Amazon Web Services, Inc. (AWS) is a company owned by Amazon that offers cloud computing services and application programming interfaces (APIs) to individuals, businesses, and governments. These services are available on a pay-as-you-go basis, meaning users only pay for the resources they use.
Many customers use AWS with a feature called autoscaling, which helps increase computing power during times of high activity and reduce it when demand is lower to save costs. AWS provides a wide range of services, including networking, computing, storage, software tools, and support for internet-connected devices (IoT). These services are managed through AWS server farms, which allow customers to avoid the need to manage, expand, or update physical hardware and software themselves.
A key service is Amazon Elastic Compute Cloud (EC2), which gives users access to virtual computers that are highly reliable and can be controlled through the internet using tools like REST APIs, a command-line interface, or the AWS console. These virtual computers mimic real computers by offering features such as central processing units (CPUs), graphics processing units (GPUs), memory, storage options, operating systems, and pre-installed software like web servers and databases.
AWS delivers its services through a global network of server farms. Pricing depends on factors like how much a customer uses the service, the type of hardware and software selected, and the level of availability, security, and reliability required. Customers can choose to use a single virtual computer, a dedicated physical computer, or groups of computers. Amazon handles some security tasks, such as protecting data centers, while customers are responsible for other security aspects, like managing accounts and updating software. AWS operates in many regions worldwide, including nine in North America.
Amazon promotes AWS as a way for customers to quickly and affordably access large-scale computing power without building their own physical server farms. All services are billed based on usage, though each service measures usage differently. As of the first quarter of 2023, AWS holds 31% of the global cloud infrastructure market, while Microsoft Azure and Google Cloud have 25% and 11%, respectively, according to Synergy Research Group.
Services
As of 2025, AWS includes more than 200 products and services, such as computing, storage, networking, databases, analytics, application services, deployment tools, management tools, machine learning, mobile services, developer tools, RobOps, and tools for the Internet of Things. The most commonly used services are Amazon Elastic Compute Cloud (EC2), Amazon Simple Storage Service (Amazon S3), Amazon Connect, and AWS Lambda, which is a tool that runs code without needing a server and can be triggered by many events, such as web requests.
These services provide features through APIs, which allow users to use them in their own applications. APIs are accessed using the HTTP protocol and follow the REST architectural style. Older APIs also use the SOAP protocol, while newer ones use JSON exclusively. Users can interact with these APIs in several ways, such as through the AWS console (a website), by using software development kits (SDKs) written in languages like Python, Java, and JavaScript, or by making direct REST calls.
History
The beginning of AWS started in the early 2000s. After creating Merchant.com, a platform that helped third-party retailers build their own online stores, Amazon worked on service-oriented architecture to help its engineering team grow. This effort was led by Allan Vermeulen, who was then Amazon’s Chief Technology Officer.
At the same time, Amazon wanted to improve how quickly its software engineers worked. Matt Round, an engineering leader, suggested changes like giving engineering teams more independence, using REST, standardizing infrastructure, reducing bureaucracy, and using continuous deployment. He also encouraged engineers to spend more time building software instead of doing other tasks. To help, Amazon created a shared IT platform. This allowed engineering teams, which spent 70% of their time on tasks like IT and infrastructure, to focus more on customer-related work. Additionally, to handle traffic spikes, especially during holidays, Amazon moved services to commodity Linux hardware and used open source software. Tom Killalea, Amazon’s first Chief Information Security Officer, led this effort, making data centers and services fast, reliable, and affordable.
In July 2002, Amazon.com Web Services, managed by Colin Bryar, launched its first web services, opening its platform to all developers. By 2004, over 100 applications were built using these services. This unexpected interest from developers showed Amazon that there was a strong demand for more tools.
By summer 2003, Andy Jassy took over Bryar’s work after Allan Vermeulen, who was Jeff Bezos’ first choice, declined the role. Jassy planned an "Internet OS" made up of basic infrastructure tools that would help speed up software development. By fall 2003, Amazon identified databases, storage, and compute as the first services to launch.
