How Amazon grew an awkward side project into AWS, a behemoth that’s now 4 times bigger than its original shopping business

If so, you did it with the help of Amazon Web Services. The less glamorous sibling to Amazon’s operations in e-commerce, streaming video, and smart devices, AWS is no less ubiquitous, deploying millions of computers worldwide, humming away somewhere in the cloud.

For all those AWS customers the on-demand cloud computing platform isn’t just another vendor. They rely on it so heavily that it resembles a public utility—taken for granted, but essential to keep the machinery humming. In the past 12 months each of the companies mentioned above has stated in Securities and Exchange Commission filings that they “would be adversely impacted” if they lost their AWS service. Hundreds more companies—Netflix, Zoom, Intuit, Caesars Entertainment—have reported the same risk factor to the SEC in the past year. By the way, the SEC uses AWS. (So does Fortune.)

And those are but the tiniest fraction of AWS customers. AWS—initially run by Andy Jassy, who went on to succeed Jeff Bezos as Amazon’s CEO—won’t say how many customers it has, only that it provides computing power, data storage, and software to millions of organizations and individuals. Now, even as Amazon lays off a reported 10,000 workers, Wall Street analysts expect another blowout performance from its web services division. That’s probably why few if any of those staffing cuts will affect this relatively recession-proof part of Bezos’s empire. (Amazon won’t say how many of its 1.5 million employees work for AWS.)

For years AWS has brought in more profit than all other divisions of Amazon combined, usually by a wide margin. AWS’s operating profit last year, $18.5 billion, was nearly three times the operating profit reported by the rest of the company ($6.3 billion). AWS pulled in $58.7 billion of revenue in this year’s first nine months; if it were independent, it would easily rank in the Fortune 100.  

How did this offshoot of an online retailer come to rule the lucrative cloud-computing industry, towering over tech giants such as Microsoft and Google, which might have seemed better positioned to dominate?

AWS’s ascent is so unlikely that it demands an explanation. It reveals the power of a truly iconoclastic culture that, while at times ruthless, ultimately breeds innovation and preserves top talent by encouraging entrepreneurship.

The best-known origin story of AWS is that it started when Amazon had some spare computer capacity and decided to rent it out to other companies. That story won’t die, but it isn’t true. The real story traces a circuitous path that could easily have ended in a ditch. It’s grounded in a philosophy that still guides AWS’s progress.

“To me, it’s the concept of insurgents versus incumbents,” says Adam Selipsky, who became AWS’s CEO last year when his predecessor, Jassy, took over as Amazon CEO. Selipsky, 56, speaks quietly, conveying an understated intensity. “One thing that I think is really important, that we intentionally worry about all the time,” he says, is that “we continue to keep the customer need dancing in front of our eyes at all times.”

The real story of the AWS insurgency began with Amazon’s innovative responses to two problems. First: By the early 2000s, Amazon—still known mainly as an online bookseller—had built from scratch one of the world’s biggest websites, but adding new features had become frustratingly slow. Software engineering teams were spending 70% of their time building the basic elements any project would require—most important, a storage system and an appropriate computing infrastructure. Building those elements for projects at Amazon scale was hard, and all that work merely produced a foundation on which to build the cool new customer-pleasing features Amazon was seeking. Every project team was performing the same drudgery. Bezos and other Amazon managers started calling it “undifferentiated heavy lifting” and complaining that it produced “muck.” 

In response, Selipsky recalls, company leaders began to think, “Let’s build a shared layer of infrastructure services that all these teams can rely on, and none of them have to spend time on general capabilities like storage, compute capabilities, databases.” Amazon’s leaders didn’t think of it as an internal “cloud”—the term wasn’t widely used in the tech world yet—but that’s what it was.

The second problem involved other websites wanting to add links to Amazon products on their own pages. For example, a website about cooking might recommend a kitchen scale and include a link to the Amazon.com page for the product. Amazon was all for it, and would send them a bit of code they could plug into their site; if someone bought the product through the link, the site owner earned a fee. But as the program grew, cranking out bits of code for every affiliate site became overwhelming, and those affiliates’ website developers wanted to create their own links and product displays instead of the ones Amazon sent them. So in 2002 Amazon offered them a more advanced piece of software, enabling them to create far more creative displays. The new software was complicated. Users had to write software rather than just plug it in. Yet thousands of developers loved it immediately.

