On March 14, 2006, in an era when Twitter (now X) was just emerging and Facebook was just opening up to the public, Amazon Web Services (AWS) quietly launched a service called Amazon Simple Storage Service (Amazon S3). Without a grand launch event, but with just a blog post, it completely revolutionized the way startups and enterprises worldwide deploy IT infrastructure. Today, AWS has grown from an experimental project initially designed to solve internal pain points within Amazon into a massive cloud empire with annual revenue exceeding $142 billion.
From pre-recorded software to cloud computing: The starting point of reversing the 80/20 rule
The birth of AWS stemmed from an unsolved "mathematical problem." In the early 2000s, Amazon's internal engineering team discovered that they were spending as much as 70% to 80% of their time on underlying infrastructure such as computing, storage, and databases, leaving only 20% to 30% of their time for actual product innovation.
Andy Jassy, who was then chief of staff to Amazon founder Jeff Bezos (and later became Amazon's CEO), thought: if this ratio could be reversed, or if Amazon could build a highly scalable and cost-effective underlying architecture that handled all the undifferentiated, heavy lifting, then global developers could focus 80% of their time on creating real business value. This initial idea ultimately led to the official establishment of AWS in 2006.
回顧20年前，創辦一家科技公司的門檻極高，需要龐大的資金購買實體伺服器與建置資料中心。Amazon S3的出現卻徹底打破這個局面。它讓雅加達的個人開發者能與紐約的財富500強企業，享有同等級別、無限擴充且「按需付費」 (Pay-as-you-go)的儲存服務。這也間接催生了過去20年來全球高達1600多家估值超過10億美元的「獨角獸」新創公司的蓬勃發展。
Behind the Scenes: 4 Little-Known Facts About AWS
Besides Amazon S3 and EC2 execution entities that transformed the industry ecosystem, there are many little-known and interesting anecdotes in the history of AWS:
• AWS's first product manager rose from intern to CEO:Matt Garman, the current CEO of AWS, began his career as an MBA intern at Amazon in 2005. When he was officially hired as AWS's first product manager in 2006, the global sales team consisted of only three people. He personally wrote product pages, developed pricing strategies, and participated in building the core Amazon EBS (Elastic Block Store) service, perfectly embodying AWS's pioneering spirit of creating something from scratch.
• The S3 architecture sketches were born in a bar:The core system design sketches for Amazon S3, AWS's first official service, were actually conceived by the team in the Six Arms bar in Seattle. They established a decentralized principle of "Design for failure," treating hardware failures as the norm. Today, the system stores over 500 trillion data objects.
• EC2 was built by an 8-person team based in South Africa:Five months after the launch of S3, Amazon EC2 Execution Instances (virtual servers), which laid the foundation for cloud computing, were initially developed by eight engineers in Cape Town, South Africa. They traversed more than 10,000 kilometers and a 10-hour time difference, using a bunch of laptops in the corner of their office to simulate the prototype of a data center rack.
• Blogs existed even before products:Jeff Barr, AWS's first Web Services Advocate, published his first AWS blog post in November 2004, long before the launch of S3 and EC2. In an era when the concept of "enterprise blogs" was not yet widespread, he accumulated more than 3300 articles over the past 20 years through his unadorned and direct dialogue with developers, becoming a classic model for communication between the technology industry and developers.
The 20-Year Challenge of AWS: Microsoft Azure vs. Google Cloud
Although AWS has successfully defined the modern cloud computing infrastructure (IaaS) standard over the past 20 years and remains the market leader, as the focus of the cloud battlefield shifts from "underlying computing power and storage" to "generative AI and data applications," Microsoft Azure and Google Cloud are exerting unprecedented competitive pressure on AWS with their unique moats.
This battle is no longer simply about "who has the cheaper servers," but rather "who can monetize AI the fastest."
Microsoft Azure: A game-changer with the help of the OpenAI Alliance and the SaaS ecosystem.
The biggest threat that Microsoft Azure currently poses to AWS lies in its pervasive "binding of enterprise-grade AI applications and office ecosystems".
