AWS suits teams that want one vendor for infrastructure, databases, and AI model access instead of stitching together specialized providers. Its Trainium and Graviton chips undercut third-party GPU and CPU pricing by 20 to 40 percent, and its roughly 127,000-person organization runs data centers across more than 30 regions worldwide.
AWS runs more than 200 cloud services across compute, storage, databases, and AI, operated as Amazon's dedicated cloud division. Its distinguishing feature is custom silicon built in-house: Trainium chips for AI training and Graviton chips for general compute, positioned as lower-cost alternatives to third-party GPUs. Amazon Bedrock gives access to models from Anthropic, Meta, Mistral, and OpenAI.
Founded: 2006 · HQ: Seattle, WA, USA · Team: 100000-150000 · CEO: Matt Garman · Funding: Division of Amazon.com (NASDAQ: AMZN). FY2025 revenue $128.73B; Q1 2026 revenue $37.59B (+28% YoY). Amazon 2026 capex: $200B, primarily AWS expansion.
About Amazon Web Services
Amazon Web Services launched publicly in 2006 with Amazon S3 (object storage) and Amazon EC2 (compute), services championed by Andy Jassy and a small team at Amazon who saw internet companies rebuilding the same infrastructure from scratch instead of renting it on demand. AWS grew from that side project into a distinct Amazon business unit headquartered in Seattle, Washington, and now generates the majority of parent Amazon.com's (NASDAQ: AMZN) operating income, reporting $128.73 billion in AWS revenue for full-year 2025. At AWS re:Invent in December 2025, the company launched Trainium3 UltraServers, packing 144 Trainium3 chips into a single system for 4.4 times the compute performance of Trainium2, and previewed Trainium4 for broad availability by late 2026. Amazon Bedrock's model catalog expanded to include Google, Kimi AI, MiniMax AI, Mistral AI, NVIDIA, OpenAI, and Qwen models, and gained reinforcement fine-tuning that shows 66% accuracy gains over base models. In April 2026, AWS announced Bedrock Managed Agents powered by OpenAI in limited preview and confirmed Anthropic now trains its most advanced models on AWS Trainium and Graviton chips. AWS AI Factories launched the same period to bring dedicated AWS AI infrastructure into enterprise and government data centers, and in May 2026 AWS announced AI-agent stablecoin payment infrastructure with Coinbase and Stripe. AWS CEO Matt Garman joined the division as an intern in 2006, spent over a decade managing Amazon EC2 and Compute Services, and was appointed CEO in May 2024, reporting to Amazon CEO Andy Jassy, who led AWS from 2003 to 2021 before succeeding Jeff Bezos atop Amazon. AWS employs roughly 127,000 people directly across engineering, sales, and support. In early 2026, Amazon cut about 30,000 corporate roles, with AWS management layers specifically targeted, while separately planning to hire 11,000 software engineers and developers. AWS infrastructure spans more than 30 geographic regions and over 100 Availability Zones across North America, Europe, Asia-Pacific, South America, the Middle East, and Africa. AWS's long-term research priorities center on custom silicon (Trainium, Graviton, Inferentia), agentic AI infrastructure through AgentCore and the Strands framework (downloaded more than 25 million times), and sovereign AI deployments for regulated industries through AI Factories. The company does not publish a standalone AI safety policy but is a strategic investor in Anthropic and distributes Claude models through Amazon Bedrock under commercial agreements. AWS holds an estimated 29 percent of the global cloud infrastructure market per third-party analyst estimates (Q3 2025), ahead of Microsoft Azure and Google Cloud, a lead built on service breadth, chip investment, and two decades of enterprise trust.
Mission
Enable any organization to access reliable, scalable, and inexpensive cloud computing infrastructure and AI services on demand.
