AI Compute Monetization Could Drive AliCloud to 13%-20% ROIC

Overview

This report assesses the capital-return outlook for Alibaba's cloud business using an AI compute unit economics model. Morgan Stanley believes Alibaba can progressively increase cloud revenue and margins during a period of heavy capital expenditure, supported by its AI infrastructure scale, cloud capabilities, and Qwen model capabilities, creating a path to 13%-20% ROIC.

Core views

The report divides the commercialization paths into self-built GPU IaaS, leased/neocloud IaaS, and self-built MaaS. Self-built IaaS has higher incremental margins but requires upfront capital expenditure; the leasing model has lower margins but immediate cash returns; MaaS has the greatest margin and ROIC potential when inference throughput, the inference mix, and pricing improve. The report believes the key to improving Alibaba Cloud's margin from its current approximately 11%-12% toward its long-term target lies in AI demand, advanced compute utilization, the MaaS revenue mix, and a greater share of inference workloads.

Analysis framework

Drawing on the US Internet team's framework, the report uses a bottom-up approach calculated per 8-GPU AI server. It separately analyzes the effects of revenue pricing, server costs, depreciation, IDC and energy costs, operating expenses, utilization, token throughput, the training/inference mix, and token pricing on margins, ROIC, and cash payback periods. Valuation uses a discounted cash flow model.

Methodology notes

• Unit Economics ModelSelf-built GPU IaaS

  Calculating the economics of leasing out owned compute capacity on a per-8-GPU-server basis

  Key variables include server lease pricing, upfront server capital expenditure, depreciation, IDC and energy costs, and utilization, which are used to derive operating margin, ROIC, and cash payback period.

• Unit Economics ModelLeased/neocloud IaaS

  Leasing third-party compute capacity and then renting it out to end customers

  Profitability in this model mainly depends on the spread between customer lease pricing and upstream compute lease costs; because it does not bear underlying server capital expenditure, cash returns can be generated immediately once the spread turns positive.

• Unit Economics ModelSelf-built MaaS

  Monetizing owned compute capacity through model APIs and MaaS

  Core drivers are token throughput per GPU, the share of inference in workloads, and token pricing; training does not directly generate revenue, but affects model capabilities and subsequent pricing.

• Valuation methodsDiscounted Cash Flow Model

  Estimating value by discounting future cash flows

  The base case assumes a WACC of 10% and a perpetual growth rate of 3%.

Asset mapping & comparison

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

• BABA.N

  Research Subject

  Strengths

  Has large-scale AI infrastructure, AliCloud cloud capabilities, and Qwen model capabilities, enabling multi-layered AI compute monetization through IaaS and MaaS.

  Weaknesses

  Current cloud business margin is approximately 11%-12% and remains affected by traditional lower-margin workloads, early-stage infrastructure, utilization ramp-up, and R&D and personnel costs.

  Comparison

  China's self-built IaaS base-case ROIC is approximately 13%, versus approximately 31% in the US; China's MaaS ROIC is approximately 19.5%, versus approximately 46.2% in the US, mainly due to higher hardware capital expenditure and depreciation.

  Risks

  Intensifying competition, capital expenditure exceeding expectations, weak consumption, slower enterprise digitalization, and increased regulation of internet platforms.

• 09988.HK

  Alibaba's Hong Kong-listed shares

  Strengths

  Together with BABA.N, reflects the value of Alibaba's cloud computing, AI, and e-commerce businesses.

  Weaknesses

  The report does not provide an independent valuation or target price for these listed shares.

  Comparison

  The report primarily presents ratings, price, and target price for BABA.N.

  Risks

  The same risks related to group operations, cloud business monetization, and the regulatory environment.

Key data

• Investment RatingOverweight; Top PickThe rating expectation is based on risk-adjusted total return performance relative to the industry coverage universe over the next 12-18 months.
• Target PriceUS$180.00Represents 45% potential upside versus the US$123.81 closing price on 2026-08-14.
• Self-built GPU IaaS Base CaseOperating margin of approximately 44%, ROIC of approximately 13%, and a cash payback period of 3.1 yearsAssumes server capital expenditure of more than RMB8mn per server and long-term lease pricing of approximately RMB250k/server/month.
• Leased/neocloud IaaS Base CaseOperating margin of approximately 20%Assumes monthly customer rent of RMB250k and monthly upstream compute rent of RMB200k; no upfront server capital expenditure.
• Self-built MaaS Base CaseGross margin exceeding 76%, operating margin of approximately 53%, ROIC of approximately 19%, and a cash payback period of approximately 2.5 yearsAssumes 4,000 tokens/second/GPU, a 50% inference mix, and blended token pricing of RMB9.5/million tokens.
• AliCloud Long-Term TargetExternal revenue of US$100bn and EBITA margin of 20%The report believes this target represents approximately 50% upside versus its estimate.
• MaaS Annual Recurring Revenue TargetMore than RMB30bn by year-endManagement targets more than RMB10bn by the end of June and more than RMB30bn by year-end.

Impact & implications

For Alibaba, rising capital expenditure may weigh on earnings and free cash flow in the near term, but if advanced compute maintains high utilization and the company successfully expands MaaS and inference businesses, the cloud business's incremental margins and capital returns could improve significantly. For the industry, the main return constraint for China's AI infrastructure is higher server and hardware costs; declining hardware costs, improved model-iteration efficiency, and a shift from training to monetizable inference will determine whether the China-US unit economics gap can converge.

Risks

• Intensifying competition could pressure cloud-service and token pricing.
• Capital expenditure and reinvestment costs could exceed expectations, extending cash payback periods and reducing ROIC.
• AI demand, advanced compute utilization, or enterprise digitalization could fall below expectations.
• An excessively high training mix and a slower-than-expected increase in the inference mix could limit MaaS monetization and margins.
• A weak consumer recovery could weigh on core e-commerce monetization and earnings growth.
• Internet platforms could face additional regulatory scrutiny.

What to watch

• AliCloud's external revenue growth, EBITA margin, and changes in advanced compute utilization.
• The share of AI revenue in cloud business revenue and progress toward the MaaS ARR target of more than RMB30bn.
• Changes in token throughput, inference mix, pricing, and model efficiency for Qwen and related models.
• GPU server procurement costs, IDC and energy costs, and resulting changes in capital expenditure intensity.
• Improvement in Alibaba's core e-commerce monetization and the sustainability of enterprise digitalization demand.
• Whether the China-US gap in AI infrastructure unit economics narrows due to lower hardware costs and inference monetization.