AWS (Amazon Web Services)

Definition: The largest and most mature cloud platform, run by Amazon, offering several hundred on-demand computing, storage, networking, and machine learning services from a global network of data centers. Launched in 2006 starting with S3 and EC2, AWS grew out of infrastructure Amazon had already built to run its own e-commerce operations at scale, and it became profitable well ahead of the rest of Amazon’s business, effectively bootstrapping the modern cloud computing industry. It remains the revenue and market-share leader among the hyperscalers, commonly cited as holding roughly a third of global cloud infrastructure spend.

Core Services & Concepts

EC2 — Virtual Machines (VMs), resizable virtual servers, the original AWS compute product and still the backbone of most AWS architectures
S3 — Cloud Storage Systems, object storage used across almost every AWS architecture, commonly the default choice for backups, static assets, and data lakes
Lambda — Serverless Computing and Cold Starts, the service that popularized serverless functions, billed per invocation and per millisecond of execution
EKS — Kubernetes (K8s), managed Kubernetes control plane, one of several container options alongside ECS (AWS’s own simpler orchestrator) and Fargate (serverless containers)
IAM — Identity and Access Management (IAM), AWS’s own permission system, famously granular and famously easy to misconfigure, a leading root cause of publicized cloud breaches
VPC — Virtual Private Cloud (VPC) and Subnets, isolated network environment per account, the networking foundation almost every other AWS service deploys into
RDS / DynamoDB — managed relational (RDS) and NoSQL (DynamoDB) database services, removing patching, backups, and failover from the customer’s responsibility
CloudFront — Content Delivery Network (CDN) and Edge Computing, AWS’s CDN, commonly paired with S3 to serve static assets and cached content close to end users

How Pricing Works

On-demand pricing is pay-as-you-go by the second or hour with no upfront commitment, the default and most expensive way to run most services
Reserved Instances and Savings Plans offer roughly 30-70% discounts in exchange for a 1- or 3-year usage commitment, the standard lever for predictable, steady-state workloads
Spot Instances sell unused capacity at steep discounts, commonly up to 90% off, in exchange for the instance being reclaimable on short notice, well suited to fault-tolerant batch jobs
A modest free tier covers small amounts of EC2, S3, and Lambda usage for the first 12 months on a new account, useful for evaluation but not production traffic
Data egress, transferring data out of AWS to the internet or another cloud, is billed separately and commonly the least-understood line item on an AWS invoice

Pros

Largest service catalog by far, commonly cited at 200+ distinct services covering nearly every conceivable infrastructure need
Most mature ecosystem and highest job market demand of any cloud platform, translating into the deepest pool of tooling, tutorials, and hired expertise
Best documentation and community support by sheer weight of market share and tenure
Broadest global footprint of data center regions and availability zones, useful for latency-sensitive or data-residency-constrained deployments
Deep enterprise contract flexibility (Enterprise Discount Program) available for large committed spend
Consistent multi-year track record of new service launches, commonly cited at 100+ announcements a year at re:Invent, keeping the platform’s edge over competitors

Cons

Pricing is notoriously complex and easy to get wrong, with hundreds of SKUs and pricing dimensions spread across services
Steep learning curve from the sheer number of overlapping services solving similar problems in slightly different ways
IAM misconfiguration is a leading cause of real-world cloud security breaches, the permission model’s flexibility is also its biggest footgun
Console and service UX quality varies significantly across the catalog, older services feel noticeably less polished than newer ones
Egress fees and cross-service data transfer costs are a common source of unexpectedly large bills on data-heavy or multi-region architectures
Vendor lock-in risk is real for teams leaning heavily on AWS-specific services like DynamoDB or Step Functions rather than portable, open-source-compatible alternatives