Jeff Barr, an early AWS employee, said the idea for EC2, S3, and RDS came from a group of people, including Vermeulen, Jassy, Bezos, and others. Jassy said the idea was the result of brainstorming with ten top technology experts and ten top product managers. Werner Vogels said Amazon wanted to make the process of "invent, launch, reinvent, relaunch, start over, rinse, repeat" as fast as possible. This led to changes like using "two-pizza teams" (small teams) and distributed systems. These changes helped create AWS and its mission to share the basic parts of the Amazon.com platform. Brewster Kahle, co-founder of Alexa Internet (acquired by Amazon in 1999), said his company’s compute tools helped Amazon solve big data problems and later influenced AWS innovations.
Jassy gathered a team of 57 people from engineering and business backgrounds to start these projects. Most of the team members came from outside Amazon. They included Jeff Lawson, CEO of Twilio; Adam Selipsky, CEO of Tableau; and Mikhail Seregine, co-founder of Outschool.
In late 2003, the idea for compute, which later became EC2, was revised when Chris Pinkham and Benjamin Black shared a paper describing a plan for Amazon’s retail computing infrastructure. This plan was fully standardized, automated, and used web services for storage. They also suggested selling access to virtual servers as a service. Pinkham, Willem van Biljon, and Christopher Brown later developed EC2 with a team in Cape Town, South Africa.
In November 2004, AWS launched its first public infrastructure service: Simple Queue Service (SQS).
On March 14, 2006, AWS launched Amazon S3 cloud storage, followed by EC2 in August 2006. Pi Corporation, a startup co-founded by Paul Maritz, was the first outside company to use EC2. Microsoft was one of the first large companies to use EC2. SmugMug, an early AWS user, saved about $400,000 in storage costs using S3. Werner Vogels said S3 started with 8 microservices in 2006 and had over 300 by 2022.
In September 2007, AWS started its annual Start-up Challenge, offering $100,000 in prizes to U.S. entrepreneurs and developers using AWS services like S3 and EC2. Justin.tv, which Amazon later bought in 2014, participated in the first event. Ooyala, a media company, won the contest.
As of June 16, 2022, AWS offers two block-storage options: EC2 Instance Store and Elastic Block Store (EBS). EBS features include:
– EBS volume tagging to help users find and organize EBS resources.
– Software-level RAID arrays to create high-performance groups of EBS volumes.
Other services from this time include SimpleDB, Mechanical Turk, Elastic Beanstalk, Relational Database Service, DynamoDB, CloudWatch, Simple Workflow, CloudFront, and Availability Zones.
In November 2010, all of Amazon.com’s retail sites moved to AWS. Before 2012, AWS was part of Amazon.com, so its revenue was not listed separately in financial reports. In 2012, industry experts estimated AWS revenue to be over $1.5 billion.
On November 27, 2012, AWS held its first major annual conference, re:Invent, in Las Vegas. The event had over 150 sessions and attracted around 6,000 attendees. Andy Jassy and Werner Vogels gave keynotes, and Jeff Bezos joined Vogels for a discussion. Reed Hastings, CEO of Netflix, announced plans to move all of Netflix’s infrastructure to AWS.
In 2013, AWS started offering certifications for engineers to show cloud computing expertise. That same year, AWS launched Activate, a program to help startups use AWS credits, third-party tools, and free expert support.
In 2014, AWS created the AWS Partner Network (APN) to help companies using AWS grow and succeed.
In January 2015, Amazon Web Services bought Annapurna Labs, an Israeli microelectronics company, for about $350–370 million.
In April 2015, Amazon reported that AWS was profitable, with $1.57 billion in sales and $265 million in operating income during the first quarter. Jeff Bezos called AWS a fast-growing business.
Availability and topology
As of October 2025, AWS operates in 38 separate geographical "regions": nine in North America, two in South America, nine in Europe, four in the Middle East, one in Africa, thirteen in the Asia–Pacific, and three in Australia and New Zealand.
