Its always hard to pick exactly where to start in a conversation like this especially since this entire process really represents a changing life-cycle. Its more of a circular spiral that moves out (or evolves) as new data is introduced than a traditional life-cycle because new data can fundamentally shift the technology or approach. That being said I thought I would start our conversations at a logical starting point. Where does one place your infrastructure? Even in its embryonic “idea phase” the intersection of government and technology begins its delicate dance to a significant degree. These decisions will ultimately have an impact on more than just where the Capital investments a company decides to make are located. It has affects on the products and services they offer, and as I propose, an impact ultimately on the customers that use the services at those locations.
As I think back to the early days of building out a global infrastructure, the Site Selection phase started at a very interesting place. In some ways we approached it with a level of sophistication that has still to be matched today and in other ways, we were children playing a game whose rules had not yet been defined.
I remember sitting across numerous tables with government officials talking about making an investment (largely just land purchase decisions) in their local community. Our Site Selection methodology had brought us to these areas. A Site Selection process which continued to evolve as we got smarter, and as we started to truly understand the dynamics of the system were being introduced to. In these meetings we always sat stealthily behind a third party real estate partner. We never divulged who we were, nor were they allowed to ask us that directly. We would pepper them with questions, and they in turn would return the favor. It was all cloak and dagger with the Real Estate entity taking all action items to follow up with both parties.
Invariably during these early days - these locales would always walk away with the firm belief that we were a bank or financial institution. When they delved into our financial viability (for things like power loads, commitment to capital build-out etc.) we always stated that any capital commitments and longer term operational cost commitments were not a problem. In large part the cloak and dagger aspect was to keep land costs down (as we matured, we discovered this was quite literally the last thing we needed to worry about) as we feared that once our name became attached to the deal our costs would go up. These were the early days of seeding global infrastructure and it was not just us. I still laugh at the fact that one of our competitors bound a locality up so much in secrecy – that the community referred to the data center as Voldemort – He who shall not be named, in deference to the Harry Potter book series.
This of course was not the only criteria that we used. We had over 56 by the time I left that particular effort with various levels of importance and weighting. Some Internet companies today use less, some about the same, and some don’t use any, they ride on the backs of others who have trail-blazed a certain market or locale. I have long called this effect Data Center Clustering. The rewards for being first mover are big, less so if you follow them ultimately still positive.
If you think about most of the criteria used to find a location it almost always focuses on the current conditions, with some acknowledge in some of the criteria of the look forward. This is true for example when looking at power costs. Power costs today are important to siting a data center, but so is understanding the generation mix of that power, the corresponding price volatility, and modeling that ahead to predict (as best as possible) longer term power costs.
What many miss is understanding the more subtle political layer that occurs once a data center has been placed or a cluster has developed. Specifically that the political and regulatory landscape can change very quickly (in relationship to the life of a data center facility which is typically measured in 20, 30, or 40 year lifetimes). It’s a risk that places a large amount of capital assets potentially in play and vulnerable to these kinds of changes. Its something that is very hard to plan or model against. That being said there are indicators and clues that one can use to at least play risk factors against or as some are doing – ensuring that the technology they deploy limits their exposure. In cloud environments the question remains open – how liable are companies using cloud infrastructure in these facilities at risk? We will explore this a little later.
That’s not to say that this process is all downside either. As we matured in our approach, we came to realize that the governments (local or otherwise) were strongly incented to work with us on getting us a great deal and in fact competed over this kind of business. Soon you started to see the offers changing materially. It was little about the land or location and quickly evolved to what types of tax incentives, power deals, and other mechanisms could be put in play. You saw (and continue to see) deals structured around sales tax breaks, real estate and real estate tax deals, economic incentives around breaks in power rates, specialized rate structures for Internet and Cloud companies and the like. The goal here of course was to create the public equivalent of “golden handcuffs” for the Tech companies and try to marry them to particular region, state, or country. In many cases – all three. The benefits here are self apparent. But can they (or more specifically will they) be passed on in some way to small companies who make use of cloud infrastructure in these facilities? While definitely not part of the package deals done today – I could easily see site selection negotiations evolving to incent local adoption of cloud technology in these facilities or provisions being put in place tying adoption and hosting to tax breaks and other deal structures in the mid to longer timeframe for hosting and cloud companies.
