The meteoric rise of both the public cloud and SaaS have brought along a strong preference for OpEx vs. CapEx. To recap: OpEx means you stop paying for a thing up front, and instead just pay as you go. If you’ve bought almost any business software lately you know the drill: you walk away with a monthly or annual subscription, rather than a DVD-ROM and a permanent or volume license.
But the funny thing about business trends is the frequency with which they simply turn upside down and make the conventional wisdom obsolete.
Recently, we have started seeing interest in getting out of pay as you go (rather unimaginatively often shortened as PAYGO) as a model, and moving back toward making upfront purchases then holding on for the ride as capital items get amortized.
Why? It’s all about economies of scale.
Imagine, if you will, that you are able to rent an office building for $10 a square foot, then rent out the space for $15 a square foot. Seems like a decent deal at 50% margin; but of course you’re also on the hook for servicing the customers, the space, and so forth. You’ll get a certain amount of relief as you share janitorial services across the space, of course, but your economic ceiling is stuck at 50%.
Now imagine that you purchase that whole building for $10M and rent out the space for $15M. Your debt payment may cut into profits for a few years, but at some point you’re paid off – and every year’s worth of rent thereafter is essentially all profit.
The first scenario puts an artificial boundary on both risk and reward: you’re on the hook for a fixed amount of rental cost, and can generate revenues only up to 150% of your outlay. You know how much you can lose, and how much you can gain. By contrast, in the second scenario, neither risk nor reward is bounded: with ownership comes risk (finding asbestos in the walls, say), as well as unlimited potential (raise rental prices and increase the profit curve).
This basic model applies to many cloud services – and to no small degree explains why so many companies are able to pop up – their growth is scaled with provisioned services.
If you were to decide to fire up a new streaming video service, say, that showed only the oeuvre of, say, Nicolas Cage, you’d want to have a fairly clear limit on your risk: maybe millions of people will sign up, but then again maybe they won’t. In order to be sure you’ve maximized the opportunity, though, you’ll need a rock solid infrastructure to ensure your early adopters get everything they expect: quick video start times, low re-buffering ratios, and excellent picture resolution. It doesn’t make sense to build that all out anew: you’re best off popping storage onto a cloud, maybe outsourcing CMS and encoding to an Online Video Platform (OVP), and delegating delivery to a global content delivery network (CDN). In this way you can have a world-class service, without having to pony up for servers, encoders, points of presence (POPs), load balancers, and all the other myriad elements necessary to compete.
In the first few months, this would be great – your financial risk is relatively low as you target your demand generation at the self-proclaimed “total Cage-heads”. But as you reach a wider and wider audience, and start to build a real revenue stream, you realize: the ongoing cost of all those outsourced, opex-based, services is flattening the curve that could bring you to profitability. By contrast, spinning up a set of machines to store, compute, and deliver your content could set a relatively fixed cost that, as you add viewers, would allow you to realize economies of scale and unbound profit.
We know that this is a real business consideration because Netflix already did it. Actually, they did it some time ago: while they do much (if not most) of their computation through cloud services, they decided in 2012 to move away from commercials CDNs in favor of their own Open Connect, and announced in 2016 that all its content delivery needs were now covered by their own network. Not only did this reduce their monthly opex bill, it also gave them control over the technology they used to guarantee an excellent quality of experience (QoE) for their users.
So for businesses nearing this op v. cap inflection point, the time really has arrived to put pencil to paper and calculate the cost of going it alone. The technology is relatively easy to acquire and manage, from server machines, to local load balancers and cache servers, and on up to global server load balancers. You can see a little bit more about how to actually build your own CDN here.
Opex solutions are absolutely indispensable in getting new services off the starting line; but it’s always worth keeping an eye on the economics, because with a large enough audience going it alone is the way to go.