One of the interesting aspects of platform shifts in computing is the way they end up redrawing platform boundaries, reorganizing what is inside the platform versus built on top of it. The seemingly simple act of changing platform boundaries resets incumbent advantages and fundamentally reshapes industry structure. When IBM dominated the mainframe market the platform was a vertically integrated hardware and software stack. In the PC era, the platform boundary shrank to the microprocessor on the hardware side and Microsoft Windows on the software side. Companies went from saying our application runs on an IBM 360 mainframe computer to saying our application runs on Wintel (i.e. Windows on Intel x86). IBM lost its platform dominance and became a supplier (along with Compaq, HP, Dell and others) on top of the Intel platform. Microsoft and Intel in turn became huge franchises with massive market power.
With cloud computing, the platform boundary has changed once more and expanded to encompass the whole cloud stack with all its software services. Developers now talk about how their app runs on the Amazon cloud (AWS). The frame of reference is now the AWS API. The Intel processors and Linux operating systems powering AWS are mere components inside the platform.
The competitive implications of changing platform boundaries are hard to tease out in the present and often only become obvious in hindsight. In the early days of the PC revolution, a computer manufacturer like Compaq looked every bit as important as Intel or Microsoft. Compaq was on a hyper-growth trajectory riding the PC wave and became the fastest company in history to $1B in sales. But a few years after hitting that milestone its business plummeted in the face of stiff competition from low-cost computer suppliers. Compaq’s problem was that it became one of many suppliers in a market where the platform providers controlled all the levers for product differentiation. If you see the cloud as just a bunch of datacenters with the latest software running on the fastest Intel hardware, then it’s easy to conclude that the world hasn’t really changed that much and it’s business as usual. But if you instead look at it from a new frame of reference centered on the AWS platform API (and all the apps running on it), a very different world emerges where power has radically shifted to the cloud platform provider. Today AWS has over a million customers and is on a $7B annualized revenue run-rate growing 80% year-over-year. It is on a trajectory to overtake the $550B combined market cap of two giant platform companies (Intel and Microsoft).
The comparison between AWS and Intel/Microsoft is more than superficial. It’s not a stretch to call Amazon AWS the Intel of the cloud era. Indeed, the two businesses share some fundamental similarities. For one, Moore’s law is the engine that drives them both. Much like Intel, AWS continues to provide infrastructure services with higher performance at lower costs roughly on pace with Moore’s law.
Second, both businesses are capital intensive to start up but benefit from increasing economies of scale. Intel’s fabrication plants cost billions of dollars in upfront capital but once ramped up the marginal cost of producing an additional microprocessor rapidly diminishes with volume. Similarly, AWS needs significant upfront capital to buy large amounts of hardware equipment (servers, networking switches, hard disks, etc.) and build datacenters but the unit costs of the business come down as more and more customers drive higher utilization of its resources. In both markets, leaders have scale efficiencies that act as powerful barriers to entry.
The comparison to Microsoft of the 1980s and 90s runs just as deep. First, AWS provides the base software infrastructure to power cloud applications in the same way Microsoft Windows provided the operating system platform that powered all PC applications.
In addition, much like Microsoft did in the PC era, Amazon is aggressively moving up the stack by offering a range of software services from databases, developer tools, productivity apps, email and Business Intelligence built on its cloud infrastructure platform (see table below).
Amazon’s ambitions should come as no surprise. After all, this is the company that started selling books online and now runs an ecommerce megastore selling everything under the sun. What is interesting about this move up the stack is that it has created a dynamic where Amazon is competing with its own ecosystem partners (see table below).
It’s too early to tell how this dynamic will play out and if it will stifle innovation on the platform and make ecosystem partners move to competing cloud offerings from Google or Microsoft. AWS today has more than 10x the capacity of the top 14 cloud providers combined but we are in the early stages of a new market that is still largely untapped. Both Google and Microsoft are mounting a big effort to catch up with Amazon (see chart below), and certainly have the balance sheet to make it happen.
We could very well end up in a cloud provider oligopoly composed of Amazon, Microsoft and Google. It is unlikely there will be other players because market entry is going to require deep pockets, deep software and hardware expertise and experience operating businesses at scale.
Either way, if you’re a startup today building an enterprise product that runs in the cloud you’ve got your work cut out for you. I’ll make a few observations here:
Things may appear bleak to some. Platforms have a way of dominating and capturing disproportionate value, but history tells us platform power notwithstanding there will always be a market for companies with innovative new products. The Wintel platform was a monopoly but it yielded many large software companies that built on its platform (Lotus, Adobe, Intuit, Symantec and McAfee to name a few). There is no reason the cloud era should be any different.