The mega-clouds are coming for your data center

In the future, enterprises may be as unlikely to stand up their own data centers as build their own power plants

The mega-clouds are coming for your data center

There's no longer a question of whether public cloud will win out over private cloud. The only remaining question is how any cloud vendor can hope to compete with the "relentless economies of scale" of the mega-clouds.

A year ago RedMonk analyst Stephen O'Grady dubbed such economies as "daunting" for traditional data center vendors, in one of the most polite understatements ever uttered. After all, a year later we're seeing mega-cloud vendors -- Amazon Web Services, Google, and Microsoft -- drive server costs to lows impossible for any enterprise data center to achieve.

Oh, and they're doing it while retaining significant margins for themselves -- all of which means we may be on the verge of seeing a complete changing of the guard in enterprise IT that will play out over the next decade.

How low can you go?

The most obvious reason that the mega-cloud vendors can charge less for more computing, storage, and other resources is that they pay less per server than enterprises running private data centers, thanks to the huge volumes of hardware they buy. As O'Grady describes it:

Their variable costs decrease due to their ability to purchase in larger quantities; their fixed costs are amortized over a higher volume customer base; their relative efficiency can increase as scale drives automation and improved processes; their ability to attract and retain talent increases in proportion to the difficulty of the technical challenges imposed; and so on.

Consider the relative size of these data centers. According to a 2012 report in Wired, Google was the fifth largest server maker in the world. Facebook wasn't far behind. Today the top five would include AWS and probably Microsoft.

No wonder, given the relative size of those mega-clouds' data centers. According to a source familiar with AWS operations, a typical AWS data center includes 50,000 hosts or 1,000 racks. A traditional enterprise data center, by contrast, has a few thousand hosts, yet incurs similar physical plant and power consumption costs -- except they're not the same costs.

Call it vertical integration or whatever you want, but the mega-cloud vendors are building everything related to their data centers.

Google, for its part, details how its roll-your-own approach results in tremendous efficiencies. When I asked one Googler familiar with the company's cloud infrastructure whether he thought an enterprise could operate anywhere near the efficiency of Google, he responded that the idea was "hysterical." As one example, he pointed out that "we take power direct from transmission lines and build out our own substations."

The reason, as Timothy Prickett Morgan explains, is that "infrastructure definitely matters [because] it is absolutely a competitive advantage." In short, no data center goes to the lengths of the mega-clouds to ruthlessly remove inefficiencies and drive greater performance.

The everyday low-cost leaders

This leaves those who make the case for private data centers with a steep hill to climb. The idea that they can operate at anywhere near the cost of public cloud mega-vendors is, increasingly, crazy talk.

As DataStax co-founder Matt Pfeil put it to me:

There's an implication ... that cloud should be cheaper than your own data center, which is huge long term. The best example I heard was: will it be cheaper for a company to run a data center and pay all their vendor margins or AWS/GOOG/MS to do it all at massive scale without those margins?

Which leaves us with governance, security, and the other purported advantages of private data centers. Such advantages, which once seemed ironclad, are fading fast. Already we're hearing executives like Capital One CIO Rob Alexander proclaim more trust in AWS security than in his company's own in-house security.

Developers, after all, are already moving there. Gartner analyst Thomas Bittman captures this with his analysis of application workload trends. As he explains, "New stuff [workloads] tends to go to the public cloud ... and new stuff is simply growing faster" than more traditional data center workloads:

Gartner cloud workloads Gartner

In other words, public cloud workloads are growing three times faster than private cloud workloads, and both are growing dramatically faster than traditional data centers. Why? Because developers, serving their lines of business, are chasing convenience and agility advantages that cloud delivers.

The fact that it is cost-competitive today, and cost-superior tomorrow, is gravy, but gravy that will fuel a complete changeover in the enterprise IT landscape over the next 10 years.