In 1998, a group of senior engineers at Amazon began planning a radical overhaul, separating presentation, business logic, and data while improving reliability, scale, performance, security, and cost efficiency. Their proposal became the distributed computing manifesto, and they needed to solve what Jeff Bezos called the muck. The 70 to 80% of unglamorous, uncreative, annoying work that was consuming engineering teams at Amazon.

If you believe companies and developers would build applications from scratch on top of these infrastructure services if the right selection existed, then the operating system becomes the Internet. When we looked at that in the 2003, we realized that none of the key components of that Internet operating system had been built.

But let's zoom out. One day, Werner Vogels was asked to give a talk at Amazon on systems design, but he almost didn't go. He remembers thinking web servers and databases, how hard could it be? But he did go. And what he saw at Amazon blew his mind.

I kind of discarded Amazon. And it wasn't until I I visited Amazon at one time, the things that I would consider research at the at Cornell. At at Amazon, they were just going with production. They were doing it day in, out.

This was 2004, and of course, the rest is history. Werner is now CTO of Amazon.

Who? Me?

But what happened in between? In this episode, we're gonna dig in to the very origins of Amazon Web Services and try to understand where this idea to build Amazon Web Services came from. And in the later episodes, we're gonna dig into how Amazon Web Services was developed, was launched, and how it grew into what it is today, something that is so important that when it goes down, it's world news. Scaling DevTools is sponsored by WorkOS. If things start going well, some of your customers are gonna start asking for enterprise features.

You could spend ages tearing your hair out, building these things yourself, or you could use WorkOS.

We use WorkOS for our SAML and OIDC integrations. It's a pretty exceptional product. It it makes everything regarding authentication pretty seamless, and it's been instrumental for us to onboard our enterprise customers much faster.

WorkOS helps your dev tools start selling to enterprises much faster. If you use their user management or you can get your first million monthly active users completely free. The year was 2004, two years before Amazon Web Services launched. Back then, if you wanted to launch a web product, you had to go buy physical servers, rent space at a data center, install the OS yourself, set everything up yourself, hope you'd got the right number of servers. If you didn't have enough servers and your website spiked, it's gonna crash.

On the other hand, if no one visited your site, you've wasted all that money on upfront costs. This was just how it was. Amazon was living in this world too and at a scale that almost no one had seen before. Werner described it this way, the scale and diversity of their operation was unlike anything I had ever seen. Amazon's architecture was at least a decade ahead of what I had encountered at other companies.

It was more than just a high performance website. We are talking about everything from high volume transaction processing to machine learning, security, robotics, binning millions of products. Anything that you could find in a distributed systems textbook was happening at Amazon, and it was happening at an unbelievable scale. When they offered me a job, I couldn't resist. But let's rewind to the very start.

In 1994, Jeff Bezos launched Kadabra Inc, later renamed Amazon, and launched the website in July 1995. In its first year, Amazon made half $1,000,000 in revenue. By a second year, it made $15,000,000 in revenue. By 1997, it was doing a $150,000,000 of revenue per year. At this time, it had one and a half million customers in over a 150 countries, and they had more than 600 employees.

It IPO ed in 1997 at a valuation of about half $1,000,000,000. It was an unbelievable success story. Personally, I remember the first time my uncle told me about Amazon because he'd been using it to buy books and the experience was incredible. It was just unlike anything that was out there. Here's how Werner remembers it.

Amazon was moving at a rapid pace, building and launching products every few months. Innovations we take for granted today, and that's just the customer facing side. Amazon began with a two tier architecture, a monolithic stateless application called OPDOS, plus a growing set of shared relational databases. As Amazon expanded internationally and into more product categories, these shared databases became the bottleneck. In 1998, a group of senior engineers at Amazon began planning a radical overhaul, separating presentation, business logic, and data while improving reliability, scale, performance, security, and cost efficiency.

Their proposal became the distributed computing manifesto, and they needed to solve what Jeff Bezos called the muck. The 70 to 80% of unglamorous, uncreative, annoying work that was consuming engineering teams at Amazon. Here's Jeff Barr, the man who would become AWS's first developer evangelist.

Jeff Bezos came up with this phrase he calls muck. Basically, the the 70 or 80% of the of the the non the non sexy just kind of things you have to do in order to to be able to have a a stable platform to actually do the creative part. So so once we did all this, we realized that if it was so much work for us, it's probably gonna be as much work for any other developer that would like to do something similar.

