How Much Does a Single Search Cost Google?

I was wondering how much it costs Google to perform a single search. So I searched in Google, and was not able to find any readily available data that looked reliable. So I used up the better part of Saturday morning working on creating my own number.
I was able to find some ComScore data on total searches from July 2009, and I got Google's financials from 2009 from the SEC site.
My calculation? I estimate that it costs Google .68 cents per search in July 2009 in direct costs, and another .50 cents in overall operating costs, or a total of 1.18 cents per search. That's considerably more than I would have expected. (Please check my math and sources!)
I also found that for 78 billion searches in the month of July 2009, Google needed 3.442 B of capital equipment to support those searches. (Excludes YouTube and other non-search capital). Thus each search performed that month needed 4.4 cents of capital equipment. I did notice on Google's later financials that their capital costs didn't go up that much, so their capital cost per search is likely dropping. (plummeting?)
I'm wondering if anyone has newer data or more accurate data that they could plug in? Please let me know in the comments section.
Here is a summary of my estimate:


July 2009






Worldwide Google Searches K


Age 15+ Home & Work Worldwide

Cost of Revenues $K


Represents 70% that is search related

Direct Cost/Search




Search Related Operating Costs $K


Note: Click on Income Statement

Op Cost/Search




Total Cost Per Search



July 2009 - My estimate


Capital Required $K



Represents 70% that is search related

Cap/Per Search $



   Google Q3 09 Monthly Revenue $K


 2009 Q3 earnings release  Not quite true, but I'm assuming all revenue from search

Revenue Per Search



Cap/Per MM Searches/Mo



$44K of capital gets you 1 million searches/mo capacity


(that’s 33K searches a day; 1.4K searches/hr


23 searches/minute - hmmm. Seems like a lot of 


capital to support just 23 searches per minute! Remember, though; this includes the cost to index the ENTIRE WORLD of web pages to support those searches.


# Seconds in a Month



Searches Per Second



In July 2009

K Searches/Sec



Cap Cost per K Search/sec



It takes $116 million in capital to support 1,000 searches per second and the indexes to all the web pages in the world (roughly)


Attached below is a PDF of my background calculations. This PDF was made from an Apple Numbers Spreadsheet.


Here is the Apple Numbers Spreadsheet with all my source information. Please let me know if you find errors. Apple Numbers is nice because it lets you combine many spreadsheets on one "canvas."


Here is the Excel spreadsheet that Numbers generates for compatibility. I regret I didn't name the tables before exporting, but you'll get the idea as you click to the various tables. (Note: There may be problems with the Posterous viewer for this file type, so if you can't see this file, it isn't you! Also, for some reason, Posterous only lets you view the Excel file, while it lets you download the Numbers file.)
Update: 6PM 1-15-2010
Found this data from Nielson Us Rankings for February 2010

Top 10 Search Providers for February 2010, Ranked by Searches (U.S.) Rank Provider Searches (000) Share of Searches   All Search 9,174,408 100.0% 1 Google Search 5,980,116 65.2% 2 Yahoo! Search 1,294,261 14.1% 3 MSN/Windows Live/Bing Search 1,142,344 12.5% 4 AOL Search 206,969 2.3% 5 Search 175,074 1.9% 6 My Web Search Search 91,288 1.0% 7 Comcast Search 55,122 0.6% 8 Yellow Pages Search 27,002 0.3% 9 NexTag Search 26,461 0.3% 10 Network Search 24,681 0.3% Source: The Nielsen Company

They show Google having done 9.2 billion searches in the US in February 2010, vs 78 billion searches worldwide for Google reported by ComScore for July 2009
This ComScore data shows that North America (more than just the US) represents 22.1% of worldwide search traffic. By this ratio, the Nielson US ranking divided by .22 would be 26.8 billion searches worldwide for Google if the US represented all of North America. If the US represents 50% of North America, then the Nielson data represents a worldwide number of 53.6 billion searches worldwide.

Worldwide Search Market Overview by Region
July 2009
Total Worldwide – Age 15+, Home/Work Locations
Source: comScore qSearch

  Searches (MM) Share (%) of Searches Search Usage Days Per Searcher Searches Per Searcher Worldwide 113,685 100.0% 11.0 103.3 Europe 36,446 32.1% 11.8 116.9 Asia Pacific 35,001 30.8% 9.3 84.7 North America 25,095 22.1% 12.5 110.6 Latin America 10,524 9.3% 13.0 130.4 Middle East - Africa 6,619 5.8% 10.5 97.3


A-B Comparisons of the Concrete and Steel Kind

There's something so, well, concrete about the real world. While Google can run millions of A-B comparisons to compare how well people respond to a new feature, it's not that often that you get to see an A-B comparison with 400 vertical feet of concrete and two suspension bridges spanning the same body of water, just feet away from each other. That A-B comparison is available for anyone to see at the Penobscot Narrows Bridge and Observatory in Bucksport, Maine.  (See Google Map here.)

