There is a bunch of ink out there about locating data centers to climate zones that are as cold as possible to save on the expense of cooling the center. Cooling is one of the key costs of operating a data center. Instead of doing that, why aren’t we moving data centers higher in the atmosphere (where it is colder) to gain the same (or better) cooling effect? It doesn’t strike me any more logistically difficult to either run cabling or build a power generating facility near the top of a peak than to locate a facility out in sticks of a northern climate when everything is added in. We have the mining capability and there is no real need to permanently disrupt the environment. Over tree line there isn’t that much and you would want to build underground anyway. You would probably have to have a piece of single track railway that ran to the top for equipment but that could be pretty well hidden. There are mountains located conveniently near to major population centers so the distance to connect to major data hubs could be relatively short. The cooling effect even at the peak of summer days would still be high; highs may only get into the 50’s on a 14er on summer days meaning a heat pump could still be extremely efficient in those conditions. Heck, you already have most of the infrastructure at most ski resorts to pull this off.
So why not locate data centers on top of mountains?
There is an interesting dimension to how a competitive advantage spreads through an industry, particularly when it has a technology component.
A typical way this expresses itself is in adoption patterns for a new idea. New idea X comes on the scene either through a firm internal to the industry, a firm entering the industry, or a vendor of some type recognizing an opportunity. For early adopters there is quite a bit of flexibility in implementing the new idea. False starts, dead ends, project failures or slow development while painful don’t place the firm at a disadvantage competitively. Follow on adopters, depending on how close they are to the forefront, gain knowledge from others who went before them at a tradeoff in flexibility. Those same false starts etc are less and less tolerable because the advantage of the idea has begun to accrue to the firms who went earlier. By the time you get to late adopters, that window of flexibility has shrunk down to where false starts and the like are completely intolerable because more and more firms are benefiting from the idea as it becomes standardized.
Depending on the idea and the industry, the pace at which adoption happens can vary considerably and firms don’t necessarily know what that will look like for any given idea. It is entirely possible to see everything from fast early adoption and rapid diffusion to slow variants of the same or a mixture. A further variable is how pronounced the advantage will be. Firms are faced with a risk matrix of adoption somewhat specific to their industry but always with the looming threat that if they get it wrong they may be late to the party on an idea with firm damaging prospects.
The questions for the firm become for any given idea X:
How is the firm coming to know about it?
Does the firm have any visibility to how seriously others are pursuing the idea?
How fast does this industry normally move to adopt?
If the idea works exactly as advertised, how big an advantage will this idea be for firms that adopt?
There is a perception that firms need to be focused on areas where other firms aren’t and that is a fair idea. That premise though seems best applied to the circumstance where the firm has active control and intelligence that it is matching competitor moves in the idea space so as not to end up competitively disadvantaged.
If you were given a blank sheet of paper and told to perform the function of the DMV what would you develop?
I’m thinking mostly of the function of licensing drivers to drive. Why do we even bother with paper tests? That probably made sense up to about 25 years ago. By now, given the state of technology, shouldn’t we be taking driving exams in simulators? There are good driving simulators that have been around for a while that could be adapted to this purpose. Put me in the simulator and have me drive for 30 minutes. Change the conditions to nighttime, snow, rain, fog, heavy traffic and n number of other random scenarios. Throw up different road signs and see how I react. Have me come up to a blind train crossing and see how I do on that. For beginning drivers, set up training simulators in various locations so they can log some number of hours before qualifying to take the live test. The beauty is that this can be some pretty long lived hardware. The basic function doesn’t need that much updating over time and even if it eventually does (self driving cars?) you drop in some new scenarios for that and you are good to go. The best part about this is you get good visibility about how I’m likely to perform at the real world function of driving. If I have developed bad habits over time, this is a good opportunity to correct them.
The other crazy thing at the DMV is entering all this data by hand on forms. That is pretty much the highest error rate way to capture information possible. Why can’t I fill stuff in on web forms, phone apps, etc that is just electronically submitted? For folks who may not have access to these means, you can put equipment at the DMV locations or direct them to the local library branch to fill out the forms. That can’t be any worse than either having forms mailed to you or going to the DMV to get the form. Realistically, I should be able to do all the back and forth remotely and then only have to show up for the bare minimum of tasks to complete the process.