Then again, Google claims that 7g CO2 volume per search "is *many* times too high". It would be good if the assumptions and methodology for these claims are also published. But then again, I'm not sure that "grams of CO2 emitted per search" on the server side is necessarily the right way to evaluate this.
The (not insignificant) carbon footprint of large data centers was covered (somewhat misleadingly I think and again using Google as the example) in March last year by Ginger Strand in Harper's Magazine, which prompted me to respond with a letter to the editor (never published in the magazine due to I'm going to guess my tendency to outlandish word count that I just can't seem to suppress).
Here's that text (I think it's once again relevant given today's meme):
It's important to realize that the move to an information- and information-services-based economy isn't as environmentally benign relative to "heavy industry" as is commonly conceived, and Ginger Strand performs an important service in pointing this out in "Keyword: Evil".
But I do feel compelled to point out that some of the numbers used to quantify the impact of data centers like Google's could be a little misleading and ought to be put in perspective. In particular the notion of how much economically useful work is performed for every watt consumed is an important one when impact is assessed and weighed against societal benefit. After all, no human activity regardless of economic or political organization is without some impact on the environment.
I would suggest that a more meaningful metric isn't that - as the author puts it - "thousands of servers" spring into action on each search query, or that "tens of billions of CPU cycles" are allocated to that task. Since "frivolous" or "wasteful" energy consumption is the putative consequence here (and presumably in contrast to other economic sectors or even individual lifestyles taken in aggregate such as watching television or commuting long distances to work), a more useful metric would be to examine total data center CPU-seconds per useful task divided by idle CPU-seconds where a CPU is doing nothing and waiting for something to do. Okay, maybe you can argue that the "American Idle" - sic - query example isn't a "useful task", but never mind that much more subjective angle for now...
Given this so-called "utilization" metric, best practices for data center software and infrastructure engineering practice are intended to optimize things so that the CPUs and other hardware consuming power are utilized as much as possible. This means that these CPUs don't sit idle very long, and can be quickly and efficiently reallocated from sub-second to sub-second as they complete their (subdivided) work for any given search query.
It is true however that these data centers are sized to peak loads, which are no doubt significantly higher than the average ones and so this reduces utilization. But even at non-peak periods these idle CPUs can do something else (like serving Gmail or Google Docs or YouTube, which don't in general share the same peaks) or can be put into a lower-power ready-to-use standby state.
Such strategies for energy consumption management and reduction - which I'd expect is taken into account in the quoted 103 megawatt figure - should also be contrasted with what happens on the receiving end of these queries. Similar CPUs in end-user PCs (and home routers and other microprocessor-intensive accessories) essentially sit idle most of the time waiting for the consumer to click the next web link (or even sit down at the keyboard). Multiply this times the hundreds of millions of internet-connected PCs in North America alone (even when in low-power standby mode which can run around 5 watts) and the data center power consumption pales in comparison.
To conclude, it's absolutely correct to point out that data centers such as Google's have a significant energy consumption impact, on par with aluminum smelting, which is traditionally considered to be a "heavy-industrial" benchmark for point electricity consumption. And it's also worthwhile to point out how an unregulated global economy with sharp differentials between states and countries in terms of environmental protection or rob-Peter-to-pay-Paul subsidization provides incentives to chase the lowest cost and often environmentally harmful arrangements. But to mention this without contrast to other aspects of environmentally harmful energy waste both within the information services sector as well as outside of it is somewhat misleading and might imply that merely eliminating the concentration in providers of information services would significantly reduce total energy consumption and waste, which I don't think would be the case at all.