The Myth of the Carbon Cost of Googling

Posted by | January 12, 2009 | Uncategorized | 2 Comments

This article on the BBC today claims that “Two search requests on the internet website Google produce as much carbon dioxide as boiling a kettle”, according to Harvard physicist Alex Wissner-Gross.

Yet in the same article, Google say that “in the time it takes to do a Google search, your own personal computer will use more energy than we will use to answer your query”. Google have a blog post which goes into more detail about the environmental cost of search.

Given that laptop power is somewhere between 20 and 50 Watts and a desktop somewhere between 50 and a 150 Watts and takes fractions of a second to search, whereas the average kettle requires 1800 Watts and takes over a minute to boil, there is something wrong here (by a factor of tens of thousands). In fact, the act of wandering over to a kettle and switching it on uses more energy than a Google search. I assume this is bad journalism rather than bad science.

Elsewhere in the article is the statement “A recent study estimated the global IT sector generated as much greenhouse gas as the world’s airlines put together.”

This is a largely meaningless assertion since without the global IT sector people would have to use planes more. In addition, ‘global IT sector’ includes all of the computers required to design, build and operate aircraft, all of the computers used to search for and book trips all of the computers used for in flight entertainment systems and all of the computers used by air traffic control.

In terms of moving people around to exchange information, if you wanted to make plane travel several million times more efficient, an Internet enabled computer is a viable alternative.

Since the availability of efficient access to information makes us access it wastefully (i.e. more people search for Paris Hilton than travel to Paris to stay in the Hilton) the idea of computer waste has some validity. However, to curb carbon emissions, fighting spam (denial of service attacks alone, account for more Internet traffic than all the world’s email) and improving software efficiency (searching Outlook is several thousand times slower than Gmail) would be a priority above making computer hardware more efficient, which has happened on its own through Moore’s Law.

Assuming that we should look at computer hardware and Internet infrastructure efficiency as a priority, there are two aspects to computer power use, calculation and bit transfer. If we measure the average bandwidth of a human being’s speech in bits and compare that to the energy used for bit transfer via the internet the latter is several million times more efficient. To illustrate the point, imagine the inefficiency of shipping a computer across the word to deliver an email. This is what shipping people to meetings via plane is like.

The calculation cost is also more efficient by a factor of several million, in many cases. The cost of calculating by slide rule and human requires feeding humans. If there were some equivalent unit of measure for computers, analogous to horse power, (human power) that measured serial calculations in terms of what a slide rule enables the brake human power of my laptop would be the equivalent of a football stadium full of snack eating, kettle boiling, tea drinking, human calculators.

For massively parallel calculations, human power is still more efficient, but it is realistic to assume that we will be able to develop machines that will outperform us. When we do, we should look at what energy saving that will allow as well as what it will consume.

2 Comments

  • Sometimes I despair that journalists can’t run figures through a “common sense” filter. The BBC article read like a secondary school essay and missed any attempt at analysis.

  • admin says:

    Yes, although the BBC is not alone and the worst culprits this morning are UK newspapers (Guardian, Times). By the time the US presses run, later in the day, things might be more objective.