If you glance at headlines today about antitrust scrutiny of Amazon's cloud computing business, it's easy to feel like it's the late 1990s all over again. Back then, Microsoft was facing similar charges, which culminated in a 2001 antitrust settlement with the federal government. Is Amazon heading toward a similar collision with government regulators? Or is it an exaggeration to draw comparisons between Microsoft's 1990s-era "embrace, extend, extinguish" playbook and Amazon's present-day cloud strategy?
The answer, I think, lies somewhere in between. Here's why.
Microsoft and Monopoly: A Very Short History
During the 1990s, Microsoft established near-total dominance of the operating system market. Apple, which sold the only real alternative to PCs, was flailing on both the hardware and software fronts, eroding its market share to almost zilch. Meanwhile, IBM's OS/2 operating system offered an alternative to Windows for PC users for a time in the early 1990s. But, by 1995, Microsoft had backed IBM into a corner, too: Big Blue continued developing OS/2, but it shipped more and more of its own PCs with Windows 95.
Microsoft leveraged its control of the PC operating system market to cow other hardware manufacturers into similar positions by forcing most of them to pre-install Windows on the machines they sold.
Microsoft's business tactics weren't limited to the operating system market, of course. By controlling which OS came preinstalled on the vast majority of PCs, Redmond also ensured that its Web browser, office software, networking protocols and much more were foisted upon consumers.
And in the rare cases where competitors in the software market did manage to hold their own against Microsoft, Microsoft squelched them through its infamous "embrace, extend, extinguish" strategy. That strategy centered on making competitors' software compatible with Windows in ways that crippled the software's functionality and/or made it difficult for end-users to switch between Windows and alternative platforms. The policy helped kill the Netscape browser (until it was reborn as Firefox, at least), among other important applications and platforms. Microsoft also wielded the policy against various open source technologies, including Linux.
Today, of course, Microsoft has abandoned these sorts of practices, which reached their peak around the turn of the millennium. The Microsoft of 2020 is a company that is in love with Linux and open source, and is almost unrecognizable from its past self. But its business strategies in the late 1990s nonetheless left a major impression on the tech industry.
Is Amazon the New Microsoft?
In certain key ways, it's easy to hear echoes of Microsoft's past in Amazon's cloud computing strategy today.
Amazon's share of the cloud computing market is about 50%. That puts it well ahead of the next-largest company in this space (which happens to be Microsoft, whose Azure cloud enjoys market share of around 16%).
To protect its cloud dominance, Amazon employs various strategies that make it hard for customers to leave. The most basic involves ensuring that most of Amazon’s cloud tools and services work only with the AWS cloud, which means it is difficult for Amazon cloud customers to extend their workloads into other clouds. But Amazon’s efforts to limit competition go much further than mere tooling.
Embracing and extending in the AWS cloud
Like Microsoft in the 1990s, Amazon engages with third-party platforms and tools in ways that suggest that it wants to lock users into its own versions of platforms that could theoretically work with any cloud.
Take container orchestration, for instance. Originally, Amazon offered a proprietary container orchestration solution, ECS. Then, when it became clear by 2017 that Kubernetes, an open source container orchestration platform, was eating the world, Amazon finally embraced Kubernetes, which it used as the basis for another container orchestration solution, EKS. (Amazon continues to offer ECS in addition to EKS.)
Yet EKS is arguably far removed from open source Kubernetes. It has its own CLI tool, eksctl, instead of using generic kubectl. It is tightly integrated with Amazon's IaaS offerings (namely, EC2), making it quite different from other Kubernetes derivatives (like Rancher and OpenShift) that work with a variety of cloud platforms.
You might argue, then, that EKS is an effort to embrace and extend Kubernetes--and, perhaps, eventually extinguish competing Kubernetes distributions by getting users hooked on the proprietary tooling and dependencies that come with EKS.
You could draw similar conclusions about Amazon’s database services. AWS supports open source databases (like MySQL), while also offering its own proprietary takes on the same types of databases (like Aurora). A critic might contend that Amazon wants to draw users into its cloud databases with promises of open source compatibility, then get them hooked on Amazon’s proprietary solutions. Once you are invested in a database like Aurora, it becomes harder to move to a different cloud than it would be if you stick with a pure-play open source database.
