Cloud demand continues to exceed supply, and CY27 capex expectations still have room for upward revision

Overview

The report tracks C2Q26 results of U.S. hyperscale cloud providers and capex for the top 14 global cloud service providers. The latest data show that artificial intelligence inference, enterprise artificial intelligence applications and traditional cloud migration are jointly driving compute demand to continue exceeding supply. CY27 cloud capex consensus expectations have been raised significantly, but Morgan Stanley believes the implied non-artificial-intelligence cloud investment growth and overall growth path remain too conservative.

Core views

First, all four U.S. hyperscale cloud providers are supply-constrained, and new capacity can usually be monetized quickly. Second, cloud revenue growth no longer depends only on new capacity, but also benefits from higher utilization, high-value artificial intelligence services, and artificial-intelligence-driven consumption of storage, databases and networking. Third, backlog, multi-year contracts, pricing power, in-house chips and infrastructure efficiency improvements have strengthened management confidence in CY26-CY27 return on investment. Fourth, sources of capital are diversifying, as companies reduce free cash flow pressure through operating cash flow, debt and equity financing, leases and third-party capacity. Fifth, given multiple consecutive quarters of upward capex forecast revisions, CY27 consensus expectations still have further upside risk.

Analysis framework

The report combines quarterly results from the four major U.S. hyperscale cloud providers, management capex guidance, consensus expectations for the top 14 global cloud service providers and Morgan Stanley forecasts. It compares changes in cloud revenue, cash capex, non-artificial-intelligence capex and capital intensity, and validates the durability of the spending cycle through supply-demand constraints, order commitments, financing capacity and return-on-investment commentary.

Methodology notes

• Data trackingGlobal cloud capex tracker

  Summarize capex forecasts and year-over-year growth rates for major cloud service providers by quarter

  By comparing the direction and magnitude of consecutive forecast revisions, assess whether the cloud infrastructure investment cycle is accelerating and whether consensus expectations fully reflect management guidance.

• Expectations gap analysisConsensus expectations versus Morgan Stanley forecast comparison

  Compare market consensus expectations with MSe capex estimates

  CY27 consensus expectations are about $1.2 trillion, while MSe is about $1.4 trillion; the 17% gap forms the core basis for further upward revisions.

• Supply-demand analysisCloud compute supply-demand gap

  Measure demand strength using capacity constraints, reservations and the speed of monetization of new capacity

  All four major U.S. hyperscale cloud providers said demand exceeds available capacity, and some CY27 capacity has already been reserved, indicating no clear near-term cooling in demand.

• Capital efficiency analysisCapital intensity and return-on-investment assessment

  Assess capex as a percentage of revenue, contract duration, pricing power and asset payback periods

  Capital intensity is near historical highs, but multi-year contracts, efficiency improvements, pricing power and some server investments reaching breakeven within three years support management confidence in long-term returns.

Asset mapping & comparison

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

• GOOGL / Alphabet Inc.

  One of the four major U.S. hyperscale cloud providers and a direct vehicle for GCP capex and cloud revenue acceleration

  Strengths

  GCP grew 82% year over year, driven jointly by artificial intelligence products, infrastructure services, core cloud workloads and TPU system sales; long-term demand, backlog and returns on third-party capacity support investment confidence.

  Weaknesses

  Internal capacity still cannot meet demand, requiring the use of higher-cost third-party capacity and bearing high capital intensity and financing needs.

  Comparison

  CY26 capex guidance was raised to $195 billion-$205 billion, reflecting accelerated capacity delivery.

  Risks

  Capacity expansion delays, artificial intelligence demand or monetization weaker than expected, rising third-party capacity costs and pressure on free cash flow.

• META / Meta Platforms Inc.

  A major internal demand source for artificial intelligence infrastructure and a hyperscale capex entity

  Strengths

  It can monetize artificial intelligence infrastructure through consumer products, enterprise services, application programming interfaces, agent solutions and direct compute, and has strong cash generation capabilities.

  Weaknesses

  Capex scale is expanding rapidly, and returns depend relatively heavily on internal product monetization and infrastructure deployment efficiency.

  Comparison

  The CY26 capex range was narrowed and moved up to $130 billion-$145 billion, while still emphasizing long-term industry compute tightness.

  Risks

  Artificial intelligence product monetization falling short of expectations, infrastructure partnership execution risk, increased debt financing and free cash flow pressure.

• CRWV / COREWEAVE INC

  Artificial intelligence cloud and compute capacity provider benefiting from tight hyperscale compute supply and third-party capacity demand

  Strengths

  Industry compute shortages, long-term capacity commitments and rapid growth in artificial intelligence inference demand help improve capacity utilization and order visibility.

  Weaknesses

  The business is capital intensive, and expansion depends on continued financing, infrastructure delivery and realization of customer demand.

  Comparison

  Compared with large integrated cloud platforms, it has higher direct sensitivity to the artificial intelligence compute cycle.

  Risks

  Financing costs, customer concentration, capacity delivery, equipment depreciation and risk of a reversal in the supply-demand cycle.

• NBIS / NEBIUS GROUP NV

  Artificial intelligence infrastructure and cloud compute provider benefiting from external compute demand exceeding the internal supply of large cloud providers

  Strengths

  Enterprise artificial intelligence adoption is still in the early stage, and third-party capacity can supplement large cloud providers during expansion.

  Weaknesses

  Scale, financial strength and ecosystem are weaker than those of large U.S. integrated cloud platforms.

