Amazon Web Services (AWS)
The world's largest cloud service provider, offering GPU compute rental, the machine learning platform SageMaker, self-developed AI chips Trainium/Inferentia, and the AI application building platform Bedrock.
Unparalleled underlying infrastructure scale and global coverage, with extremely high customer trust and switching costs accumulated from over 20 years of enterprise-grade cloud service operations. The richest PaaS component ecosystem (over 200 full-stack services) creates strong platform lock-in effects.
Company Overview
Amazon Web Services (AWS), founded in 2006, is the pioneer of global public cloud services. AWS was initially built as a platform for Amazon to absorb its internal e-commerce infrastructure capabilities, and later unexpectedly became Amazon's most profitable and fastest-growing business line.
AWS is currently the undisputed leader in the global cloud computing market, with revenue exceeding the combined total of second-place Microsoft Azure and third-place Google Cloud. In the AI era, AWS faces intense competition from Azure (which bundles OpenAI), but it maintains its leadership position in the cloud computing market through its broader customer base, richer PaaS services, and self-developed chip strategy.
Core Business
Elastic Compute (EC2 & GPU Instances) | Core Revenue (~40% of Revenue)
- GPU Compute Instances: Offers multiple GPU instance types based on NVIDIA H100/B200, AMD MI300X, and self-developed Trainium chips, covering both training and inference scenarios.
- Spot Instances: AWS's unique bidding instance model that allows customers to use idle GPU compute at significant discounts. The checkpoint fault-tolerance mechanism commonly used in large model training makes Spot Instances an effective way to reduce AI training costs.
AI Platforms (SageMaker + Bedrock) | Fastest Growing
- Amazon SageMaker: The industry's most mature full-lifecycle machine learning management platform, covering the entire workflow from data labeling to model training, deployment, and monitoring.
- Amazon Bedrock: Provides managed API services for multiple foundation models (including Anthropic Claude, Meta LLaMA, AI21, etc.), allowing customers to access top-tier AI capabilities without deploying models themselves.
Self-Developed AI Chips | Long-Term Strategic Positioning
- Trainium2: Custom chips designed for AI training scenarios, with superior compute density and energy efficiency compared to same-generation GPUs. AWS is building an ultra-large-scale training cluster (Project Rainier) based on Trainium2.
- Inferentia2: Custom chips designed for inference scenarios, offering far superior cost-performance compared to general-purpose GPUs.
Data & Storage
S3 (object storage), Redshift (data warehouse), Kinesis (real-time data streaming), and more form the underlying infrastructure of the AI training data pipeline.
Technology Moat
Infrastructure Scale Effects: AWS operates 105 availability zones across 33 geographic regions worldwide, with infrastructure scale far exceeding any competitor. This scale delivers lower unit costs and better elasticity and redundancy.
PaaS Ecosystem Lock-in Effects: AWS offers over 200 services covering all IT domains including compute, storage, databases, networking, security, and AI. Once enterprises deeply adopt AWS PaaS services (such as DynamoDB, Redshift, Kinesis), the cost and complexity of migrating to other cloud platforms becomes extremely high.
Market Landscape
| Dimension | Data |
|---|---|
| Global Cloud Market Share | ~31%, Ranked #1 |
| Annualized Revenue | ~$100 Billion |
| Key Competitors | Microsoft Azure (~24%), Google Cloud (~11%) |
| Core Customers | Industries ranging from startups to government agencies |
Financials & Growth
| Metric | Data |
|---|---|
| AWS Revenue (2024) | ~$100 Billion |
| Operating Margin | ~30% |
| Capital Expenditure | ~$65 Billion (including AI infrastructure) |
| Growth Drivers | Surging AI compute rental demand + Growth in Bedrock model hosting services |
Risks & Summary
Key Risks:
- Microsoft Azure, leveraging its exclusive access to OpenAI, is growing rapidly in the enterprise AI market and eroding AWS's share.
- Ongoing cloud price wars continue to pressure profit margins.
- Shipment volumes and performance of self-developed chips still need validation.
Core Industry Value: AWS is the definer of AI infrastructure-as-a-service. It does not pursue owning the most advanced AI models; instead, it makes compute, storage, and models as readily available as utilities like water and electricity. From startups to Fortune 500 companies, a vast number of AI applications are built on AWS.