Global cloud capital expenditure continues to accelerate, with AI demand still significantly exceeding supply
AI summary card
Global cloud capital expenditure continues to accelerate, with AI demand still significantly exceeding supply
Morgan Stanley believes the four major US hyperscalers remain supply-constrained; CY27 consensus cloud capex has risen to approximately $1.2 trillion but could still be revised higher.
- CY27 consensus cloud capex is approximately $1.2 trillion, up about 29% year over year and significantly above the previous forecast of +14%.
- Morgan Stanley estimates aggregate cloud cash capex at $1.4 trillion, 17% above consensus.
- All four major US hyperscalers have indicated that demand exceeds available capacity, with external cloud customers and internal AI workloads jointly driving demand.
- Cloud revenue growth at Azure, GCP, and AWS continues to accelerate under supply constraints, reflecting higher utilization, AI service expansion, and broader infrastructure consumption.
- Funding sources are becoming more diversified, with operating cash flow, debt/equity financing, leases, ASICs, and infrastructure efficiency improvements collectively easing free cash flow pressure.
Report interpretation
Overview
This report tracks global cloud capex trends. Its core conclusion is that AI and cloud demand continues to exceed supply, driving continued upward revisions to cloud capex expectations for 2026-2027. The report notes that 2027 consensus cloud capex has risen to approximately $1.2 trillion, or about +29% year over year, but Morgan Stanley believes this expectation remains conservative as hyperscalers' confidence in demand, order visibility, AI monetization, and investment returns continues to strengthen.
Core views
The report's core views include: first, three of the four major US hyperscalers have raised CY26 capex or guidance, and all have indicated insufficient capacity; second, cloud revenue growth continues to accelerate despite capacity constraints, indicating that revenue is driven not only by new capacity but also by higher utilization and increasing value from AI services; third, management confidence in CY26-CY27 ROIC is strengthening, supporting a longer investment cycle; fourth, capex funding sources are becoming more diversified, easing but not eliminating free cash flow pressure; fifth, if qualitative commentary on cloud spending remains strong, CY27 capex expectations still have room for further upward revisions.
Analysis framework
The report uses a cloud capex tracking framework, combining commentary from the four major US hyperscalers' earnings reports, changes in CY26 guidance, CY27 market consensus expectations, Morgan Stanley estimates, cloud revenue growth, AI token processing volumes, capital intensity, and financing methods to assess whether capex expectations retain an upward bias.
Methodology notes
Comparison of consensus expectations with Morgan Stanley estimates
By comparing Consensus and MSe forecasts for the cash capex of the Top 14 cloud service providers, the report identifies whether market expectations are overly conservative.
Demand exceeding available supply
The report uses management commentary from AWS, Azure, GCP, Meta, and others regarding demand and capacity as key evidence for assessing the severity of supply tightness.
Returns on AI infrastructure investment
The report focuses on backlogs, multi-year commitments, pricing power, efficiency improvements, and AI monetization pathways to assess whether high capex is supported by adequate returns.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- GOOGL.USUS hyperscaler and operating entity of GCP
- Strengths
- CY26 capex guidance was raised to $195-$205B, GCP growth accelerated to +82% Y/Y, and the company benefits from enterprise AI products, AI infrastructure services, core GCP workloads, and TPU system sales.
- Weaknesses
- Demand exceeds internal supply, requiring the use of higher-cost third-party capacity as a transitional measure.
- Comparison
- Compared with other cloud providers, GCP has the highest growth rate and a notable contribution from TPU system sales.
- Risks
- Continued capex increases, high third-party capacity costs, and AI demand or cloud migration proceeding below expectations.
- META.USA major capex contributor to AI infrastructure and internal computing demand
- Strengths
- The CY26 capex range was narrowed to $130-$145B, with continued focus on maximizing capacity in 2026/2027. The company emphasizes that AI infrastructure can be monetized through consumer experiences, enterprise products, APIs, agentic solutions, and direct computing capacity.
- Weaknesses
- High investment intensity could pressure free cash flow.
- Comparison
- Compared with providers that directly monetize cloud service revenue, Meta relies more on internal AI applications and diversified monetization pathways to demonstrate investment returns.