Products
- Amazon EC2 + S3 + Core Infrastructure (Compute, storage, and networking cloud infrastructure): https://aws.amazon.com/products/
- Amazon Bedrock (Fully managed foundation model access (Anthropic, OpenAI, Google, Mistral, Meta)): https://aws.amazon.com/bedrock/
- Amazon SageMaker AI (Managed ML model development, training, and deployment platform): https://aws.amazon.com/sagemaker/
- AWS Trainium + Graviton (Custom AI training and general compute silicon): https://aws.amazon.com/machine-learning/trainium/
- Amazon Q (Enterprise AI assistant for code, data, and business workflows): https://aws.amazon.com/q/
- AWS AI Factories (On-premises sovereign AI infrastructure for enterprises and governments): https://aws.amazon.com/solutions/ai-factories/
Compliance
SOC 2 Type II, SOC 3, ISO 27001:2022, ISO 27017:2015, ISO 27018:2019, ISO 27701:2019, FedRAMP High, PCI DSS, HIPAA-eligible, HITRUST CSF, CMMC, DoD IL2-IL6, FIPS 140-3, CSA STAR Level 2
Links
Website · GitHub · Twitter · LinkedIn · Blog · Docs
Frequently Asked Questions
How much funding has Amazon Web Services raised?
AWS is a wholly owned division of Amazon.com, Inc. (NASDAQ: AMZN) and does not raise outside funding or report standalone AWS financials beyond revenue. Its capital comes from Amazon's corporate budget, which commits roughly $0.2 trillion in 2026 capital expenditure, primarily for AWS data centers and AI infrastructure. Outside that internal funding, AWS's biggest AI-related capital moves are strategic investments rather than raises: $25 billion into Anthropic and an OpenAI compute commitment expanded to $100 billion over eight years.
What products does Amazon Web Services make in 2026?
AWS's core infrastructure lineup includes Amazon EC2 (virtual machines), Amazon S3 (object storage), Amazon RDS and Aurora (managed databases), Amazon Lambda (serverless compute), and Amazon VPC (private networking). Its AI lineup centers on Amazon Bedrock, which offers managed access to foundation models from Anthropic, Meta, Mistral, Google, OpenAI, and Qwen; Amazon SageMaker AI for model development; and Amazon Q, an AI assistant built into the console for developers and business teams working with AWS resources. Custom silicon (Trainium for training, Graviton for compute, Inferentia for inference) underpins most of these services, and AWS Marketplace lists 42,240 third-party software products on top of them.
How does Amazon Web Services handle data privacy and compliance?
AWS holds FedRAMP High authorization for US federal workloads and serves agencies including the CIA and Department of Defense under DoD Impact Levels 2 through 6. Its certification portfolio also covers SOC 2 Type II, ISO 27001:2022, ISO 27018:2019 (cloud privacy), HIPAA-eligible services with signed BAAs, HITRUST CSF, and CMMC. EU customers get GDPR coverage through Standard Contractual Clauses and Data Processing Addenda, with resident data center options in Ireland and Frankfurt. AWS runs a shared responsibility model: it secures the infrastructure, and customers secure what they build on top of it.
Who are Amazon Web Services' main competitors?
Microsoft Azure holds roughly 20% of the cloud market versus AWS's estimated 29%, and wins Windows-heavy enterprise accounts through deep integration with Active Directory, Office 365, and Teams. Google Cloud trails both in share but leads on TPU-based training infrastructure for specific machine learning research workloads. Oracle Cloud competes mainly on database migrations, where its autonomous database and existing licensing agreements create incumbent advantages. On raw AI chips, NVIDIA's GPUs still dominate training workloads globally, and AWS positions Trainium as a lower-cost alternative rather than a like-for-like replacement.
How do you start using Amazon Web Services' products?
Entry points differ by product: Amazon EC2 and Amazon S3 are self-serve through the AWS Management Console for compute and storage, while Amazon Bedrock offers API access to managed foundation models including Anthropic's Claude without infrastructure setup. Amazon SageMaker AI targets teams that want to train, fine-tune, or deploy their own models rather than call a third-party one. AWS AI Factories, aimed at enterprises and governments, brings that same AWS infrastructure on-premises instead of into AWS's own data centers.
Top Alternatives
- Google Cloud: Pick AWS for Bedrock model variety; pick Google Cloud for Vertex AI and TPU access.
- Microsoft: Pick AWS for model choice; pick Microsoft Azure for OpenAI exclusivity and enterprise integration.