IaaS Is Currently Quite Profitable
Several months ago, Amazon Web Services announced revenue of $2.88 billion in the second quarter, and earnings of around $700 million. From a revenue perspective, they’re up 58% year-over-year. By my back of the envelope calculations, they’re also pulling in roughly 25% margins (exactly 23.5% in Q1 of this year), making the AWS easily the most profitable piece of Amazon. Were AWS it’s own company, it would also be one of the most profitable in the world. So where is the “commoditization of Infrastructure as a Service (IaaS)” that’s apparently inevitable?
I frequently hear the argument that “infrastructure is rapidly becoming a low-margin business.” I’ve actually asserted this on several occasions myself. Heck, AWS CEO Andy Jassy made the same assertion back in 2010, stating “there is no doubt in my mind that we will be a high volume, low margin business”. So why, 6 years later (an eternity in technology) is AWS still making money hand-over-fist? The prediction of inevitable commoditization is based upon the assumption that IaaS will perform in a manner similar to that of comparable industries, with the most direct comparison being hardware makers coupled with historical economic data of profitability in maturing industries. The assertion seems perfectly logical and defensible…but what if it’s wrong?In the afore-linked WSJ article the same analyst compares IaaS to “cement poured into the foundation of a building” and goes on to say the “real value is built on top at the platform and software layers.” While I don’t dispute the considerable value at the platform (PaaS) and software (SaaS) layers, Infrastructure as a Service (IaaS) too has the potential to drive considerable profitability longer term. Here’s why:
The benefits of scale in the cloud game are hard to overstate. Not only do traditional economies of scale hold true in terms of driving costs down, but scale is also a competitive differentiator that can be used to drive prices up as well as further customer adoption. For example, you simply can’t run massive workloads on Rackspace the way you expect to on AWS or Microsoft. The scale simply isn’t there. Not only does Amazon have far greater purchasing power on the underlying hardware, but their ability to scale provides a significant competitive advantage – one that can protect pricing power longer term.
2. Oligopoly Pricing
To dust off another economics 101 term, the cloud is rapidly becoming an oligopoly by virtue of rising barriers to entry. Scale is certainly one of these barriers as trying to construct a series of globally distributed datacenter operations alone would cost billions of dollars. More importantly, however, is the intellectual property associated with a differentiated IaaS. Ask anyone who has been in the IT industry for a while how easy it is to offer on-demand, virtualized infrastructure at 99.9%+ availability at the click of button. Most will share with you war stories of trying (and failing miserably) to accomplish this panacea of virtualization. It takes a lot of time, money, and know how that the “Big 3” (AWS, Microsoft, Google) already posses – in abundance. “But the airlines are an oligopoly, and they aren’t very profitable”, you might counter. True enough, but cloud providers aren’t airlines and what they’re providing is far more differentiated than a cramped seat, mediocre snacks, and abysmal service.
3. Vertical Offerings and Integration
The low-margin argument makes a further faulty assumption that IaaS, as any other kind of infrastructure, is effectively a commodity. While true in the very simplest sense (a virtual server by any other name is still just a server), this comparison breaks down very rapidly once you start looking at the immediate periphery beyond core storage and compute. The options for scaling up that capacity on-demand, the API’s for integrating it into your product or service, and the online marketplace for buying and selling unused compute (known in AWS parlance as Spot Pricing) are all proprietary offerings to each cloud vendor. Brilliantly, most of the larger cloud providers don’t charge a thing for these capabilities and encourage you to use them liberally and freely in the design of your application.
4. Cost of Being Vendor Agnostic
“But can’t you design systems that use 3rd-party tools to accomplish these tasks so that you aren’t subject to vendor lock-in or just avoid using these proprietary tools altogether?” Sure, you could absolutely do either of those. Trouble is, your competitors (especially small, nimble start-ups on tight budgets) may not posses the same proclivities for supplier diversity. Going back to the airline example, ask any of the major carriers how hard it is to compete with Southwest Airlines. The company that long ago not only made Boeing its sole supplier of aircraft, but the 737 the only airframe it flies. Sure, there are some disadvantages to this approach, but from an infrastructure and support perspective, it massively simplifies and speeds operations and helps them keep costs at a bare minimum. Leveraging the proprietary offerings of a cloud vendor is a very effective means of streamlining IT operations. While building cloud-vendor-agnostic solutions in the cloud can be done, the opportunity cost compared with the limited potential for savings is simply too high. To compound the problem, the scale that the Big 3 occupy means that a newcomer beating them on price is going to be unlikely over the mid to long term.
A Faulty Analogy
So the “concrete foundation upon which more valuable offerings can be built” seems, after all, to be a faulty analogy. Cloud infrastructure would better be thought of as the entire infrastructure of the house that’s waiting for you to do the finishing work. You can use the foundation, electrical, and plumbing already in place and get your “house” up in a couple weeks (and oh, by the way, the infrastructure provider is going to maintain and upgrade all this stuff for you while you’re in the house). Alternatively, you can ignore all of it and build your own “portable house” that you can easily move from foundation to foundation to save a few bucks every time a new infrastructure provider starts a price war. If you’ve ever tried to build a house (or design perfectly platform agnostic software), the choice is pretty obvious. The cloud providers are working hard to make it even more of a no-brainer and they aren’t doing it in pursuit of a low-margin, commodity business, despite their 6 year old statements to the contrary 😉 Perhaps even Andy is surprised at the margins his business is pulling in.
Footnote: In doing research for this blog post I came across a post by Gartner analyst Lydia Leong (Does Cloud Infrastructure Become Commoditized) that discusses similar themes. While these have only recently occurred to me, she wrote her post back in 2009 and it holds true today.