The original suspension bridge, called the Waldo-Hancock Bridge, opened in 1931, with a tower height of 235 feet. It uses the tensile strength of steel in two giant cables that carry the load of the roadway. Vertical cables attach the roadway to the suspension cables. The whole bridge looks delicate and light. The Penobscot Narrows Bridge and Observatory (where this sign was photographed) has a pylon height almost double, at 447 feet. The two bridges just couldn't be more different, even though they perform the same purpose, in the same place: to carry traffic across the Penobscot Narrows.

I love the process of design, and especially the downstream effects of a few design choices. Here, the first key choice is material. The original bridge was steel, and the new one is concrete. Steel is great in tension. Concrete loves compression. The older bridge is the classic suspension bridge where a few large suspension cables carry the load, and the bridge deck is hung from the cable. It all looks rather delicate, and the steel is mostly in tension, except for the two support towers which act as pillars to hold up the main cables (not the roadway itself).

Meanwhile, the cable-stayed bridge is the perfect solution for a concrete structure. Look at how these cables are pulling the roadway towards the pillars. The cables are not only holding the roadway up, but they are also compressing the roadway with great force. If you made the roadway structure from steel, you'd have to design it to take a heavy compression load, which would be very difficult. For concrete, compression is a walk in the park. 

The towers of a cable-stay bridge have to be very high to get the cables to be at the correct angle to the roadway. If the towers are too low, the cable will provide too much compression and not enough roadway lift. But seldom do you see the results of design decisions this starkly presented side by side. Look at how incredibly massive the new tower is. 

This photo shows the new bridge under construction. The cable-stay design has a cool feature: each side is self-sufficient! The two halves can be built incrementally until they meet. And it appears you can even build the tower and the bridge together as you go. 

The two designs also played out very differently when something goes wrong. Seventy years of salt air had done damage to the main suspension cables of the Waldo-Handcock Bridge. Maine DOT tried to rehabilitate the cables, but along the way realized they were unsound. An emergency contract was put out to add new cables to keep the bridge safe, and at the same time the load limit was cut dramatically. A replacement bridge would need to be designed and built in record time.

What happens inside a cable-stay bundle is quite different than on a normal suspension bridge. Here, each individual strand runs in parallel. And each can be replaced individually. And each is coated in epoxy to resist corrosion. 

They're even testing new, carbon fiber cables that will altogether eliminate the issue of corrosion. Here, we see the steel cable on the left, and the equivalent carbon fiber cable on the right. They've actually removed a few steel cables and replaced them with carbon fiber to test how they do. They can even track individual cables and assess how well they are holding up.

If you click on the picture, it should get big enough to read the sign.

Here's the view from the tower. It's a beautiful vista. The paper plant may not be scenic, but it provides lots of local jobs.

The old bridge lives on, perversely saved by budget cuts. The state lacks funds needed for demolition. So go see them both while the bad economy allows it!

And next time you're doing design, keep in mind that the decisions you make now will have impacts as great as you see in these two bridges. Even in software, the downstream nature of the software is often tightly tied to the key decisions you made in the very, very beginning of the project. So give those basic decisions lots of thought, and learn how they play out before you ...well...cast them in concrete.

I love Boston. I love Cambridge. I love it here. What an amazing day! #masstlc

It's 5am and I already woke up. So many thoughts from today!

I fell in love with Boston the first time I came here, in 1976. My sister was going to school here. I grew up in New Jersey, and my exposure to a city was New York. Huge, imposing...yes exciting... but surrounded by a moat and you had to pay a toll to get in.

My first impression of Boston was that it was so beautiful, so compact, so walkable, and so inviting. I came in 1977 to go to MIT, and I've been here ever since. The city, the region, and the people just keep getting more amazing. I love it here.

Today, we had the MassTLC Innovation 2010 unConference. It was an amazing event. I'll say more later, but for now:

Thanks so much to everyone who made this event just awesome!

Check out this tweet roll (Thanks Chris Myles)

Check out these amazing photos: (Thanks Dan Bricklin) 

Okay, back to bed. Thanks everyone! Wow.