• Exclusive OpenAI computing power and model advantages:Microsoft's massive investment in OpenAI has made Azure the preferred, and often the only, way for enterprises to access top-tier large language models like GPT-4o. When enterprises want to implement the most powerful generative AI, Azure OpenAI Service is often the option with the least resistance.
• The ultimate binding of the Copilot ecosystem:AWS excels at selling "Lego bricks" (underlying APIs and computing power) for developers to assemble themselves; while Microsoft sells "pre-built castles." By deeply integrating AI into Microsoft 365 (Word, Excel, Teams), GitHub Copilot, and the Windows operating system, Microsoft has successfully transformed AI from a "development tool for the IT department" into a "productivity tool for all employees."
• Painless migration of traditional B2B customers:Many large traditional enterprises are already deeply entrenched in Microsoft's Active Directory (AD) and SQL Server systems. Microsoft's highly attractive Enterprise License Bundle (EA) program makes it almost "natural" for these customers to allocate their AI and cloud budgets to Azure, a pain point that AWS, lacking a large number of SaaS (Software as a Service) end-user applications, cannot directly compete with.
Google Cloud: Dominating the discourse among "data scientists" and the "open source development ecosystem"
If Azure wins in the integration of business applications, then Google Cloud wins in the hearts of "open source believers and data scientists".
• Open source architecture and Kubernetes' home-field advantage:As the inventor of Kubernetes (K8s) and TensorFlow, Google enjoys immense prestige in cloud-native and open-source architectures. Many startups and AI developers who started with microservice architectures naturally prefer to use GCP's GKE (Google Kubernetes Engine), giving GCP a more "approachable" brand image in the open-source community than AWS.
• The most powerful data analysis engine:Data is the fuel for AI, and GCP's BigQuery is still widely recognized as one of the most powerful and user-friendly enterprise data warehousing services. Google has successfully integrated BigQuery seamlessly with its Gemini machine learning models, allowing enterprises to train and deploy AI models directly in the database without having to move massive amounts of data between different services.
• Embrace the open source model and startup community:Faced with the closed alliance between Microsoft and OpenAI, Google has chosen to play the "open card." In addition to its own Gemini model, GCP is actively integrating the Hugging Face ecosystem and vigorously promoting the open-source Gemma model. For AI startups that do not want to be tied to a single vendor (vendor lock-in) and are extremely focused on model autonomy, GCP provides a highly user-friendly development sandbox.
Hybrid Cloud and Multi-Cloud Architecture: A Breakthrough Tool for Challengers
In the past, AWS, as the dominant player, tended to keep its customers firmly locked within its own ecosystem (such as using AWS Outposts). However, Microsoft and Google, as challengers, realized early on that "multi-cloud architecture" is the true norm for large enterprises.
Google's Anthos (now part of Google Cloud) and Microsoft's Azure Arc both allow businesses to manage applications and data on "competitors' clouds" (or even AWS) or in on-premises data centers. This open strategy of "I allow you to use other people's infrastructure, but control and the AI brain remain on my platform" has successfully broken down some of AWS's infrastructure moat and attracted many financial and healthcare giants who are afraid of cloud lock-in.
Analysis of viewpoints
Looking back at AWS's 20-year history, it is essentially a history of the evolution of modern internet technology.
As Dr. Werner Vogels, Vice President and Chief Technology Officer of AWS, stated, the real revolution brought about by cloud computing in 2006 was not just server virtualization, but the "democratization of access rights." AWS transformed supercomputing power, which was previously only affordable to tech giants, into a basic resource that college students in dormitories could access simply by swiping their credit cards. This "democratization of resources" directly ignited the wave of mobile internet, video streaming, and big data computing in the Web 2.0 era.
Standing at this 20th anniversary juncture, AWS, while having grown into a tech behemoth with annual revenue exceeding $100 billion, faces entirely different challenges. In the past, its competitors were traditional data center and hardware vendors; now, it must prove itself, amidst the fierce competition from Microsoft Azure and Google Cloud, to continue playing its crucial underlying role in "lowering the development threshold and enabling ubiquitous innovation" in the era of generative AI. In the next 20 years, the focus of the cloud battlefield will officially shift from "the widespread adoption of computing power and storage" to "the widespread adoption of AI intelligence."