AWS Outposts is another interesting example of a product that in theory gives Amazon cloud customers more freedom and flexibility, but only in a gotcha kind of way. Outposts makes it possible to deploy various Amazon cloud services on customers' own servers. That means you can use AWS services in a data center that you own, rather than sharing hardware with other AWS customers in one of Amazon's data centers. The caveat, though, is that--unlike other cloud vendors' hybrid cloud frameworks, such as Azure Arc and Google Anthos--Outposts requires users to purchase hardware directly from AWS.
So, you get to own the Outposts hardware. But does ownership mean much when Amazon decides exactly which hardware you'll be using? To be sure, Outposts does simplify some compliance needs by keeping workloads out of Amazon's public data centers. But it still imposes rigid restrictions on choice and flexibility, in ways that other cloud vendors' competing platforms don't.
Acquiring and killing with cost
Another interesting facet of Amazon’s strategy is the way it goes about acquiring certain third-party platforms that could potentially compete with it.
Take, for example, Macie, Amazon's data loss prevention solution. Macie originated as an independent startup, which Amazon quietly purchased in 2017. Amazon didn’t kill the technology; it makes Macie available to all AWS customers. However, it does so at a price that is so high, you have to wonder if Amazon actually expects anyone to use Macie.
Another possible interpretation of Amazon’s Macie pricing strategy is that the company has simply decided that it can basically charge whatever it wants because there are few competitors for the product--which is exactly what Microsoft did when it priced Windows 95 at more than $200, or about $350 in today’s dollars. (By comparison, Windows 10 Home currently retails for under $150, no doubt because Microsoft knows that customers today have viable alternatives to Windows.)
Differences between Microsoft and Amazon
Despite these parallels between Amazon today and Microsoft 20 years ago, there are some very important differences between the two companies and their business strategies.
For one, Amazon's dominance of the cloud computing market is much smaller than Microsoft's control of the PC software market ever was. Amazon may enjoy a dominant position with 50% market share, but that's not at all the same as Microsoft's near-monopoly over PC platforms in the 1990s, when the market share of competitors to Windows sank into the single digits.
For another, Amazon isn't exactly alone in many of the strategies it uses to keep customers locked into its cloud. All of the major cloud vendors charge steep data egress fees, for example, which can be a barrier to moving data from one cloud to another. All of them also design most of their tools to work only with their own clouds. It would be almost as hard to migrate from Azure or Google Cloud to AWS as it would be to go in the opposite direction.
It’s also hard to point fingers too singularly at Amazon for the way it incorporates certain third-party platforms into its cloud portfolio. For example, you can’t be too critical of EKS given that Red Hat also took Kubernetes and tacked on extensions and dependencies to create OpenShift. OpenShift does run on multiple clouds, so it doesn’t lock users into a particular infrastructure in the same way that EKS does. But it does encourage tooling dependency in ways that make it similar to EKS.
Finally, it's important to note that Amazon's cloud business is just a side operation that reinforces Amazon's online retail business. In contrast, Microsoft was and always has been purely a software-focused company (its experiments with hardware devices notwithstanding).
This means that Amazon today and Microsoft in the 1990s have very different end games. Amazon wants to use its cloud business to supercharge its dominance in online retail--which is what really has the government worked up. A company that wants to monopolize one industry (like PC software) is one thing. A conglomerate that monopolizes multiple, almost totally unrelated industries (like retail and cloud computing) is much worse, from a regulator's point of view.
On top of this, Amazon faces particular pressure right now because authorities on both sides of the Atlantic Ocean are increasingly worried about what they perceive as big tech’s threat to data privacy. Arguably, Amazon has less potential to harm consumers’ privacy rights than do companies like Google and Facebook, because Amazon doesn’t directly collect as much personal information as these other companies. Nonetheless, the regulatory ire surrounding privacy rights and big tech has spilled over into Amazon’s territory, too, reinforcing other types of complaints about Amazon’s business practices.
In short, it may be fair to say that Amazon is using Microsoft's former anti-competitive playbook as a guide for some of its cloud computing strategies. But a wholesale comparison between the two companies misses the mark. Amazon doesn't dominate the cloud to the same extent that Microsoft dominated PC software. Nor are Amazon’s cloud business practices unique in the way that Microsoft’s strategies were in the 1990s: Other tech companies behave similarly to Amazon today, whereas in the 1990s only Microsoft was aggressively locking out competitors.
Plus, if Amazon were only focused on cloud computing, I doubt we'd be seeing as many headlines as we are about looming antitrust action against a company that began as a humble online bookseller. It's only because of Amazon's retail business strategy, combined with the fervor surrounding big tech’s alleged threat to privacy rights, that AWS has come so squarely into regulators' crosshairs.