When Amazon launched a fuller, free version of the software building block a few months later, it enabled anyone, not just affiliates, to incorporate Amazon features into their sites. The surprise: A lot of the downloads were going to Amazon’s own software engineers. The building block turned out to be a proof of concept for the labor-lightening innovations that Amazon itself was looking for.

A picture was emerging. Amazon desperately needed to free its software developers from creating muck. Developers everywhere, not only its own, were starving for new tools that did just that. “We very quickly figured out that external developers had exactly the same problems as internal developers at Amazon,” Selipsky says.

But was that a business for Amazon? During a 2003 offsite at Bezos’s house, the company’s top managers decided that it could be. That decision was the turning point, especially significant because it could so easily have gone the other way. Amazon’s customers were consumers who bought “new, used, refurbished, and collectible items,” as the company told investors at the time. Why would anyone imagine this company could build a business selling technology services to software developers?

The decision to plunge ahead revealed a subtle distinction that outsiders didn’t understand. The world saw Amazon as an online retailer, but the company’s leaders never thought of it that way. They thought of it as “a technology company that had simply applied its technology to the retail space first,” Jassy later told Harvard Business School professors who were writing a case study. For that kind of company, AWS looked like a promising bet.

Coming out of the 2003 offsite, Jassy’s job was to build a team and develop AWS. He wrote a proposal for it as a cloud-computing business. The document, one of the famous six-pagers used at Amazon’s executive meetings instead of PowerPoint (which is banned), reportedly went through 31 revisions.

It took three years before AWS went live. In 2005 Jassy hired Selipsky from a software firm to run marketing, sales, and support, Selipsky recalls: “Amazon called and told me there was this initiative for something about turning the guts of Amazon inside out, but other companies could use it.” AWS’s first service, for data storage, “was such a novel concept that it was even hard to explain and hard for me to understand,” he says.

Wall Street didn’t get it. “I have yet to see how these investments are producing any profit,” a Piper Jaffray analyst said in 2006. “They’re probably more of a distraction than anything else.”

The rest of the world didn’t get it either. “I cannot tell you the number of times I got asked, with a quizzical look on people’s faces, ‘But what does this have to do with selling books?’ ” Selipsky recalls. “The answer, of course, was: AWS has nothing to do with selling books. But the technology we use to sell books has everything to do with AWS and what we can offer customers.” Those customers were software developers, an entirely new target market that baffled outsiders.

AWS was prepared for that reaction. One of Amazon’s principles reads in part: “As we do new things, we accept that we may be misunderstood for long periods of time.”

On the day in March 2006 when AWS finally launched its inaugural service—S3, for Simple Storage Service—Selipsky was at a trade show in Santa Clara, Calif., “in a windowless, internet-less conference room,” as he describes it, unable to learn how the launch was going. At day’s end he and a colleague ran outside to call Seattle for news. 

“We were told that 12,000 developers had signed up,” he says, a note of marvel still in his voice. “On the first day. It was just amazing.”

Five months later AWS launched its other foundational service, EC2, for Elastic Compute Cloud, which was also instantly popular. The revolution had begun. Instead of raising millions of dollars to buy servers and build data centers, startups could now get online with a credit card, and pay a monthly bill for just the computing power and storage they used. If their new app was a hit, they could immediately engage all the cloud services that they needed. If it bombed, they weren’t stuck with rooms of junk equipment. As a Silicon Valley entrepreneur and early AWS customer told Wired in 2008: “Infrastructure is the big guys’ most powerful asset. This levels the field.”

In response to that historic shift, AWS’s potential competitors did … nothing. “A business miracle happened,” Bezos told a conference years later. “This is the greatest piece of business luck in the history of business so far as I know. We faced no like-minded competition for seven years. I think the big established enterprise software companies did not see Amazon as a credible enterprise software company, so we had this incredible runway.”

