Cloud Giants' Backlog Reaches $2 Trillion, 2026 Capex Raised to $770 Billion

Overview

This report summarizes Jefferies' joint analyst conference call minutes regarding Q1 2026 hyperscalers, data centers, and power sectors. Key conclusions indicate AI infrastructure construction is still in early stages, with demand far exceeding supply. Order backlog totals reached approx. $2 trillion for major cloud vendors (Google, Amazon, Microsoft, Meta, Oracle), justifying continued massive capital expenditure. The institution raised 2026 combined capex expectations to $770 billion and supports the 'Compute equals Revenue' logic, maintaining buy ratings on relevant leaders.

Core views

Demand side shows explosive growth, order backlog (Backlog) becomes key indicator to measure future revenue. First quarter main cloud vendor total order backlog increased quarterly by $376 billion to about $2 trillion. Among them, Google contributed the most, single quarter increase of $225 billion; Amazon increased $120 billion (if including Anthropic transaction reaches $220 billion). This huge backlog volume indicates that current capital expenditure is not only not excessive, but far lagging behind actual needs. Performance among cloud vendors diverged, with Google Cloud (GCP) becoming the biggest highlight. GCP first quarter revenue growth as high as 70%, market share significantly improved, and cloud business margins nearly doubled to about 33% over past four quarters. Amazon AWS growth accelerated 4 percentage points to 28%, operating margin expanded to 38%, reflecting its AI, fulfillment, and utilization efficiency flywheel effect. Microsoft Azure grew 39% (constant currency), met guidance upper limit, expected second half to have gentle acceleration, its overall operating margin expanded to 46.3%. Capital expenditure (Capex) ramp-up continues, and scale is surprising. Based on Q1 reports, Jefferies raised 2026 top 5 cloud vendors (Big 3 + Meta & Oracle) capital expenditure expectation by $52 billion (increase 8%) to $770 billion, year-over-year growth rate maintained at high position of 74%. Specifically, Microsoft raised 2026 guidance to $190 billion (partially affected by component price increases); Amazon maintained $200 billion guidance, emphasizing cost advantages brought by self-developed Trainium chips; Meta raised guidance to $125-145 billion; Google raised guidance to $180-190 billion, and announced 2027 capital expenditure will "significantly increase". Revenue growth of AI model vendors is staggering, further confirming urgency of compute demand. Anthropic added annual recurring revenue (ARR) exceeding $31 billion in first four months of this year, accounting for more than half of ARR new additions in 63 public software companies covered by Jefferies (except Microsoft) in full 2025. OpenAI first quarter revenue and usage annualized growth reached 80x, expected this year compute cost expenditure reaches $50 billion. These unlisted giants' wild expansion directly translates to rigid demand for cloud vendor infrastructure.

Analysis framework

Institution adopts a method combining 'top-down' industry sentiment tracking with 'bottom-up' company fundamental verification. First, through aggregating head cloud vendors' order backlogs (Backlog) and capital expenditure guidance, judging industry overall supply-demand gap and investment cycle position. Second, dissecting each company's revenue growth, market share changes and margin performance, assessing evolution of competitive landscape (such as GCP rise). Finally, through tracking AI native application vendors' (such as Anthropic, OpenAI) compute procurement behavior and income explosiveness, cross-verifying authenticity and sustainability of cloud vendor upstream demand. This analytical framework emphasizes direct transmission logic of 'Compute = Revenue', and constraining effect of supply chain bottlenecks (such as Power, GPU) on capital expenditure scale.

Methodology notes

• Industry / Industrial Analysis FrameworkSupply-demand framework

  Supply-Demand Framework

  Report compares cloud vendor order backlogs (demand proxy variable) with capital expenditure (supply proxy variable), pointing out current AI computing market is in severe shortage of supply and demand state, demand growth speed far faster than supply capacity construction speed.

• Company Fundamentals & Financial FrameworkFree cash flow analysis

  Capital Expenditure & Free Cash Flow

  Although significant capital expenditure increase may suppress short-term free cash flow, report emphasizes high order backlog predicts future strong cash inflow, and current margins still expanding, indicating investment efficiency not diluted due to scale expansion.

