The BBC reports today on a study by Harvard physicist Alex Wissner-Gross. Wissner-Gross claims that performing a standard Google search on a desktop computer produces 7g of CO2. A quick session with two searches will produce 14g of CO2 – the same as that from boiling a kettle.
From the BBC article:
Although the American search engine is renowned for returning fast results, Dr Wissner-Gross says it can only do so because it uses several data banks at the same time.
Speaking to the BBC, he said a combination of clients, networks, servers and people’s home computers all added up to a lot of energy usage.
“Google isn’t any worse than any other data centre operator. If you want to supply really great and fast result, then that’s going to take extra energy to do so,” he said.
According to Google Web History, I’ve performed 9,308 Google searches and it’s only counted the searches I’ve performed whilst I was logged on.
I’m guesstimating I perform about 40 searches a day; that’s 15,000 Google searches per year (sounds scary when you put it like that). My annual Google carbon footprint would be 105kg of CO2 (0.15 tons).
Google have disputed this figure; saying that a search only produces 0.2g of CO2.
I’m not able to comment on what I think of the methodoly as I don’t know how either figure was reached. But I think it is important to point out the difference between average cost and fixed cost.
As an example, imagine a server farm which was responsible for 100g of CO2 emissions every day. If ten people perform searches, the average carbon cost of a search is 100g divided by 10 searches = 10g of CO2 per search. This is the average cost of the search.
Whereas, the marginal cost would be the CO2 cost of performing one more search. If we then performed an 11th search, the CO2 emissions of the server farm stay the same (we assume it’s running with spare capacity). The marginal cost of performing a search of zero grams of CO2.
With eleven searches, you could claim each search had a carbon cost of 9g. But that’s a bit unfair – considering the CO2 output of the server farm if you had made the search and if you had not, you find the CO2 output it exactly the same. Your search had a marginal cost of zero grams of carbon.
Whether Wissner-Gross and Google stated the average cost or the marginal cost I don’t know (although I suspect the first may have been the average cost and the second the marginal cost).
With Google’s server farms, we know that they will be running regardless of whether we perform searches or not. The important thing then is the marginal cost of a search – this being so close to zero, I don’t think any of us should feel a guilty conscience from using Google.