Selipsky suspects an additional motivation: “They either didn’t believe this could be a real business, or they were so threatened by what it would do to their own business models, and the way they were overcharging customers, that they didn’t want to believe it.”

No one, not even at Amazon, foresaw how massive a business cloud computing would be, or AWS’s dominance in the space. To understand how this happened, it’s worth examining the company’s guiding principles.

Eye-roll alert: Every company has principles, missions, visions, values; the vast majority are indistinguishable and sound as if they were written by committees, which they probably were. Some of Amazon’s leadership principles, as they’re called—there are 16—sound that way, until they get a little “peculiar,” to use a favorite Amazonian word.

For example, principle No. 11 begins, “Earn trust.” Leaders, it explains, “are vocally self-critical, even when doing so is awkward or embarrassing. Leaders do not believe their or their team’s body odor smells of perfume.” This peculiarity is a badge of pride at Amazon; its web page for job seekers even says that its use of the principles “is just one of the things that makes Amazon peculiar.”

Not every Amazonian observes every principle all the time; in a company of 1.5 million employees, that’s not realistic. But Amazon’s batting average is high.

To answer the basic question of why a retailer would even think of creating AWS, consider principle No. 1, seemingly the hoariest of them all: “Customer obsession.” Amazon sees itself as a tech company and sees the world as 8 billion potential customers. That’s one reason AWS made sense for a bookseller.

“I cannot tell you the number of times I got asked … ‘But what does this have to do with selling books?’”

Adam Selipsky, CEO, Amazon Web Services

Amazon allows new projects lots of time, as with AWS, in part to make sure decisions are based on data. An unusual principle states that leaders “work to disconfirm their beliefs.” Groupthink is comforting, contagious, and dangerous. Being able to invoke one of the principles enables doubters to speak up. 

“We have senior engineers who will stop a meeting and say, ‘We’ve got to disconfirm our beliefs—we’re going too far here without checking,’ ” says Mai-Lan Tomsen Bukovec, who oversees AWS’s storage services. “That’s actually kind of revolutionary in terms of corporate culture.”

It’s not a culture for everyone. Amazon is a famously demanding place to work, and there are plenty of stories of employees who found it to be too much. Media reports have criticized Amazon’s treatment of workers, and the company is battling unionization efforts at some of its e-commerce warehouses. It’s noteworthy that last year Amazon added a new leadership principle: “Strive to be Earth’s best employer.”

“It’s not good for our business and not good for our customers if we turn out great employees and burn them out, and they leave after a couple of years,” says Matt Garman, an early AWS employee who now oversees sales and marketing. “Sometimes there are people who don’t like the culture, don’t like those leadership principles. It’s not a good fit for them. People like the culture or they don’t like the culture, and I think that’s okay. But we want people here for the long term.”

Asked to describe AWS’s strategy, Tomsen Bukovec says, “That’s not a word we use a ton.” 

The foundation of conventional strategy, the subject of hundreds of books and articles, is understanding a company’s industry and competitors. That approach gets us nowhere with Amazon. What industry is it in? No one industry encompasses selling dog food and selling computing power. 

So does Amazon even have a strategy? “Yes,” says Ram Charan, an adviser to CEOs and boards, and coauthor of a book on Amazon’s management system. But “it’s not a competitive strategy,” he says. “It’s a customer strategy.” 

That’s a mind bender. Business is competition, and business strategy is inherently competitive strategy. Except that at Amazon it isn’t. If it had been—if Amazon had been conventionally competitor-focused—AWS probably wouldn’t exist. 

Colin Bryar, a former Amazon executive, says he’s often asked what Amazon is going to build next. Can it repeat what it did with AWS, create an out-of-the-blue business, unexpected and underestimated, in which it becomes dominant? “That’s not the first question Amazon asks,” Bryar says. “They ask, ‘What’s the next big customer problem we can go try to solve?’ ”

The word “big” is key. At Amazon’s size—analysts expect revenue exceeding $500 billion for 2022—small problems are simply not of interest. When company leaders identify a sufficiently big problem, they must then conclude that Amazon can solve it, and that customers will adopt the solution. Those are not easy or quick questions to answer.

