Amazon is giving people a chance to rent a copy of the net for themselves.
The move will help Amazon compete with rivals Google
Via its subsidiary Alexa, the e-commerce firm is letting people get at a regularly updated copy of much of the information found on the web.
Via the Alexa service, anyone with a basic knowledge of programming will be able to search 4.5 billion web pages from more than 16 million websites for whatever they want.
Prices for the service start at $1 per processor per hour to crunch the data.
Those signing up for the service get access to the copy of the web that Alexa refreshes every month, which comprises up to 300 terabytes of data. Users also get at the infrastructure Alexa maintains to crunch through the data for interesting information.
In a statement Alexa said that it expected a lot of interest from entrepreneurs keen to use web search systems for their own applications or services.
As examples of what can be done with a copy of the web Alexa showed off an image search application and a service called Musipedia.
The image search allows people to query all the metadata that digital cameras attach to snapshots which record when a picture was taken, which model and make of camera was used and the image's size.
WEB SEARCH COSTS
$1 per processor hour
$1 per gigabyte/year of user storage
$1 per 50 gigabytes of data processed
$1 per gigabyte uploaded/downloaded
The Musipedia service lets people search for song melodies and lets them whistle a query and submit it to the database.
Industry experts said Alexa's move could change the search industry because the cost of setting up a global infrastructure to scan and index the web was prohibitive.
By contrast Alexa said that running searches via a copy of the entire web should cost a few thousand dollars.
Users pay for the amount of computer processors they use to crunch through data, by gigabyte of storage they need and how much data they have processed.
The move comes as a challenge to search giants such as Google and Yahoo which do let coders get at their net data but only via closely controlled programming interfaces.