How to Sell Newspapers and Promote Your Service

Someone at the office goaded me into this (which is the equivalent of "The Devil Made Me Do It", but very different from "My Dog Ate It"). But there is a meta-topic here which has been a pet peeve of mine.

It happens with great regularity and predictability that we root for the underdog, and should the underdog one day grow big and powerful and become an ... overdog (sp?), it gets to be our favorite whipping boy. Or in the case of Google, a target that satisfies two constraints at once: (a) its newsworthiness and (b) the looming specter of an evil empire. How in fact newsworthy and how in fact evil are of secondary importance.

So, when someone pinged me with a link to The UK Sunday Times' "Revealed: the environmental impact of Google searches", my BS antenna went up. Long story short, a Dr. Wissner-Gross (Physics PhD 2007, Harvard) of http://www.co2stats.com/ appeared to have been ghost-written in an article with Jonathan Leake, Environment Editor at the UK Sunday Times. It claims, among other things, estimates showing a Google search emits between 7-10g of CO2. To put that in context, it cites (fairly accurately) 40-80 g CO2 per hour of PC use.

This claim translates to an astonishing 5-15 minute-equivalents of PC power use per search!

Separately, Dr. Urs Holzle of Google (CS PhD 1994, Stanford, SVP Google who owns such numbers) wrote that "[the estimate] is *many* times too high", at that it is actually about 0.2g-CO2-equivalent/search (or 9-18 seconds of equivalent PC use).

So, following my "trust-no-one-(blindly)" philosophy, I did my own calculations based on what I know, and my estimate came to 0.285 (call it 0.3) g-CO2 per search, just about 50% higher than Holzle's number, but no where near the 35-50x estimate mentioned in the Wissner-Gross-Leake article.

So AFAIAC, Google's data is more accurate. You don't have to take my word for it---I'm satisfied with my own calculations.

Nobody is releasing their methodology, so based on what I know, and what I think people know and don't know, here are the likeliest sources of errors:

1. An over-estimate of the number of machines in Google's data centers.
2. An over-estimate of the energy intensity per machine in Google's data centers.
3. An inaccurate estimate of the CO2-equivalent emission per unit power consumed by Google.
4. An under-estimate of how many searches Google really serves up per unit time.
5. The difference between the baseline and marginal amount of energy used in search-related activities (e.g., idle time)
   
6. An inaccurate accounting (if accounted for at all) of how much energy power Google's searches, and how much for other things (e.g., GMail, Calendar, Adwords, Adsense, YouTube, etc.)

I realize the last 2 points are subject to some contention of principles: one could argue from an aggregate business perspective that since most of Google's revenues come from search, all ancillary (ads) and related (apps) energy consumption is fair game for inclusion (*). But by that same aggregate standard, then, consider that Google is or is close to being carbon-neutral, so if we expanded the scope to include all this "larger perspective", the g-CO2 equivalent estimates would go way down, not up.

For people wondering why Google's gotten its dander up on this, it's simple: the company really cares and has done a lot about being green, so this is like accusing the Wiesenthal Center of being anti-Semitic.

Even the article's reference to the comparison of CO2 emissions by the IT industry to that of the airline industry misses the basic point that air travel is far greener per passenger-mile than any other mode save for train. The point is, in a numbers-rich and understanding-poor world, it's easy to make impressive statements absent of any sensible contextual unit (e.g, some measure of utility per g-CO2 emitted). I for one am glad IT has reached the 2% share of CO2 equivalents of the airline industry---it probably means we are applying more IT in the age of ... IT and becoming more energy efficient per unit GDP.

Wissner-Gross himself has disclaimed ownership of the claims, lamenting that the Times did it, because "The short answer is, it's a really easy way to sell papers. Google is a very successful company and it's a very easy way to get readership by making grandiose claims about them."

cf. Behan, Brendan F.

----
(*) and if I did include them, it's still within a factor of 2x, not 35-50x.