Surging AI CAPEX Drives Divergence in Hyperscale Cloud Vendor Performance in Q1

Overview

This report provides an in-depth analysis of the performance of leading global hyperscale cloud service providers (hyperscalers), including Microsoft, Amazon, Google, Oracle, Alibaba, and the emerging player CoreWeave, during the first quarter of 2026. It focuses on the relationship between their substantial capital expenditures (CAPEX) and revenue growth amid the surge in generative AI demand, highlighting investor concerns about return on investment (ROI). Despite recent acceleration in cloud revenue growth, it remains insufficient to keep pace with escalating capital investments. The competitive landscape has shifted from simple GPU availability to physical data center capacity and power supply constraints, reshaping each vendor's position in the AI cloud sector.

Core views

Capital Expenditures and Financing Pressures: In the first quarter of 2026, the top four hyperscale cloud vendors collectively incurred cash CAPEX of approximately $130 billion. Bernstein forecasts that the total cash CAPEX for these four vendors will reach $623 billion in 2026, with Meta bringing the total close to $760 billion when included. Given that operational cash flow growth lags behind CAPEX, most vendors may need to resort to debt or equity financing in the coming years to fund infrastructure development, as evidenced by recent fundraising activities from Amazon, Google, and Oracle. Microsoft (MSFT): Azure demonstrated robust performance even under constrained supply conditions, posting constant currency revenue growth of 39%, surpassing guidance. Its AI annual recurring revenue (ARR) exceeded $37 billion, with M365 Copilot paid seats reaching 20 million and growing rapidly. Due to internal capacity limitations, Microsoft even rents AWS capacity to meet GitHub demands. Commercial RPO stood at $627 billion, indicating stable long-term demand. Oracle (ORCL): Transitioning into a primary provider for AI training and inference, OCI reported a year-over-year revenue increase of 93% to $58 billion. RPO skyrocketed to $638 billion, with approximately $300 billion attributable to commitments from OpenAI. Although short-term margins face pressure, the company exhibits significant potential for long-term profitability and free cash flow growth. Plans include adding 1GW of data center capacity next quarter and another 10GW over the following three years. Amazon (AMZN): AWS revenue grew by 28%, with an annualized run rate of approximately $150 billion, marking a 400 basis point acceleration in growth velocity. Custom chip offerings emerged as a standout feature, with Trainium revenue commitments totaling $225 billion and a chip-specific ARR exceeding $200 billion. Data gravity effects encourage enterprises to expand AI workloads within the AWS ecosystem. Operating margins have held steady around 38%, reflecting strong cost control capabilities. Google (GOOGL): The cloud division performed notably well, achieving revenue growth of 63% to exceed $20 billion, outpacing net additions from AWS. Backlogs doubled quarter-on-quarter to $462 billion, fueled by enterprise AI demand and new revenue streams from third-party TPU hardware sales. Operating margins significantly improved to 33%, though investors should monitor rising depreciation costs and short-term integration expenses related to the Wiz acquisition. Alibaba (BABA): Alibaba Cloud saw revenue growth of 38.2%, with external revenue增速 accelerating to approximately 40%. AI-related products experienced triple-digit growth for 11 consecutive quarters, accounting for 30% of external cloud revenue. Management projects future external revenue增速 will exceed 40% over the next few quarters, with adjusted EBITA margins targeting low double digits. CoreWeave (CRWV): As an emerging player, CoreWeave recorded revenue growth of 112% to $21 billion, with backlogs amounting to $994 billion. However, its business model encounters severe challenges, with a meager 0.5% after-tax operating margin and CAPEX projections ranging from $310-350 billion. The report questions its lack of structural advantages and long-term sustainability, assigning an underperform rating.

Analysis framework

The report adopts a horizontal comparative analysis framework, positioning traditional hyperscale cloud vendors alongside emerging players to dissect the 'CAPEX-revenue growth-backlog-cash flow' transmission logic. Emphasis is placed on the shift in competitive factors from 'GPU availability' to 'physical data center capacity/power,' using Remaining Performance Obligations (RPO) and contract durations to assess revenue visibility. Additionally, the analysis evaluates the economic viability of custom chips (such as Trainium, TPU, Graviton) to determine each vendor's long-term cost advantage and moat in AI infrastructure.

Methodology notes

• Industry/Industrial Analysis FrameworkSupply-demand framework

  Core contradiction in the cloud infrastructure industry shifting from GPU chip supply shortages to physical data center capacity and power supply constraints

  As AI compute demand surges, bottlenecks have evolved from initial GPU shortages to limitations in data center physical space and power availability, determining which companies can roll out capacity faster and capture customer contracts.

• Corporate Fundamental and Financial FrameworkFree cash flow analysis

  Gap analysis between CAPEX and operating cash flow

  By comparing massive CAPEX projections with the pace of operating cash flow (CFO) growth, the assessment determines whether vendors need external financing and how such intensive investment impacts short-term financial health and long-term returns.

• Competition and Strategy FrameworkMoat / competitive advantage

  Data Gravity and Economic Efficiency of Custom Chips

  The report highlights the 'data gravity' effect, where enterprises prefer running AI applications where their data resides—a moat for leaders like AWS—and emphasizes that self-developed chips (e.g., Trainium, TPU) improve cost efficiency, reducing AI inference costs and boosting profitability.