Most AWS regions are automatically available for all AWS accounts. Regions added after March 20, 2019, are opt-in regions, meaning users must manually enable them to use them. For opt-in regions, Identity and Access Management (IAM) resources, such as users and roles, are only available in regions that are enabled.
Each region is located entirely within one country, and all data and services remain within that region. Every region contains multiple "Availability Zones," which are groups of one or more separate data centers. These data centers have backup power, networking, and connectivity and are housed in different buildings. Availability Zones are intentionally isolated from each other to prevent problems in one zone from affecting others. Some services, like S3 and DynamoDB, can work across Availability Zones, while others can be set up to copy data across zones to manage traffic and avoid service interruptions.
As of December 2014, AWS operated about 1.4 million servers across 11 regions and 28 Availability Zones. The AWS global network includes over 700 Edge locations worldwide, located in North America, Europe, Asia, Australia and New Zealand, Africa, and South America. The AWS Cloud includes 120 Availability Zones across 38 regions, with plans to add 10 more Availability Zones and three additional regions in Saudi Arabia, Chile, and the AWS European Sovereign Cloud.
As of March 2024, AWS announced plans to launch six new regions in Malaysia, Mexico, New Zealand, Thailand, Saudi Arabia, and the European Union. In mid-March 2023, AWS signed an agreement with the New Zealand government to build large data centers in New Zealand.
Pop-up lofts
AWS has "pop-up lofts" in various places around the world. These lofts help introduce AWS to entrepreneurs and startups in different technology industries through physical spaces. People can work, relax, or learn about AWS services inside the lofts. In June 2014, AWS opened its first temporary pop-up loft in San Francisco. In May 2015, it expanded to New York City, and in September 2015, it opened in Berlin. The fourth location, in Tel Aviv, was open from March 1 to March 22, 2016. A pop-up loft in London was open from September 10 to October 29, 2015. The lofts in New York and San Francisco are closed permanently due to the COVID-19 pandemic, while the Tokyo location remains open in limited capacity.
Charitable work
In 2017, AWS started a program called AWS re/Start in the United Kingdom to help young adults and military veterans learn new skills related to technology. AWS worked with the Prince's Trust and the Ministry of Defence (MoD) to offer training for young people facing challenges and former military members. AWS also partnered with companies such as Cloudreach, Sage Group, EDF Energy, and Tesco Bank to support these efforts.
In April 2022, AWS announced it had pledged over $30 million over three years to support early-stage start-ups led by Black, Latino, LGBTQIA+, and women founders through its AWS Impact Accelerator program. The program provides eligible start-ups with up to $225,000 in cash, credits, training, mentorship, and technical help. It also includes up to $100,000 in AWS service credits.
Reception
In 2014, AWS stated its goal to use 100% renewable energy in the future. In the United States, AWS partnered with renewable energy providers to support its operations. These partnerships include Community Energy of Virginia for the US East region; Pattern Development, in January 2015, to build and operate Amazon Wind Farm Fowler Ridge; Iberdrola Renewables, LLC, in July 2015, to build and operate Amazon Wind Farm US East; EDP Renewables North America, in November 2015, to build and operate Amazon Wind Farm US Central; and Tesla Motors, to use battery storage technology to meet power needs in the US West (Northern California) region.
In 2016, Greenpeace evaluated major technology companies, including AWS, Microsoft, Oracle, Google, IBM, Salesforce, and Rackspace, based on their use of "clean energy." Greenpeace assessed companies on their use of renewable energy sources, transparency, renewable energy commitments, energy efficiency, greenhouse gas reduction efforts, renewable energy purchases, and advocacy. AWS received an overall "C" grade. Greenpeace acknowledged AWS’s progress toward greener computing and its plans to build multiple wind and solar farms in the United States. The group noted that Amazon does not provide clear information about its carbon footprint.
