If you follow the buzz around all things cloud, you've seen the "scale" argument in different forms: cloud operators will win in the long term simply because they have economies of scale.
And then the author will invariably point at compute, or power, or some other commoditized economic input as evidence of the cloud model's inherent advantage over traditional IT.
Well, I beg to differ.
Yes, I think there's an inherent 'scale' advantage that service providers have in their favor. But it has little to do with things like compute, storage, data centers, etc.
It has to do with people and process.
And then the author will invariably point at compute, or power, or some other commoditized economic input as evidence of the cloud model's inherent advantage over traditional IT.
Well, I beg to differ.
Yes, I think there's an inherent 'scale' advantage that service providers have in their favor. But it has little to do with things like compute, storage, data centers, etc.
It has to do with people and process.
Was Nicholas Carr Right?
In his seminal book, "The Big Switch", Nicholas Carr uses the evolution of the power generation business to create a historical parallel with how IT will be delivered in the future.
To paraphrase, at the beginning of the 20th century, if you were in the manufacturing business (the economic engine of the day), you generated your own power. He presents an interesting history on how the model shifted from do-it-yourself to our current use-the-electric-utility model.
Later in the book, he tries to use some of the economies-of-scale arguments to presage what will likely happen in IT. He uses familiar examples from Google and Amazon to underpin his story line.
While it's all directionally correct, I think the real economies of scale are turning out to lie elsewhere. And, if you agree with my line of thinking, it might change how you think about your SP business model, and its long-term economic advantages over alternatives.
Declining Economies of Physical Scale
I've seen all sorts of IT pricing curves around declining unit costs as scale increases -- for CPUs and servers, for networks and bandwidth, for storage, for data centers, for power and cooling, etc. -- all the "hard costs" associated with data center operations at scale.
While it's generally true that unit costs decline as scale increases, most of the big savings are at the front of the curve, with progressively less incrementaly savings as the curve moves outwards.
Put differently, as you move up the this particular set of scale curves, the marginal savings tends to decrease over time -- you get bigger and bigger, but are proportionally rewarded less and less for it in most cases.
Another way of looking at things is from the buyer's perspective. Two service providers offer largely the same infrastructure capabilities. One of them is 1% or 2% cheaper. Does the cheaper provider always win the business, or do other factors come into play?
To make matters interesting, many enterprise IT organizations can operated at significant scale: thousands of servers, multiple modern data centers, bulk buys for network bandwidth, etc. So if you're going to compete with internal alternatives at this sort of scale, you'll have to be very, very large indeed!
Clearly, SPs need some decent scale just to be cost-effective, but its attractiveness as an underlying competitive principle declines over time.
Declining Economies of Geographic Scale
One could make a related argument around geographical scale, i.e. multiple data centers around the globe.
The demand side is easy to see: as more companies do more business in more and more countries around the world, there is increasing demand for sole-source service providers who can stand up infrastructure and/or applications where business is getting done.
However, start-up costs for outside players are enormous compared to in-country players. Let's say you had to stand up a data center in, say, Vietnam to take care of your customers. Compare your cost and effort of doing so compared to, say, the Vietnamese telco or some similar player.
To say that they've got a "home field advantage" would be a bit of an understatement.
Far more attractive for a service provider with global aspirations to simply sub-contract for the required services in the geography of interest, and see how it goes. If business grows, perhaps reconsider the proposition.Towards People Scale
Most IT leaders won't come right out and say it, but the expense of IT that really bothers them is people cost. Now, that's not the sort of thing that an organizational leader would usually say in a public forum, but -- over a nice bottle of wine -- it usually comes out.
Let's face it: people are expensive. They're hard to find, hard to train, hard to keep productive, hard to retain, and tend to get their nose out of joint on a periodic basis. Figure out how to deliver a better level of IT services with fewer people, and that's a win.
So, let's make a quick inventory of skills required to deliver modern, enterprise-class IT these days. Infrastructure people: server, storage, network, hypervisor. Software people: databases, middleware, packaged applications, internally developed applications. User-facing people: desktops, apps, consulting, etc. Security and compliance people.
The list keeps going on, and on, and on.
Do you think the required portfolio is getting bigger, or smaller? If IT is all about delivering new value to the business, that means you'll always need new skills that you didn't have before.
Put differently, the natural tendency is for the IT roster to balloon over time. There's always demand for a new skill set, or more depth in an existing one. It was true twenty years ago, and it's just as true today.
One historical approach has been the increased use of contractors. Well, that makes it a bit better -- they don't all have to be full-time, do they? -- but the underlying costs for people haven't changed much, it's just a bit more flexible to consume.
Or perhaps outsourcing, or off-shoring? That's essentially simple cost shifting and cost arbitrage -- and not fundamentally changing the cost curves in a fundamental or sustainable manner.
Can service providers provide a more economic solution to "people scaling"? I think the answer is most decidedly "yes", and will explain my reasoning a bit further on.Towards Process Scale
The history of our economy shows one overriding fact: most processes are far more efficient when they are designed to scale.
That's true for manufacturing, telecommunications, defense, etc. Conversely, running efficient and/or automated processes on a small scale is somewhat useless, if you think about it.
One clear sustainable advantage that service providers have is this notion of "process scale". If they do a good job of standardization and automation, their processes can scale very, very large and use very, very few people -- in addition to being more robust and higher quality.
Achieving this level of process efficiency turns out to be incredibly difficult for most enterprise IT organization. They either don't have the motivation to invest in it, or -- when they do -- find it torn to pieces by the tidal forces of "unique business requirements" that don't fit the standardized model.
The Real SP Advantage: People and Process Scale
You can see where I'm going with this: I believe that many SPs can carve off a sustainable competitive advantage vs. traditional IT approaches by focusing on scaling their people and process.
One crack database team can do it all: requirements, architecture, implementation, performance tuning, migrations, etc. etc.
That same team of a few people can cover the needs of dozens -- if not hundreds -- of clients.
Ditto for the other infrastructure disciplines: storage, networking, servers, security, etc.
When it comes to specific application environments (e.g. SAP, Microsoft, etc.) the same scaling effect can come into play.
Put together a small team of, say, SAP experts -- and their expertise can be applied to dozens -- or hundreds -- of clients.It's cheaper and better than hiring full time staff. Arguably cheaper and better than bringing in the occasional contractor or consultant, too. Or moving it offshore or outsourced.
These teams of experts also have the incentive and luxury of designing and implementing repeatable processes that scale efficiently.
Ideally, there's one way of doing things (or, at least, a few ways of doing things!) which lower costs, improve quality, increase responsiveness, etc. Moreover, the SP has a vested interest in creating and sustaining these scalable processes, simply because they don't have to be everything for everyone -- like most enterprise IT organizations have to do!
The Bottom Line
Do I agree with Nicholas Carr and others that service providers (or cloud operators, if you prefer) have a sustainable economic advantage?
I do.
But I think it will come less from scaling the hard costs (hardware, power, data centers, etc.) and more from models that scale the ultimately valuable commodities: people and process.
We'll see, won't we?
Comments