Asset mapping & comparison

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

• Microsoft (MSFT.US)

  Beneficiary: Strong growth in AI ARR supported by diverse application base enhancing capacity utilization

  Strengths

  Stable Azure growth, accelerated Copilot seat growth, balanced mix of in-house chips and third-party contracts

  Weaknesses

  Capacity constraints requiring renting competitor capacity

  Comparison

  Comparable to Google in terms of incremental AI cloud dollars, superior to pure infrastructure providers

  Risks

  Long CAPEX payback cycles, risk of AI bubble burst

• Oracle (ORCL.US)

  Beneficiary: Rapid OCI growth backed by high-value contracts providing certainty

  Strengths

  Massive RPO scale, execution capability, expanding financing channels

  Weaknesses

  Short-term margin pressure, tight cash flows

  Comparison

  Outpaces peers in growth, though smaller base creates reliance on key clients

  Risks

  Financing failure risk, volatility in strategic partnerships

• Amazon (AMZN.US)

  Beneficiary: Leveraging data gravity and cost advantages from custom chips

  Strengths

  Leadership in infrastructure, competitive edge with Trainium/Graviton chips, stability from non-AI businesses

  Weaknesses

  Massive CAPEX burden, short-term free cash flow stress

  Comparison

  Leads in speed of acquiring computing capacity, resilient profit margins

  Risks

  Intensifying competition driving pricing pressures, suboptimal CAPEX efficiency

• Google (GOOGL.US)

  Beneficiary: New revenue stream from TPU hardware sales and robust AI workloads

  Strengths

  Backlog doubling, significant gross margin improvement, technological leadership

  Weaknesses

  Increased depreciation, short-term integration costs from Wiz acquisition

  Comparison

  Net revenue additions surpassed AWS, strongest growth trajectory

  Risks

  Deferred revenue recognition for hardware sales, intense competition

• Alibaba (BABA.US)

  Beneficiary: Accelerating growth in AI-driven products and external revenue

  Strengths

  Consistent triple-digit growth in AI-related services, improving margins

  Weaknesses

  Geopolitical risks, fierce domestic competition

  Comparison

  Maintains leadership in China, swift AI transformation

  Risks

  Economic slowdown, regulatory changes

• CoreWeave (CRWV.US)

  Damaged: Structural disadvantages make business model questionable

  Strengths

  High short-term demand, fast growth rates

  Weaknesses

  Extremely low margins, massive CAPEX needs, leasing-dependent operations

  Comparison

  Significantly smaller than hyperscale rivals, disadvantaged long-term

  Risks

  Financing difficulties, squeezing by dominant players, unresolved profitability issues

Key data

• Top Four Cloud Vendors' Q1 Cash CAPEX~$130 BillionActual value for Q1 2026; projected full-year 2026 total is $623 billion
• Microsoft's AI ARR> $37 BillionIncludes Azure AI and SaaS Copilot revenue, up 123% year-over-year
• Oracle OCI Revenue Growth Rate93%Year-over-year growth for fiscal Q4 2026; RPO stands at $638 Billion
• AWS Trainium Revenue Commitments~$225 BillionStrong custom chip segment; chip-specific ARR exceeds $200 Billion
• Google Cloud Backlog~$462 BillionDoubled quarter-over-quarter; over 50% expected to convert to revenue within 24 months
• CoreWeave's Operating Margin0.5%After-tax operating profit of $21 Million in Q1 2026; scalability and profitability concerns persist

Impact & implications

The report views the hyperscale cloud market as the 'shovels' business supporting the generative AI wave but cautions against risks associated with excessive CAPEX expansion affecting ROI. Microsoft and Amazon demonstrate resilience in the AI bubble scenario thanks to diversified customer bases and advantages in self-developed chips; Oracle leverages large client relationships like OpenAI for rapid growth albeit facing short-term margin pressures; Google Cloud shows strong momentum although hardware sales' deferred revenue recognition could impact short-term reporting. For CoreWeave, the report signals lack of long-term structural advantages likely putting it at a disadvantage against major players. Overall, AI is reshaping cloud IT budgets, and as AI workloads transition from training to inference, demand for CPUs and non-GPU resources will also grow simultaneously.

Risks

• Underperformance in AI CAPEX ROI dampening investor confidence
• Physical data center capacity and power supply bottlenecks faced by hyperscale vendors
• Unsustainable business models among emerging players like CoreWeave creating dual pressures on financing and profitability
• Risk of AI bubble bursting impacting long-term revenue visibility for cloud providers
• Geopolitical factors potentially disrupting cross-border cloud operations

What to watch

• Whether cloud revenue growth can sustainably accelerate to match CAPEX increases
• Progress in hyperscalers' financing activities and implications for ownership structures
• Shift in AI workloads from training to inference altering CPU and non-GPU resource requirements
• Improvement in profitability and market share movements for emerging vendors like CoreWeave
• Adoption rates and economic benefits derived from various vendors’ custom chips (Trainium, TPU, Graviton, etc.)