In January 2021, AWS joined an industry agreement to make data centers climate neutral by 2030, called the Climate Neutral Data Centre Pact. As of 2023, Amazon is the largest corporate buyer of renewable energy worldwide, a position it has held since 2020, and has over 20 gigawatts of renewable energy capacity globally. In 2022, 90% of all Amazon operations, including data centers, were powered by renewable energy. In 2024, AWS paid $650 million for a data center connected to the 300 MW portion of the 2.5 GW Susquehanna nuclear power plant. In 2025, AWS signed an $18 billion power purchase agreement with the station for 840–1,200 MW in 2029 and 1,680–1,920 MW in 2032, continuing through 2042. The power increase depends on Amazon’s expansion of the data center.
The US Department of Homeland Security uses software called ATLAS, which runs on Amazon Cloud. The software scanned over 16.5 million records of naturalized Americans and flagged about 124,000 for manual review by USCIS officers regarding denaturalization. Some data came from the Terrorist Screening Database and the National Crime Information Center. The algorithm and its criteria were not made public. Amazon faced protests from employees and activists over its collaboration with authorities on this project.
The contract for Project Nimbus faced criticism from shareholders and employees, who were concerned the project might enable human rights abuses against Palestinians during the Israeli–Palestinian conflict. Critics worried the technology could increase surveillance of Palestinians, allow unlawful data collection, and support the expansion of Israeli settlements on Palestinian land. A government procurement document listed "obligatory customers" of Nimbus, including two major Israeli state-owned weapons manufacturers, Israel Aerospace Industries and Rafael Advanced Defense Systems. This document was published in 2021 and updated periodically through October 2023.
AWS has worked with Chinese companies, such as Hikvision and Dahua Technology, which are under US government sanctions. These companies have been accused of enabling mass surveillance and human rights abuses in China and other regions.
Security incidents
In July 2025, a security researcher found a problem in the Amazon Q Developer Extension (version 1.84) for Visual Studio Code. The issue involved an unauthorized request to the project's GitHub repository, which added an instruction telling the AI assistant to delete local files and AWS resources. AWS admitted the problem in Security Bulletin AWS‑2025‑015 and released version 1.85 to remove the harmful instruction. The company said no customer systems or data were harmed. However, critics, including Corey Quinn in Last Week in AWS, noted concerns about how the issue was handled, pointing out the lack of a changelog entry, a CVE assignment, or a public statement beyond the bulletin.
In response to the Log4Shell vulnerability, AWS provided quick fix updates to reduce risks in Java applications across different environments, such as standalone servers, Kubernetes clusters, and Elastic Container Service (ECS). These fixes worked for both AWS and non-AWS systems. However, researchers from Unit 42 at Palo Alto Networks found serious flaws in these patches that could be used to gain access to systems and increase access rights, possibly allowing attackers to take full control of a host system. AWS fixed these issues by releasing updated patches on April 19, 2022. Users who used the first patches were told to update to the latest versions to reduce security risks.
In April 2024, security researchers from Miggo security found a setup issue in AWS Application Load Balancer (ALB) that could let attackers bypass access controls and gain access to web applications. The problem happened because some users configured ALB's process of passing user information to third-party services in a way that allowed unauthorized access to data. On July 11, 2024, AWS confirmed the issue affected its customers, and on July 19, 2024, AWS updated its documentation to suggest safer ways to set up ALB.
Issues
Some customers of AWS have reported getting very large bills, often called "surprise bills." This can happen for several reasons, such as mistakes in how services are set up, security issues, difficult pricing when using many AWS services together, and unexpected costs for moving large amounts of data.
Pricing
AWS charges fees for data moved between Availability Zones in the same region (inter-AZ) and between different regions (inter-region). Pricing depends on factors such as where the data starts and ends, the specific AWS services being used, and how the network is set up. Different services use different ways to move data, and the method chosen for sending data—such as VPC Peering, Transit Gateway, or AWS PrivateLink—can also change the total cost.