Kingsoft Cloud's AI cloud transition enters harvest period; MaaS brings further upside

Overview

This report is Morgan Stanley's initiation of coverage on Kingsoft Cloud Holdings. The report believes that Kingsoft Cloud transitioned early and executed decisively from a mid-tier commoditized cloud vendor toward AI cloud and neocloud, and has already benefited from training demand, GPU capacity shortages, and cloud service price increases, with further potential to improve revenue quality and margins through MaaS and inference demand.

Core views

The core views include: first, Kingsoft Cloud was one of the earliest among major Chinese public cloud vendors to bet on AI cloud, with rapid growth in AI-related revenue; second, the Xiaomi and Kingsoft Group ecosystems span smartphones, AIoT, EVs, office software, and gaming, providing differentiated AI use cases and compute demand; third, chip shortages have increased cloud vendors' pricing power, and GPU cloud unit economics are superior to traditional cloud workloads; fourth, customer prepayments and finance leases help ease the balance-sheet constraints of this capital-intensive business; fifth, if MaaS scales based on open-source LLMs, it could bring margin upside above the base-case scenario.

Analysis framework

The report uses company financial forecasts, AI cloud market sizing, IDC data, CIO surveys, cloud vendor capex cycles, unit economics, and a relative valuation framework for analysis; valuation centers on 2027e EV/EBITDA and compares the company at a discount to US neocloud peers.

Methodology notes

• Valuation methodsEV/EBITDA relative valuation

  Assigning a 5.5x EV/EBITDA multiple based on 2027e EBITDA.

  This multiple is below the roughly 10x median of US neocloud peers, reflecting uncertainty in Kingsoft Cloud's supply chain availability and return on assets.

• Industry analysisAI cloud market segmentation

  The report defines AI cloud as IaaS + MaaS and does not include AI SaaS in the market size estimate.

  Growth at the infrastructure layer is more visible, while MaaS sits between the infrastructure and platform layers; future growth in inference and token usage is an important demand driver.

• Operating analysisUnit economics and cash flow analysis

  Analyzing server life cycle, finance leases, customer prepayments, and project IRR.

  The report believes that longer server useful lives and prepayment arrangements can improve cash flow and raise equipment IRR when pricing is unchanged.

Asset mapping & comparison

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

• KINGSOFT CLOUD HOLDINGS LTD (KC.US)

  Core coverage target; beneficiary of AI cloud and MaaS transformation.

  Strengths

  Rapid AI cloud revenue growth, pricing power in GPU cloud, demand support from the Xiaomi and Kingsoft ecosystems, and improved cash flow through customer prepayments and finance leases.

  Weaknesses

  High capital intensity, uncertainty in supply chain availability, and legacy businesses such as traditional CDN may dilute overall growth.

  Comparison

  The target multiple of 5.5x 2027e EV/EBITDA is below the 10x median of US neocloud peers.

  Risks

  Insufficient chip procurement, higher-than-expected interest rates, slower AI development in China, and MaaS margins falling short of expectations.

• Xiaomi ecosystem

  Shareholder and ecosystem partner, providing Kingsoft Cloud with compute demand related to smartphones, AIoT, EVs, and AI models.

  Strengths

  The MiMo model and continued AI investment could drive upward revisions in GMV and incremental business volume.

  Weaknesses

  The market may view Kingsoft Cloud as a high-beta proxy for Xiaomi, resulting in high short-term share price correlation.

  Comparison

  Compared with a single external customer, ecosystem cooperation provides a clearer entry point for use cases.

  Risks

  Progress in Xiaomi's AI models may fall short of expectations, or pressure in consumer electronics could hurt market sentiment.

• US neocloud peers

  Valuation reference group.

  Strengths

  Provide a valuation anchor for AI compute leasing and cloud infrastructure.

  Weaknesses

  They differ from Kingsoft Cloud in supply chain, depreciation policy, return on assets, and regional demand.

  Comparison

  The report values Kingsoft Cloud at a multiple discounted to the peer median.

  Risks

  If peer valuations compress or AI compute leasing prices decline, Kingsoft Cloud's valuation upside could be constrained.

Key data

• Initiation ratingOverweightMorgan Stanley lists Kingsoft Cloud as a preferred name within its Greater China IT services and software coverage.
• Target priceUS$15.00Based on 5.5x 2027e EV/EBITDA.
• Current priceUS$9.10The report table shows the pricing date as 2026-07-03.
• Revenue CAGR35%The report forecasts 2025-2028 revenue CAGR.
• Adjusted EBITDA CAGR79%The report forecasts 2025-2028 adjusted EBITDA CAGR.
• AI revenue share34% in 4Q25, expected to exceed 40% in 2026 and 60% in 2028From the report's front-page investment summary.
• AI cloud as share of public cloud revenue49% in 4Q25, expected 61% in 2026 and 78% in 2028From the executive summary.
• 2026 AI revenue growth109%Driven by strong compute demand and Rmb15bn in server investment.
• 2026 China GenAI IaaS + MaaS market forecastFrom Rmb15bn in 2024 to Rmb218bn in 2029, CAGR 72%The report cites IDC data.
• 2026 China hyperscaler capex forecastRmb602bn / US$86bn, up 41% YoYThe report believes early-cycle cloud growth is mainly driven by supply and capital expenditure.

Impact & implications

If the report's view proves correct, the investment thesis for Kingsoft Cloud would shift from repairing the low margins of traditional cloud to high-elasticity growth driven by AI compute supply, inference demand, and MaaS monetization. The investment implication is that its valuation could move closer to neocloud assets, though near-term constraints remain from chip supply, financing costs, the pace of AI model development, and reliance on a single ecosystem.

Risks

• Supply-side constraints could lead to insufficient procurement of AI chips or servers, thereby limiting incremental revenue.
• Higher-than-expected interest rates could increase financing costs and reduce project returns.
• China's AI model development or token consumption growth may be slower than expected, potentially weakening AI cloud demand.
• A significant increase in supply or earlier launch of next-generation servers could put pressure on rental pricing for Kingsoft Cloud's installed server base.
• If leading Chinese LLMs shift from open-source to closed-source, upside in Kingsoft Cloud's MaaS margins based on open-source LLMs may be constrained.
• If the AI LLM market undergoes downstream consolidation or major players exit, short-term training demand could come under pressure.
• Legacy cloud businesses such as traditional CDN may dilute overall growth.

What to watch

• Whether AI revenue growth in 2026 can reach the 109% forecast in the report.
• Whether AI cloud's share of revenue and of public cloud revenue continues to increase.
• Whether MaaS business contribution materializes and narrows the margin gap with industry leaders.
• Progress of Xiaomi's MiMo model and Xiaomi's AI ecosystem investment.
• Changes in China's AI chip supply, domestic AI wafer capacity expansion, and availability of Nvidia H200 or equivalent chips.
• The actual improvement in cash flow and IRR from customer prepayments, finance leases, and server life cycle extension.
• Whether public cloud price increases continue and whether higher pricing can offset upstream cost inflation.