The matching of the capacity to the business requirements, as we encounter more and more price pressure brought on by the cloud guys, is going to have to change the way that CIOs look at their data centers and IT assets. If they don’t, they are just going to get replaced by cloud operators.

But what about what is already out there in the market?

Enterprise data centers

For many of these facilities most of their capital has been written off – the monies were invested some time ago, it’s been amortized and the tax has been recovered on it. It is an owned asset but has got very poor re-sale value. What this actually means is that rather than being an expensive obligation, you have what could be very cheap capacity, even if it is not low PUE. A relatively small re-investment capital can yield very cost effective capacity (certainly much cheaper than you could ever build even if your new build had a PUE of 1).

And for levels above 1MW you can build your own new capacity at a competitive TCO or have Compass or Digital Realty build it, or build your own modular building. Once you get through that capacity threshold you can be very competitive. There are also many options in the marketplace available for lease purchase and lease-back. So enterprise DCs are anything but dead; the capital sunk in them can make them a very effective asset and commoditisation means that you can build new capacity at very competitive TCO, there is no ‘special sauce’ reserved for the experts any more.

Colo data centers

These are more interesting as there are more price pressures in this market. Customers can change provider much more easily than they used to thanks to the portability of applications between IT devices and the continuing shortening lifetime of those IT devices. Opex guarantees (metered power times PUE as billing) are rapidly becoming the norm in many markets. Some of the colo demand is directly migrating to cloud demand for exactly the same reasons that people didn’t want to own the data center capacity – they don’t want to own the IT capacity either. Some colos are responding by placing cloud capacity in their colo facilities to try and capture some of that migration of customers. There is very much a reduced need to have IT equipment in the same place; the old ideas of secure private networks etc. don’t really work any more and were just as likely to call an application or service in another continent as from another rack. We also have rapidly falling market prices (and the margins that go with that). Importantly, that’s being driven by better informed customer renegotiations – even those customers that don’t necessarily want to move know that they can push on price. A really key thing here is that for any colo operator with a site that is bearing high costs (e.g. poor PUE or poor capacity utilisation) these costs will eventually come out of the colo operator’s margin as they are forced to match the market.

Wholesale Colo data centers

Similar issues to colo but there is a larger population of old, high PUE, low power density sites that really you wouldn’t choose to lease space in now against the competition that is out there. Again, it is easier for customers to migrate out of these; no more weekend moves and truck hire to shift equipment. As above, informed re-negotiation by the customer base is likely to push down price. Many of these facilities will require substantial re-investment to be competitive again, particularly in European countries were attention has started to focus on energy efficiency or carbon.

Cloud data centers

The market leaders are all following a common strategy: cheap buildings in cheap locations

They obtain a very low utility energy cost by locating to an area with a low power cost (Iowa, Norway, etc.) and then negotiating that price further downwards. As part of this cost negotiation they typically achieve low or no taxes by promising a large investment in this town you’ve never heard of. Cloud providers are building cheap data centers by creating cheap buildings (no fancy foyers here as you may find in some enterprise facilities) – it’s more like a Walmart. They use minimum M&E redundancy, run their UPS in bypass mode and have loose environmental controls. This keeps their capital and operational costs way down, and at the scale this is being done (hundreds of MWs of capacity at a time) means that it is very hard to build cheaper than these operators or even get down to their price.

Conclusion

To conclude – don’t dismiss that old enterprise facility as a lost cause – a relatively small investment could produce a useful amount of financially efficient capacity. Meanwhile colo operators, both small and large are finding the relentless emphasis on lowest cost by cloud providers eating into their customer base and their margins.