Listen to the story.
Marco Werman: Massive amounts of data, massive servers to hold it. New words like zetabyte that describe the hulking successors to terabytes. So how much energy does it take to store all this big data? John Koomey studies energy policy at Stanford University. So this new Utah data center, John, that the NSA is building will require one and a half million gallons of water a day to keep the servers cool. What's known about the NSA data farms and how much energy they use?
John Koomey: So National Geographic reported that the new data center will use about 65 megawatts, and that's the equivalent on average of how much electricity that would be drawn by a town of about 65,000 people. So when you get your electric bill it's measured in kilowatt hours. So if you have 100 watt bulb that goes for 10 hours, that's one kilowatt hour. So a 65 megawatt power plant or a 65 megawatt data center will use in one hour 65 megawatt hours.
Werman: Wow, okay so I assume you think that's a lot.
Koomey: Well, it's as much as some large industrial plants. It's as much as a, you know, a midsize city. But you have to put this in context because those facilities allow us to do lots of other things in this society that we couldn't do otherwise.
Werman: Sure, so how does that compare to say, commercial data farms like Google's well known data farms that are popping up everywhere?
Koomey: Well, so Google, Apple, Facebook, the cloud computing providers, they have data centers that are also in the tens of megawatts. 65 megawatts would be a very large one.
Werman: And how much are we each individually responsible for using energy when we say we just Google something on a smart phone? Aren't we forcing a server somewhere to spin up in order to give us the answers?
Koomey: We are, but it's a trivial amount of electricity for each Google search. So the most important thing to keep in mind is that the data centers that we hear the most about from Google, and Facebook, and Apple, and Microsoft, those actually lead the industry in efficiency. And they're not the main issue. They also have invested a lot in renewable power. There's a set of data centers in companies whose primary business is not computing, and that would be, you know, the manufacturing company or retail company, and those are the data centers that use, by far, the most electricity and are the least efficient. But you don't hear about them. They don't get pressure to improve the efficiency of their facilities.
Werman: And just for comparison, what industry uses even more energy than data centers?
Koomey: So people sometimes compare data center emissions to the emissions from airline industry. And they're the ones that I think that we can do a whole lot better in terms of efficiency.
Werman: How does this compare to years ago when we used to use reams of paper, photocopying things? Is there an environmental benefit as well to taking a lot of paper out of the mix?
Koomey: Yeah, so there's a bunch of different ways that you can improve environmental performance using information technology. One way is to substitute, you know, information for material goods. The examples I like to point to are driving to the store to pick up a CD instead of we download it. And we've studied that and the best case for buying the CD versus the worst case for the downloads, it was still a 40% savings. So moving bits instead of moving atoms is generally a better way to improve the environment.
Werman: John Koomey, a research fellow at Stanford's Steyer Center for Energy Policy and Finance. Thanks for your time.
Koomey: My pleasure.