It would have been really hard for us to understand how seriously this was taken within Amazon if we hadn't seen a leaked internal memo from a former Amazon engineer, Steve Yigi. Steve recalls a company wide directive sent out by Bezos. All teams must expose their data and functionality through service interfaces. With no direct database access, no shared memory, no backdoors, all communication between systems must happen through network accessible APIs, and services must be designed so that they can be accessed by external users. And Bezos allegedly also said, no exceptions.

Anyone not complying with this will be fired. By 2005, Amazon had fully changed its engineering culture and adopted this approach. It's not clear exactly when this happened, but at some point, Bezos realized Amazon needed to become a platform. And the Amazon systems built for selling books could become repurposable and be used by other companies, especially retailers at the beginning. And so Amazon quietly began externalizing some of its services.

This began with something called merchant.com. And in the early two thousands, Target asked Amazon to help run target.com. Werner Vogels remembers their internal reaction like this.

Target came to Amazon and says, we really love what you guys do on the web. Can you build our store? And there we were standing with the duct tape and w 40 behind our back and thinking, why would anybody want that?

It wasn't easy, and Andy Jassy remembers it like this. The decoupling was far more difficult than anticipated. We'd built so many dependencies between services. But, of course, the work that Steve described them doing helped this a lot. July 2002 was probably the time when the first Amazon Web Services service launched before it was called Amazon Web Services.

It was called the Amazon Associates Web Services, and it enabled users to query Amazon programmatically. Jeff Barr, who later became Amazon's first developer evangelist, recalls it like this.

We've got over 27,000 registered developers. We process very, very close to 10,000,000 calls to the service per day. And I see literally hundreds of different innovative sites. There were people who have taken the data that we provide, and they put it together, and they structure it in in just ways that are just, you know, beyond imagination sometimes. You look at that and say, wow.

Why didn't we think of that?

It helped them prove something profound that I think Jeff Bezos was very aware of. That external developers will outperform your internal roadmap if you just give them all the access. You make it accessible. Around about this time, there was a leadership off-site that has become a bit legendary. Andy Jassy recalls that off-site in 2003 like this.

And so we kinda took a step back and we asked ourselves, this was now the 2003. We said, well, if you believe companies and developers would build applications from scratch on top of these infrastructure services if the right selection existed, and we believe they would if the right selection existed, then the operating system becomes the Internet, which is really different from what had been the case for the last thirty years. And then we asked if there's gonna be an Internet operating system, what are the key components? What's already been built? And what given what Amazon's good at could we meaningfully contribute to?

And when we looked at that in the 2003, we realized that none of the key components of that Internet operating system had been built. And when we thought about what we were good at, Amazon has always been a technology company at its heart. One that applied the technology to the retail space first, but always a technology company.

There are so many myths and contradictory stories about the AWS origin. I wanted to give you one more perspective from someone impartial, Matt Klein. Matt wasn't there at the start, but he was very early. He joined in January 2010 and stayed on until September 2012 working on EC two in Seattle in the early days. Matt's now the founder of BitDrift, a mobile observability tool.

And his take on the AWS origins is very interesting.

I think there are a lot of folk stories that have gone around, you know, that AWS was born out of they had extra capacity that they weren't using during Christmas or something like that. And to my knowledge, that's not true. I think they just saw a need and they said, let's see if this can happen. And I think it's just a great example of being in the the right place at the right time, seeing the opportunity, and executing really well. Amazon was uniquely positioned to capture that value because Google, Microsoft, simply too slow.

Amazon just said, we're just gonna ship it. I mean, we're just gonna put it out there, and it was rough in the beginning.

Jeff Bezos, in one of his latest shareholder letters, described why he felt like Amazon Web Services was such a big opportunity because it met all three of the categories that he liked to check before Amazon would go into anything. It has to be a large and growing market. It has to have a high return on capital, and it has to leverage Amazon's expertise. And AWS checked all of these categories. And so Andy Jassy and his team of about 50 or 60 people started to work on the operating system of the Internet, what would become AWS.

In the next episode, we talk about how Amazon launched AWS, what they launched first, and we talk about the surprising history of EC two, which was developed not in Seattle, but in Cape Town, South Africa. Check back for that episode very soon.