An ‘open letter’ to data center investors
CEO - Romonet
October 08, 2014
I've had too many people ask me to publish this over the past 9 months to not finally take action.
What I outline here is nothing more than the predicable outcome of a simple economic maturing of an industry that has enjoyed explosive growth over the last decade.
Those that know me (and my business) know that I’m quietly assisting many of the smarter businesses understand both the macro economic effects in terms of how those impact their businesses, and more importantly, how they can adapt to survive in the coming years as the data center and technology industry go through some major technology driven market corrections.
This article looks primarily at hosting data centers; some of what I say here applies to the enterprise data center and true Cloud data centers too but I’ll deal with those more specifically in a future article.
Let's start by dispelling some of the most common myths in this industry.
Here we go….
1) Data Centers can’t be valued as a property business (any more)
Despite what you might believe or been told, for data center providers the dominant cost (from a TCO perspective) in these businesses is not the real-estate.
It’s actually the capital for the electrical and mechanical equipment and the operational cost for staffing, maintenance, and energy, much of which is too often wasted through poor design and operation.
What’s worse is that very little of this equipment lasts for the expected 20 year real-estate reinvestment horizon. Much of it needs replacing between 8 to10 years. Even if you got the build 100% right we’ve seen plenty of changes in market demand, so re-investments are necessary to update the asset.
People will tell you that they are ‘sweating their asset’ which is fine up to the point that the customers’ contracts come up for renewal, and you now have an uncompetitive data center that’s also not fit for purpose compared to all the more recently designed empty new space.
During the decade of ‘sweating your asset’ the customer’s technology (computing) refresh cycle was only 3 years. So within just one decade their requirements are now three generations ahead of what they were when they first signed their contract. This rate of IT refresh is also increasing; in a few years they may be on very low power servers or asking for direct liquid cooling.
If your data centers are approaching 10 years old and have not had a major reinvestment you are in for a nasty surprise one way or other!
2) Data Centers are not "The Cloud"
The thing I hear all the time is ‘data centers are where the Cloud lives’. Yes, cloud is delivered from data centers but don’t for a moment make the mistake of thinking that this means your data center is now magically worth twice what it was as a plain old ‘hotel for computers’ or even suitable for use as a cloud data center (at what you can afford to charge for it).
Many data center operators have managed to get vastly inflated valuations by dressing themselves up as a ‘Cloud’ business. Some other businesses have mistakenly paid well the over the odds to acquire them.
For the avoidance of doubt, Cloud is the commoditization of IT services, nothing more. Commoditization is nothing new; take a look at the airline or shopping mall businesses for a taste of the future margins and problems for data centers.
3) The Cloud is not your friend it's your competition!
Very specifically I’m talking about hosting data center businesses here but some of the facts apply to the CIO within an enterprise operator too.
Every time someone virtualizes a software application (disconnects the sticky link between software and physical hardware) it becomes trivial to move it. That might be move it to new hardware in the same data center or move it out into a pure Cloud data center.
Discussions of Public vs. Private vs. Hybrid clouds don’t matter here; the underlying economic issue for most data center investors is that the real Cloud providers (Google, Amazon, Microsoft, etc.) have a set of major advantages:
People scale - they can develop, maintain and operate IT management platforms that would eat your margin if you tried to implement them.
Physical scale, they buy everything from land to ‘no-brand’ servers and network devices in huge volumes
Cheap data centers - they are not saddled with ageing “Enterprise Tier III” data centers in expensive locations, they build the cheap cloud data centers out in the wheat belt. It matters which country a cloud data center is in, but not which state or city.
4) Focus on revenue and growth and it will all be fine
This is scarily reminiscent of the strategies that many companies followed during the last data center boom; the survivors of that market correction were able to acquire some valuable (and some truly awful) assets for very little money when the bubble burst.
5) Growth in data centers is a function of the explosive growth in technology, mobile communications, globalization, etc.
That as a standalone statement is true, but, if you are a colo only data center your share of that market growth is slowing down and from a real estate perspective (total sq. ft.) will eventually start to contract.
Why? Because the huge demand for colocation space (as it’s called) is a transitional phase in the market evolution, between the old model of enterprises owning everything from the IT staff down to the building, and the new model of buying services from a cloud provider for users to access with their mobile devices. When corporate CFOs decided that owning the data center was a cost risk they could eliminate, but CIOs could still justify owning the servers, software and staff to manage them, we needed somewhere to put all that expensive tin. The same financial decisions which started to kill the enterprise data center now apply to owning the servers, just look at the “Tier 1” server vendors.
6) My product and my data center business are unique and thus immune to commoditization
This is a favorite of those in denial! It used to be one of the standard BS bingo phrases, especially on investor calls. For those still in denial, sorry, but for the vast majority of you, the only way your data center is different to anyone else’s is what re-investments are hiding in the closet and how much your margin will drop as those around you get more efficient and your customers wise up to the economic impact of the maturing of this market.
The decision about where to host your compute will soon be made by a resource broker or a market which has no interest in what your data center looks like, how nice your staff are or anything much else beyond cost, jurisdiction and service level.
There will be some exceptions to this, there always are. For example some operators have high frequency trading exchanges within their data centers as they are prepared to pay well over market rate to be a few milliseconds closer.
7) Customers will stay despite the costs because moving is hard and costly
Yes, moving physical hardware out of one data center and into another is costly, hard and risky. Moving applications (software) from one homogeneous Cloud stack to another (of the same flavor) is not, it can be done literally a few clicks of a mouse.
In fact, it’s entirely possible for a company to move all their virtualized applications over the network and then have an asset disposal company remove their old compute hardware as it was probably end of life anyway (3 years lifetime remember!) and let then just let the hosting contract expire.
8) For most customers renting space in a data center is cheaper than owning a data center
For many customers we, and others, have gone through the economics and the simple fact is that somewhere between 500kW and 1MW of capacity it becomes substantially cheaper (from a TCO perspective) to build your own data center than to rent colo.
This change has come about due to the commoditization of everything from the mechanical and electrical plant through the design and construction.
The colo data centers do have some cost benefits due to scale but these are smaller than you think (none of what they buy is high margin so it can’t be discounted much) and they have to build a more expensive data center in a more expensive location than you do or the cloud operator already does.
As for not wanting to own (then lease it) and operate (then outsource the operations) your own data center once you’re close to 1MW there’s almost no reason not to do so. You can rent the staff and skills from the same people the colo operators do.
Oh, and if the reason is time-to-market and the long lead time to have your own site(s) built then think again, if the people you’re talking to are beyond the 6-9 months timeframe then stop talking to them. If you need shorter term capacity than this then rent the Infrastructure in somebody else’s cloud until yours is online.