-2027 quarters; above Visible Alpha consensus according to Goldman Sachs.
• Qwen MaaS ARRRmb20bnAt end-August 2026, versus Rmb16bn in mid-August and Rmb8bn in mid-May; target is Rmb30bn by end-2026.
• FY27E/FY28E EPS growth+58% / +27% year-on-yearGoldman Sachs expects a strong September-quarter inflection.
• BABA / 9988.HK price targetsUS$177 / HK$172SOTP-based 12-month targets, implying 52.9% and 52.8% upside from the stated prices.

Impact & implications

Goldman Sachs argues that Alibaba's infrastructure scale, custom silicon, Qwen roadmap and enterprise AI products can sustain Cloud growth and support a broader earnings recovery. The key economic question is whether capacity investment converts into external Cloud revenue and margins quickly enough to support the targeted 20GW build-out.

Risks

• Lower-than-expected GMV growth resulting from macro conditions or competition.
• Slower-than-expected monetization in China retail.
• Weaker-than-expected execution in key strategic investments.
• Deceleration in Cloud revenue growth.

What to watch

• Progress toward the 20GW data-center power-consumption target and the associated capex and funding mix.
• Alibaba Cloud revenue growth relative to Goldman Sachs' 53%/55%/55% quarterly forecasts.
• Bailian MaaS ARR progress toward the Rmb30bn end-2026 target.
• Cloud capex-to-revenue conversion, external-versus-internal revenue mix and progress toward the medium-term 20% EBITA margin.