• Valuation MethodEV/EBITDA valuation

  EV/EBITDA Multiple Comparison

  Report uses EV/EBITDA multiple comparison for Amazon (13x) and Google (19x), believes Amazon has valuation discount, providing better risk-return ratio, while Google although valuation higher possesses stronger AI flywheel effect.

Asset mapping & comparison

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

• Alphabet (GOOGL.US)

  Benefit: GCP Growth Leads, Backlog Surges, Significant AI Flywheel Effect

  Strengths

  GCP Market Share Fastest Increase, Cloud Margins Doubled, TPU Ecosystem Mature

  Weaknesses

  Valuation Relatively Higher (19x EV/EBITDA)

  Comparison

  Compared to peers, strongest growth momentum, but valuation premium requires performance continual realization

  Risks

  Capital Expenditure Execution Risk, Intensified Competition

• Amazon (AMZN.US)

  Benefit: AWS Growth Accelerated, Self-developed Chips Bring Cost Advantage, Valuation Attractive

  Strengths

  Trainium Chips Lower Compute Costs, High Operational Efficiency, Valuation Discount (13x EV/EBITDA)

  Weaknesses

  Market Share Slight Decline

  Comparison

  Best Risk-Return Ratio, Suitable for Value Investors

  Risks

  Retail Business Fluctuations, Cloud Competition Pressure

• Microsoft (MSFT.US)

  Benefit: Azure Steady Growth, Deep Enterprise Customer Base

  Strengths

  M365 & Azure Synergy, Powerful Enterprise Channels, High Margins

  Weaknesses

  Booking Growth Weaker Relative to Peers

  Comparison

  Highest Stability, but Burst Potential Slightly Inferior to Google

  Risks

  Component Price Increases Impact Capital Expenditure Efficiency

• CoreWeave (CRWV.US)

  Benefit: As Specialized AI Cloud Service Provider, Directly Benefits from Compute Shortage

  Strengths

  Focus on GPU Compute Leasing, Close Relationship with NVIDIA

  Weaknesses

  Scale Smaller Than Hyperscale Cloud Vendors

  Comparison

  Greater Elasticity, but Risk Resistance Weaker Than Giants

  Risks

  Dependence on Single Hardware Vendor, Customer Concentration Risk

Key data

• Total Cloud Backlog~$2 TrillionIncreased $376B QoQ
• Expected 2026 Capital Expenditure$770 BillionRaised $52B, 74% YoY Growth
• Google GCP Revenue Growth70%Largest Market Share Gain, Accelerated 16 Percentage Points
• Amazon AWS Revenue Growth28%Accelerated 4 Percentage Points, Operating Margin ~38%
• Microsoft Azure Revenue Growth39%Constant Currency, Met Guidance Upper Limit
• Anthropic YTD New ARR~$31 BillionOnly Used 4 Months, Over Half of Full Industry Last Year New Total

Impact & implications

Report believes AI investment cycle remains complete and in early stage, demand exceeds supply fundamental unchanged. For cloud vendors, huge order backlog provides high visibility for future revenue growth, despite high capital expenditure, profit margin expansion proves scale effect and operational efficiency improvement. For Power and Data Center REITs sector, large projects approved at 9-10GW scale (such as Wyoming and Utah projects) indicates infrastructure bottlenecks alleviating via policy support and new capacity construction, beneficial to related power utilities and digital infrastructure companies. Investors should focus on those able to efficiently convert capital expenditure into revenue, and occupy market share advantage in AI wave leader enterprises.

Risks

• Component price increases causing capital expenditure exceed expectations
• AI demand growth slowdown, causing backlog conversion below expectations
• Power supply bottlenecks limiting data center expansion speed
• Potential restrictions of regulatory policies on AI development

What to watch

• Whether Azure and AWS growth accelerates further in second half of 2026
• Specific scale of Google 2027 capital expenditure 'Significant Increase'
• Progress on compute procurement contracts for AI-native vendors like Anthropic and OpenAI
• Construction progress and power access status of large data center projects (e.g., Wyoming, Utah)