AWS is the clear leader in cloud infrastructure with 29-30% market share, generating $45.6B in operating income at 35% margins. The business faces a critical juncture: Amazon is investing $200B in 2026 capex -- the most aggressive infrastructure build in corporate history -- betting that AI workloads will drive a multi-year growth reacceleration. The $244B backlog provides demand visibility, but the capex-to-FCF conversion is the key uncertainty.
| AWS | 29-30% market share, $128.7B FY2025 revenue, 24% Q4 growth |
| Azure | 20-21% market share, ~39% growth, gaining share fastest |
| Google Cloud | 13-14% market share, $58.7B revenue, 48% Q4 growth |
Capex Risk
Amazon's FCF collapsed from $32.9B (2024) to $7.7B (2025). With $200B guided for 2026, analysts estimate negative FCF of $17-28B. The depreciation wave from this cycle will pressure margins through 2027-2028.
Does the $200B capex cycle yield returns via AI workload dominance, or create overcapacity?
AWS experienced a dramatic growth reacceleration through 2024-2025, recovering from a 12% trough to 24% Q4 growth. The recovery is driven by AI workloads reaching critical mass, a $244B order backlog, and wins with major AI companies (Anthropic, OpenAI, Apple) for custom Trainium silicon. The question is whether 20%+ growth can be sustained on a $130B+ revenue base.
Growth Driver
GenAI cloud services grew 140-180% in 2025. AWS's reacceleration from 12% to 24% is predominantly AI-driven, not traditional cloud migration.
AWS margins remain industry-leading at 35% despite the most aggressive infrastructure build in corporate history. The critical question is what happens in 2027-2028 when the depreciation wave from $200B in 2026 capex hits the income statement. Bulls argue AI workload density and Trainium cost advantages will drive margins toward 38%+. Bears point to the $200B capex creating a depreciation tsunami that compresses margins to 28-30%.
FCF Under Siege
Free cash flow collapsed 77% from $32.9B to $7.7B in FY2025. With $200B capex guided for 2026, analysts estimate negative FCF of $17-28B. This is peak investment phase -- FCF should recover by 2028-2029 as capex normalizes.
| AWS | 29% share, $128.7B revenue, 24% Q4 growth. Broadest services, custom silicon (Trainium), no frontier model. |
| Azure | 20% share, ~39% growth. OpenAI exclusive partner, Copilot 90%+ of Fortune 500, strongest enterprise AI distribution. |
| Google Cloud | 13% share, $58.7B revenue, 48% Q4 growth. Gemini/Vertex AI, 10-20% cheaper, fastest growth rate. |
AWS maintains clear market share leadership but faces a strategic question: in an AI-first world, does infrastructure scale or model ownership matter more? Azure has OpenAI, Google has Gemini -- AWS has Bedrock (multi-model) and Trainium (custom silicon). AWS's bet is that enterprises will choose infrastructure flexibility and cost advantages over model lock-in.
AWS's custom silicon strategy may be its most important competitive move in a decade. By winning Anthropic, OpenAI, and Apple as Trainium customers, Amazon has validated that custom chips can compete with NVIDIA at scale. The 30-40% cost advantage, combined with Trainium4's planned NVLink support for hybrid clusters, creates a path to decouple from the NVIDIA supply chain while reducing customer migration friction.
Strategic Significance
If Trainium succeeds at scale, AWS becomes the only hyperscaler with both infrastructure and custom silicon advantages -- a structural moat that Azure and GCP cannot replicate without years of chip development.