UBS Maintains Buy Rating on Amazon, Unpacking AI Backlog as Growth Catalyst

Overview

This report outlines UBS’s latest revision to its investment thesis on Amazon (AMZN.US), centered on a refined revenue forecasting model for its cloud computing division (AWS). Previously, UBS had overestimated 1Q26 growth and thus re-examined its methodology. By decomposing AWS’s performance obligations (PO/backlog) into three components—'core non-AI', 'Bedrock', and 'large contracts with Anthropic and OpenAI'—UBS arrives at a more optimistic and well-supported growth outlook. The report maintains a 'Buy' rating and a $333 target price, asserting that the market has yet to fully price in AWS’s AI-driven acceleration potential—particularly Bedrock’s explosive growth and the massive, long-term contracts signed with top-tier AI labs.

Core views

UBS notes that its prior model relied solely on aggregate PO and average contract duration to derive AWS revenue, resulting in forecast inaccuracies. The new analytical approach structurally disaggregates backlog: Demand-side & AI-driven growth: According to CEO Andy Jassy, AI-driven run-rate revenue reached $15 billion as of 1Q26. UBS estimates Bedrock contributed ~$1.7 billion of this. More critically, Bedrock’s backlog stood at ~$35 billion as of 1Q26, up ~340% YoY. With OpenAI models now available on the Bedrock platform and deepening partnerships with other AI labs, Bedrock’s backlog is projected to reach ~$96 billion by end-2026. Large contract overlay: UBS overlays recently announced large-scale contracts with OpenAI—including an initial $38 billion commitment and a subsequent $100 billion commitment—and Anthropic (~$100 billion)—into its model. These long-term contracts (with durations set at 7–8 years) meaningfully extend the weighted-average contract term and enhance forward revenue visibility. UBS assumes AWS will add ~$350 billion in new backlog in 2026—a scenario made substantially more credible by the public announcement of these mega-deals. Revenue conversion & share: Under revised conversion assumptions, UBS forecasts AI compute revenue’s share of total AWS revenue rising from ~10% in 1Q26 to 26% by end-2026 and 30% in 2027. Consequently, its 2026 AWS revenue forecast remains at $175.9 billion (+36% YoY), well above consensus of $166.6 billion (+29% YoY). This implies a 2027 operating profit estimate ~40% above consensus. Other business highlights: Beyond AWS, the report highlights e-commerce’s margin expansion potential driven by improved unit economics (sales growth outpacing cost growth); same-day delivery expansion and grocery investments potentially accelerating GMV growth and market share gains; and Prime Video’s advertising business—bolstered by strategic partnerships and live sports content—as having higher-margin revenue potential during global scale-up.

Analysis framework

UBS demonstrates a hybrid bottom-up/top-down 'backlog decomposition' methodology in this report. First, the firm moves away from treating AWS as a monolithic black box and instead constructs a 'book-to-bill' waterfall model using disclosed performance obligation (PO) data from financial statements and average contract duration. Second, it introduces structural segmentation. Given the distinct characteristics of AI-related business, UBS splits PO into 'core non-AI', 'Bedrock', and 'large AI lab contracts'. It reverse-engineers Bedrock’s specific contribution using the known AI run-rate revenue and triangulates the size of Anthropic’s early undisclosed contract (~$20 billion) by identifying anomalies in PO growth. Third, it adjusts conversion parameters. Different contract types are assigned distinct revenue recognition timelines (e.g., OpenAI’s early contract over 7 years, follow-on contracts over 8 years), enabling more accurate modeling of how long-term contracts contribute to current and future revenue. While this increases quarterly forecast volatility, it improves annual forecast accuracy—especially in capturing the timing of AI capital expenditure conversion into realized revenue.

Methodology notes

• Industry/sector analysis frameworkVolume-price decomposition

  Decomposing total AWS backlog into core non-AI, Bedrock, and specific large-customer contracts

  By breaking down aggregated business metrics into sub-components driven by distinct growth levers, analysts can more clearly identify sources of growth. Here, this segmentation isolates the true growth trajectory of AI-related business—especially Bedrock—avoiding its signal being smoothed out by traditional cloud business growth.

• Valuation methodologyFCFF/FCFE Free Cash Flow

  Free cash flow (FCF)-based valuation method

  Given management’s longstanding emphasis on free cash flow generation, UBS applies a price-to-free-cash-flow (P/FCF) multiple rather than conventional P/E. This better aligns with Amazon’s capital-intensive, high-reinvestment profile and yields a more accurate reflection of intrinsic value.

• Corporate fundamentals & financial frameworkThree-statement reconciliation

  Reconciling performance obligations (PO) with revenue recognition

  By analyzing the relationship between 'performance obligations' (i.e., unfulfilled contractual commitments) reported on the balance sheet or in footnotes and 'revenue' on the income statement—combined with contract duration—one can proactively forecast future revenue. This is a widely used technique in SaaS and cloud services industries.

Asset mapping & comparison

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

• Amazon.com (AMZN.US)

  Direct beneficiary and core vehicle for AWS cloud and AI infrastructure

  Strengths

  Robust cloud infrastructure; exclusive or deep partnerships with top-tier AI firms including OpenAI and Anthropic; rapid growth of the Bedrock platform; improving e-commerce margins

  Weaknesses

  Massive upfront capital expenditures, potentially pressuring short-term free cash flow

  Comparison

  Compared to other cloud providers, offers richer AI application ecosystems and retail synergies

  Risks

  Intensifying competition, deteriorating consumer sentiment, higher-than-expected capital expenditures

Key data

• Bedrock Backlog (as of end-1Q26)~$35 billionUp ~340% YoY
• 2026 AWS Revenue Forecast$175.9 billion+36% YoY, above consensus of $166.6 billion (+29% YoY)
• AI Revenue Share of AWS (2026E exit)26%Projected to reach 30% in 2027
• 2027 Operating Profit Forecast Differential~40% above consensusDriven by higher AWS revenue expectations
• Target Price$333Based on 30x 2027–2028 free cash flow estimate
• 2026 New Backlog Addition Forecast~$350 billionPrimarily driven by OpenAI and Anthropic mega-deals

Impact & implications

UBS believes Amazon’s 'coiled spring' investment theme will continue releasing momentum with each quarterly earnings release. The market currently underappreciates AWS’s AI growth potential—particularly the compounding effect of Bedrock and high-value AI lab contracts. As investors gradually recognize that 2027 operating profit could exceed current consensus by ~40%, further upside remains. Additionally, the stock currently trades at an 18x 2027 estimated P/E—UBS contends such a high-quality asset should not trade at a multiple below the market average.

Risks

• Rising competitive intensity from online e-commerce and offline retail rivals
• Deteriorating consumer sentiment, potentially impacting transaction volume and velocity
• Higher-than-anticipated capital intensity for the e-commerce platform or AWS, thereby weighing on free cash flow forecasts

What to watch

• Quarterly changes in AWS new backlog, particularly the split between non-AI core business and AI-related business
• Bedrock platform revenue conversion rate and progress on revenue recognition from OpenAI/Anthropic contracts
• Improvement in e-commerce unit economics
• Global scaling progress of Prime Video advertising business