There is still a learning curve out there as most governments mistakenly try and tie these investments with jobs creation. Data Centers, Operations, and the like represents the cost of goods sold (COGS) to the cloud business. Therefore there is a constant drive towards efficiency and reduction of the highest cost components to deliver those products and services. Generally speaking, people, are the primary targets in these environments. Driving automation in these environments is job one for any global infrastructure player. One of the big drivers for us investing and developing a 100% lights-out data center at AOL was eliminating those kinds of costs. Those governments that generally highlight job creation targets over other types typically don’t get the site selection. After having commissioned an economic study done after a few of my previous big data center builds I can tell you that the value to a region or a state does not come from the up front jobs the data center employs. After a local radio stationed called into question the value of having such a facility in their backyard, we used a internationally recognized university to perform a third party “neutral” assessment of the economic benefits (sans direct people) and the numbers were telling. We had surrendered all construction costs and other related material to them, and they investigated over the course of a year through regional interviews and the like of what the direct impacts of a data center was on the local community, and the overall impacts by the addition. The results of that study are owned by a previous employer but I can tell you with certainty – these facilities can be beneficial to local regions.
No one likes constraints and as such you are beginning to see Technology companies use their primary weapon – technology – to mitigate their risks even in these scenarios. One cannot argue for example, that while container-based data centers offer some interesting benefits in terms of energy and cost efficiencies, there is a certain mobility to that kind of infrastructure that has never been available before. Historically, data centers are viewed as large capital anchors to a location. Once in place, hundreds of millions to billions (depending on the size of the company) of dollars of capital investment are tied to that region for its lifespan. Its as close to permanent in the Tech Industry as building a factory was during the industrial revolution.
In some ways Modularization of the data center industry is/can/will have the same effect as the shipping container did in manufacturing. All puns intended. If you are unaware of how the shipping container revolutionized the world, I would highly recommend the book “The Box” by Marc Levinson, it’s a quick read and very interesting if you read it through the lens of IT infrastructure and the parallels of modularization in the Data Center Industry at large.
It gives the infrastructure companies more exit options and mobility in the future than they would have had in the past under large capital build-outs. Its an insurance policy if you will for potential changes is legislation or regulation that might negatively impact the Technology companies over time. Just another move in the cat and mouse games that we will see evolving here over the next decade or so in terms of the interactions between governments and global infrastructure.
So what about the consumers of cloud services? How much of a concern should this represent for them? You don’t have to be a big infrastructure player to understand that there are potential risks in where your products and services live. Whether you are building a data center or hosting inside a real estate or co-location provider – these are issues that will affect you. Even in cases where you only use the cloud provisioning capabilities within your chosen provider – you will typically be given options of what region or area would you like you gear hosted in. Typically this is done for performance reasons – reaching your customers – but perhaps this information might cause you to think of the larger ramifications to your business. It might even drive requirements into the infrastructure providers to make this more transparent in the future.
These evolutions in the relationship between governments and Technology and the technology options available to them will continue to shape site selection policy for years to come. So too will it ultimately affect the those that use this infrastructure whether directly or indirectly remains to be seen. In the next paper we will explore the this interaction more deeply as it relates to the customers of cloud services and the risks and challenges specifically for them in this environment.
One thought on “The Cloud Cat and Mouse Papers–Site Selection Roulette and the Insurance Policies of Mobile infrastructure”
Great write up Mike. I laugh sometimes about the cloak and dagger junk, I often think people nowadays are just trying to play around. At the end of the day, it usually doesn’t accomplish much.