  Comparison

  Compared with hyperscale cloud providers, it has greater capex cycle elasticity, but also higher operating and financing risks.

  Risks

  Demand realization, financing, project construction, equipment supply and intensifying competition risks.

• ORCL / ORACLE CORP

  Cloud infrastructure and enterprise workload provider that may benefit from enterprise artificial intelligence adoption and continued cloud migration

  Strengths

  Its enterprise customer base and cloud infrastructure exposure can capture artificial-intelligence-driven database, storage, networking and compute demand.

  Weaknesses

  The report does not provide separate capex, supply-demand or return-on-investment analysis for ORCL.

  Comparison

  Compared with AWS, Azure and GCP, this report provides less quantitative tracking evidence for ORCL.

  Risks

  Capital investment recovery, competition, capacity construction and enterprise artificial intelligence adoption slower than expected.

• Data center equipment and component supply chain

  Upstream beneficiary group of global cloud capex growth

  Strengths

  Cloud providers continue to expand capacity, demand visibility is extending, and CY27 capex expectations may still be revised upward.

  Weaknesses

  Cloud capex revenue exposure, bargaining power and capacity constraints vary significantly across suppliers.

  Comparison

  The report says its coverage spans more than 50 technology companies in the United States, Europe and Asia, but the current materials do not provide complete company-level exposure data.

  Risks

  Customer concentration, order delays, component cost inflation, capacity bottlenecks and a pullback in the capex cycle.

Key data

• CY27 cloud capex consensus expectationsAbout $1.2 trillion, up about 29%-30% year over yearYear-over-year growth is 15 percentage points higher than the previous forecast of 14%, and the amount increased by about $170 billion versus before C2Q26 results.
• CY27 Morgan Stanley cloud cash capex forecastAbout $1.4 trillion17% above consensus; corresponding year-over-year growth exceeds 50%, about 22 percentage points above consensus.
• CY26-CY27 global cloud capexAbout $2.1 trillionAbout $800 billion higher than the forecast at the beginning of the year, equivalent to about 33% of combined revenue over the same period.
• Capital intensityClose to 35%Based on consensus expectations, this reaches a record high and is about four times the average capex intensity from 2014 to 2023.
• GOOGL CY26 capex guidance$195 billion-$205 billionPreviously $180 billion-$190 billion; the upward revision stems from accelerating capacity delivery to meet demand that exceeds supply.
• AMZN CY26 cash capexAbout $220 billionPreviously about $200 billion. The incremental portion is affected by memory cost inflation, but the company still expects capacity to be insufficient to meet demand.
• META CY26 capex guidance$130 billion-$145 billionPreviously $125 billion-$145 billion; the company remains demand-constrained and focused on maximizing CY26-CY27 capacity.
• MSFT CY26 capexAbout $190 billion on the underlying basis and about $175 billion on the reported basisUnderlying spending is unchanged; the decline on the reported basis mainly comes from lease accounting classification adjustments.
• Major cloud platform revenue growthAzure 43%, GCP 82%, AWS 37%All are year-over-year growth rates; AWS recorded its fastest growth in 18 quarters.
• Expected revenue growth for the four major U.S. cloud providers over the next several quarters40%-45%Expected to be the strongest level since 2020.

Impact & implications

Continued upward revisions to cloud capex benefit the data center supply chain, including artificial intelligence compute, servers, networking, optical communications, storage, power and cooling, and also support cloud platforms with capacity, customer contracts and financing capabilities in expanding revenue. At the same time, capital intensity is at a historical high, and investors will pay closer attention to whether artificial intelligence monetization, asset utilization, free cash flow and return on investment can keep pace with spending growth. The industry view listed in the report remains cautious, meaning strong capex does not automatically imply that all related stocks have equal return potential.

Risks

• Artificial intelligence demand or monetization speed may be lower than expected, causing cloud providers to cut or delay capex.
• Supply constraints in data centers, power, networking, memory and advanced computing equipment may delay capacity going online.
• Capital intensity is close to historical highs, and depreciation, leases and financing costs may continue to pressure free cash flow.
• In-house chips, infrastructure efficiency and application-layer optimization may fail to generate expected returns, potentially weakening ROIC.
• Cost inflation in components such as memory may push up project budgets and compress cloud service margins.
• Changes in the classification of operating leases and finance leases may reduce the comparability of capex data across companies and periods.
• If demand growth slows after capacity supply improves, utilization may decline or temporary oversupply may emerge.
• The research institution has shareholdings, investment banking business and other service relationships with multiple covered companies; investors should assess potential conflicts of interest together with the disclosures.

What to watch

• Whether CY27 cloud capex consensus expectations can continue moving toward MSe’s forecast of about $1.4 trillion.
• The speed of new capacity delivery by the four major U.S. hyperscale cloud providers and the timing of easing supply constraints.
• Whether Azure, GCP and AWS can maintain or improve current cloud revenue growth rates.
• Monthly artificial intelligence token processing volume, inference demand and enterprise artificial intelligence adoption rates.
• Backlog, multi-year capacity contracts, pricing power and utilization of new capacity.
• The extent to which artificial intelligence services drive storage, databases, networking and traditional cloud workloads.
• The impact of capex growth on free cash flow, depreciation, debt financing and balance sheets.
• The actual effectiveness of cost-reduction measures such as in-house ASICs, longer server lives, leases and third-party capacity.
• Order visibility, capacity expansion and cloud capex revenue exposure of data center component suppliers.