20 Years of AWS Cloud, Time Really Flies

AWS has marked its 20th anniversary. It has provided more than 240 cloud services to date, releasing thousands of new features every year, and this blog has accumulated more than 4,700 posts.

A personal reflection is also included. On March 13, 2006, after meeting Jeff Barr as a keynote speaker in Seoul, the author's perspective on API-based services and opening up to third-party developers changed, and from then on they began to truly feel the potential of the cloud. Later, the author's company became one of AWS's early customers in Korea in 2014, and the cloud became established as a tool enabling both personal research and large-scale projects.

Over the past 10 years, the center of technology has shifted from deep learning to generative AI, LLMs, and today's agentic AI. Amid these changes, AWS emphasizes the principle of "distinguishing between conspicuous fads and real trends, keeping in mind that today's niche can become tomorrow's mainstream," and continues to innovate in a direction that first listens to and solves the problems customers actually need solved.

The major turning points of the past 20 years were also revisited.

• Amazon S3, Amazon EC2(2006), Amazon RDS, Amazon VPC(2009), Amazon DynamoDB, Amazon Redshift(2012), Amazon WorkSpaces, Amazon Kinesis(2013), AWS Lambda(2014), AWS IoT(2015)
• Containers: Amazon ECS(2014), Amazon EKS(2017), AWS Fargate
• Databases: Amazon Aurora(2017), Aurora Serverless v1(2018), Aurora Serverless v2, Amazon Aurora DSQL(2025)
• ML/AI: Amazon SageMaker(2017), the next-generation SageMaker and SageMaker AI in 2024, Amazon Bedrock, Bedrock AgentCore
• Custom chips and price-performance: AWS Graviton, AWS Inferentia, AWS Trainium, EC2 M9g, Trn3 UltraServers
• Hybrid/edge: AWS Outposts
• Developer tools: Amazon CodeWhisperer, Amazon Q Developer, Kiro, and the recent Kiro autonomous agent
• Expanded model choice: Amazon Titan, Amazon Nova, Nova Forge, Nova Act, Nova 2 Lite

Finally, the message is clear. AWS's next 10 years won't be a race to bolt on more AI features, but will depend on turning AI into tools and services so that customers can solve the problems they actually face.