Cloud computing will grow 20% annually through 2026, far faster than any other segment of infotech, according to the Gartner tech consulting firm. It’s no longer just smaller companies and startups who don’t want to invest in their own server systems. Many AWS customers are increasing their spend, and some “spend literally hundreds of millions of dollars per year on AWS,” says Gartner analyst Raj Bala, who sees the contracts. “I’m not shocked anymore to see a $200 million annual commitment, which is astonishing.”

Yet AWS’s dominance of the market will likely diminish even as its revenue grows. With a 44% share of the market, AWS has 20 points over Microsoft’s 24%—but that lead is shrinking, says Bala. “In the next five, six, seven years, that gap is going to be very, very narrow, if not equal.” That’s because “a lot of late adopter enterprises are coming to market,” he says, “and a lot of these folks will gravitate to Microsoft because they’ve got an existing contractual relationship with Microsoft.”

The narrowing gap with Microsoft is probably inevitable. AWS’s great challenge for the future is to maintain the discipline that made it a global colossus.

Losing that discipline is insidiously easy. Jim Collins, author of Good to Great, which identifies the factors shared by the world’s most successful companies, has also written an analysis of failure, How the Mighty Fall. Winners invariably maintain discipline, and loss of discipline is always an element of decline. One of the principle threats? Attempts to control workers by overregulating them. “Bureaucracy subverts discipline,” he tells Fortune. 

When a company is growing as fast as AWS, it can be tempting to weaken hiring standards. “As you grow, you start to bring in some of the wrong people,” he says, speaking of companies generally. “If they don’t get the intensity of being there, they shouldn’t be there, but if enough of them stay, you try to control them with bureaucracy. Then the right people get out, which creates a cycle.”

With success and growth come further threats to discipline. When a business is riding high, “easy cash erodes cost discipline, and that discipline is hard to recover once you lose it,” he says. Expansion brings risks, too: Responding desperately to deteriorating performance, the business bets on “undisciplined discontinuous leaps”—acquisitions or expansions for which it isn’t ready. 

At the top of its game, bigger and stronger than any competitor, AWS must now meet an enviable challenge but a challenge nonetheless: the curse of success. Its most crucial task is to maintain the unwavering rigor—the discipline—of its principles and processes.

Selipsky seems to understand the need. Asked to define his job, he is silent for several seconds. Then, quietly but emphatically, he says his job “is to ensure that the positive, productive, useful elements of what got us to this stage—that we hold those dear, and we safeguard them, and we don’t let them slip away. We don’t become incumbents.” 

Amazon’s next big, thorny problem to solve

What might be the next industry to get Amazon’s AWS-style mega-venture treatment? The leading candidate is health care. 

In 2018 Amazon bought PillPack, an online pharmacy, and last summer it paid $3.9 billion for One Medical, a membership-based primary-care provider operating across the U.S., saying in its announcement that “we think health care is high on the list of experiences that need reinvention.” 

No one would disagree. For a company that seeks big problems to solve, this may be the biggest opportunity of all. Health care is the largest sector of the U.S. economy, and the industry is growing fast worldwide. 

Data is the problem at the heart of health care’s inefficiency and unfathomable, wearisome customer experiences—and it’s possible that it could be the solution.

That data is staggering in quantity and mostly unstructured—handwritten notes and X-ray and lab reports, sometimes of life-and-death importance—in an industry that is the last bastion of fax machines. 

It’s a particularly attractive conundrum to Amazon because of the company’s dominance of cloud computing. AWS is already deeply entrenched in the industry, used by hospitals, pharma companies, equipment makers, insurers, pharmacy benefit managers, the Centers for Medicare and Medicaid Services, and more. 

Another potential advantage is Amazon’s massive international workforce and its enormous health care needs and expenses. Just as Amazon developed AWS by observing its own software needs and seeing them mirrored elsewhere, its own challenges as a growing corporate behemoth now may point the way to a new market opportunity.

This article appears in the December 2022/January 2023 issue of Fortune with the headline, “How Amazon’s cloud took the world by storm.”