Intel will build a $3 billion facility for producing microprocessors next to an existing fabrication laboratory in Chandler, Ariz., the company announced Monday.
The new factory, designated Fab 32, will start cranking out microprocessors in the second half of 2007 using the 45-nanometer manufacturing process. With this technique, the average feature size on the chips will measure 45 nanometers. (A nanometer is a billionth of a meter.) Current chips are produced using the 90-nanometer process, and 65-nanometer chips will come out later this year.
Chipmakers reduce the average feature size of their chips every two years, a la Moore's Law. This improves performance and increases total output by allowing them to punch more chips out of the same amount of silicon--and hence increase profits.
The 45-nanometer generation is expected to be particularly difficult to pull off because chip designers will need to change many of the basic structures of the silicon transistor to control power and improve performance. Often in these difficult transitions, a few companies succeed without a glitch; others are forced to delay products.
The announcement of the factory's location quells months of speculation about where Intel would build. Government leaders from China, India, Ireland and Israel all held talks with the chipmaker. Intel already has fabs in Israel and Ireland, but building a plant in China or India would have been a coup for those countries: The chip giant so far has only built less high-tech testing and assembly plants in developing nations.
The Arizona facility will create 1,000 new jobs, according to the company. Arizona got the nod in part because of tax breaks inserted into the state budget in May, according to Bob Baker, senior vice president and general manager of Intel's Technology and Manufacturing Group.
Under the new law, a company that makes a capital investment of a billion dollars or more has flexibility in regard to how it pays sales tax. The company can calculate its sales tax by the output of a factory, or by the number of employees or by the total dollars invested. Intel will sell most of the chips it produces in Fab 32, so it will not likely use total sales as a basis for calculating its liability.
"It helped create a stable environment for this kind of investment," Baker said, noting, "We looked at a number of sites."
Although semiconductor manufacturers like to take advantage of low labor rates, the most important factors in determining where to put a fab are local tax breaks and the quality of the local electric, freeway and water infrastructures. Earlier this year, Intel CEO Paul Otellini spoke at hearings in Washington about the importance of tax breaks for semiconductor manufacturers and the often lucrative deals, such as the 10-year tax holidays offered in China, available overseas.
Most of the $3 billion cost gets consumed by equipment. Lithography machines can cost $18 million or more each, and they are just one of several pieces of equipment that need to be installed.
The factory will process wafers measuring 300 millimeters in diameter. Intel already has five 300-millimeter factories, and several 200-millimeter factories that are now used for flash memory and other products.
Separately, Intel said it will invest $105 million to convert an existing inactive wafer fab in New Mexico to a component temporary test facility.
"This investment positions our manufacturing network for future growth to support our platform initiatives and will give us additional supply flexibility across a range of products," Otellini said in a statement.