- Risks
- Delayed AI monetization, continued industry-wide compute supply tightness, and execution risks related to debt financing and infrastructure partnerships.
- ORCL.USA beneficiary related to software infrastructure and cloud infrastructure
- Strengths
- It is positioned within the beneficiary chain of expanding cloud infrastructure demand and the early stage of enterprise AI adoption.
- Weaknesses
- The report does not provide detailed company-specific capex or cloud revenue growth breakdowns for Oracle.
- Comparison
- Compared with the four major US hyperscalers, the report's evidence is focused more heavily on AWS, Azure, GCP, and Meta.
- Risks
- Intensifying competition, the pace of capacity buildout, and changes in enterprise cloud migration and AI adoption.
- CRWV.USA company related to AI cloud and data center capex
- Strengths
- It is included among the cloud capex-related companies covered by Morgan Stanley and may benefit from AI compute demand and the data center investment cycle.
- Weaknesses
- The report does not provide company-level operating data or an updated price target.
- Comparison
- Compared with large hyperscalers, CoreWeave offers greater exposure to the AI compute supply side.
- Risks
- Customer concentration, financing needs, compute supply cycles, and volatility in GPU/data center costs.
- NBIS.USA company related to AI cloud and data center capex
- Strengths
- It is included among the cloud capex-related companies covered by Morgan Stanley and has exposure to AI infrastructure demand.
- Weaknesses
- The report does not provide detailed company-level financial forecasts for Nebius.
- Comparison
- Similar to CoreWeave, it has greater exposure to compute supply and AI cloud infrastructure rather than traditional hyperscale cloud platforms.
- Risks
- Execution risk, financing conditions, data center construction progress, and fluctuations in AI demand.
Key data
- CY27 consensus cloud capex~$1.2TEquivalent to approximately 30% year-over-year growth, about $170B above the level before the C2Q earnings reports.
- CY27 year-over-year cloud capex growth+29% Y/YThe previous forecast was approximately +14% Y/Y, representing a 15-percentage-point upward revision.
- Aggregate MSe cloud cash capex$1.4T17% above Consensus.
- Aggregate global cloud capex forecast for CY26-CY27$2.1TThe forecast at the beginning of the year was approximately $1.3T, implying an increase of about $800B.
- Capital intensityNear 35%Based on Consensus estimates, reaching a new historical high.
- Implied CY27 non-AI cloud capex growth+7% Y/YThe report considers this implied assumption overly conservative.
- Expected cloud revenue growth for the Top 4 US hyperscalers40-45%Expected to accelerate into this range over the next several quarters, representing the strongest level since 2020.
- Year-to-date change in Consensus CY27 capex+70%+ / ~$500BThe report believes further upward revisions remain possible if qualitative commentary stays strong.
Impact & implications
The report has positive implications for AI infrastructure, data centers, and the broader cloud capex value chain. Beneficiaries include hyperscalers, data center component suppliers, servers and storage, networking, semiconductors, ASICs, GPU-related supply chains, and software/cloud service ecosystems. Potential pressures include free cash flow, financing costs, depreciation and lease accounting treatment, and a potential decline in capex expectations if AI demand slows.
Risks
- AI demand growth falls below expectations, reversing the upward trend in cloud capex revisions.
- Supply chain bottlenecks, data center delivery issues, power constraints, and component shortages limit the conversion of capex into revenue.
- Rapid capex growth increases free cash flow and balance sheet pressure.
- Higher financing costs or a weaker capital market environment reduce cloud providers' ability to expand.
- Lease accounting, depreciation periods, and accounting reclassifications may affect financial statement comparability.
- If non-AI cloud capex growth remains low, overall investment intensity could face market scrutiny.
What to watch
- Subsequent changes in CY26/CY27 capex guidance from the four major US hyperscalers.
- Whether the gap between Consensus and MSe cloud capex forecasts continues to widen or narrows.
- Whether GCP, AWS, and Azure cloud revenue growth can remain in the 40-45% range.
- AI token processing volumes, inference demand, and the pace of enterprise AI adoption.
- Orders, delivery lead times, and revenue visibility for data center component suppliers.
- Changes in operating cash flow, debt financing, leasing strategies, and free cash flow pressure.
- Management commentary on ROIC, backlogs, multi-year commitments, and pricing power.