Comparison: AWS vs Google Cloud Platform (GCP) vs Microsoft Azure

	AWS	GCP	Azure
Primary strength	Largest service catalog, most mature ecosystem	Data analytics (BigQuery) and Kubernetes	Enterprise/Windows integration, hybrid cloud
Typical pricing model	On-demand pay-as-you-go, Reserved Instances/Savings Plans for commitment discounts	Pay-as-you-go with automatic sustained-use and committed-use discounts	Pay-as-you-go with Reserved Instances and Hybrid Benefit licensing credits
Best fit	Enterprises needing broad service coverage, teams with dedicated cloud engineers	Data-heavy and Kubernetes-native workloads, teams wanting Google’s AI/ML tooling	Organizations already invested in Microsoft/Windows/Active Directory
API/product style	Deep, sprawling catalog (200+ services), CLI/SDK-first	Smaller, more curated catalog, gcloud CLI, strong open-source/K8s alignment	PowerShell/Azure CLI, ARM templates, deep Visual Studio/.NET tooling

Best For

Enterprises needing nearly every possible cloud service under one roof, including niche and legacy-workload support
Teams with dedicated DevOps or cloud engineers who can absorb the platform’s complexity in exchange for maximum flexibility

Real Examples

Netflix, Airbnb, and a large share of the modern internet’s backend infrastructure run partly on AWS
Amazon’s own retail business and Twitch both run internally on AWS, alongside countless startups defaulting to it as their first cloud provider
Slack and Pinterest both built significant portions of their infrastructure on AWS during their early growth years

Use Cases

Large-scale enterprise backends
Data lakes and analytics pipelines
ML training and inference infrastructure via SageMaker
Hybrid cloud deployments via AWS Outposts
Disaster recovery and multi-region failover architectures
Static website and media hosting via S3 paired with CloudFront

Integration Notes & Common Pitfalls

Overly permissive IAM policies (wildcard resource/action grants) are the single most common security misconfiguration, least-privilege policies should be the default rather than an afterthought
Forgetting to tear down unused resources, idle EC2 instances, unattached EBS volumes, old snapshots, is a frequent source of quietly accumulating cost
Cross-region and cross-AZ data transfer charges are easy to overlook when designing multi-region architectures, they add up faster than compute costs on data-heavy workloads
The sheer breadth of overlapping compute options, choosing between Lambda, ECS, EKS, and Fargate, means teams often need real architectural judgment rather than a single “correct” default
Multi-account strategies via AWS Organizations are considered best practice at scale but add real setup complexity for teams that start out with a single flat account

Code Example

# AWS CLI — upload a file to S3 and list bucket contents
aws s3 cp ./build/index.html s3://my-app-bucket/index.html --acl public-read
aws s3 ls s3://my-app-bucket/ --recursive

FAQ

Is AWS the best choice for a small startup? Not always — its flexibility comes with real operational overhead, smaller teams sometimes ship faster on a more opinionated platform before migrating to AWS as they scale.

How does AWS pricing actually work for a typical web app? Costs combine compute (EC2/Lambda), storage (S3/EBS), data transfer, and any managed services used (RDS, CloudFront), most teams underestimate data transfer and monitoring costs relative to raw compute.

What is the difference between EKS, ECS, and Fargate? EKS runs standard Kubernetes, ECS is AWS’s own simpler container orchestrator, and Fargate is a serverless compute mode usable with either, the choice mostly comes down to whether a team wants Kubernetes portability or AWS-native simplicity.

Does AWS offer a simpler onboarding path for beginners? AWS has invested in guided setup tools and its Free Tier, but the platform’s breadth means most newcomers still face a steeper initial learning curve than on a more opinionated platform.

What is AWS re:Invent? AWS’s annual flagship conference, typically held in Las Vegas, where the majority of major new service and feature announcements are made each year.

History

Launched in 2006 with S3 and EC2, built on infrastructure Amazon had already developed to run its own retail operations at scale
Became profitable well ahead of the rest of Amazon’s business, and its operating income has been a major contributor to Amazon’s overall profitability for years
Introduced Lambda in 2014, popularizing the serverless computing model that competitors would later replicate across their own platforms
Has maintained the largest cloud market share since the category’s inception, commonly estimated at roughly 30% of global cloud infrastructure spend
Andy Jassy, who led AWS from its early days, became Amazon’s overall CEO in 2021, reflecting how central the cloud division had become to the company

Referenced by

Dig deeper

Spot an error in this note?