The Intel Museum is located in the Robert Noyce Building at company headquarters in Santa Clara, California. Yesterday Intel vice president Stuart Pann took seven of Robert Noyce's grandchildren on a tour of the museum, and three of us adults and a friend came along. It had been many years since we last visited with the kids, so all of us learned something.
My father started Intel in 1968 because Fairchild Semiconductor was getting too big to be fun anymore, with more employees than his hometown of Grinnell, Iowa had residents. Today Intel does $43 billion of business annually and has 85,000 employees worldwide. Intel's first chip, the 4040, had 2300 transistors. Its current top line chip contains 1.8 billion transistors and costs $200.
A silicon chip is made with a layer of silicon doped with arsenic to make it a semiconductor, which means that it can act as an on-off switch, conducting electricity in one configuration and acting as an insulator in another. Microscopic circuits are photo-etched right onto the surface of the silicon, in layers that are currently as little as 10 atoms thick. The circuits are so tiny and precise that a stray hair or flake of skin could ruin a whole batch, which is why chips have to be manufactured by people in coverall "bunny suits" in "clean rooms" 100 times cleaner than a surgical operating room.
Each new manufacturing facility or "fab" costs $2 to $3.5 billion to build. Twenty-five percent of the cost is the building itself, with its hundreds of miles of wiring and piping, and 75% is equipment. On average a fab has a million square feet and consists of several four-story buildings separated by inches of padding to minimize vibration transfer from one part of the fab to another. Intel has standardized the construction of these buildings well enough that a new fab can be constructed, up, and running within a year of breaking ground.
The computer chip has revolutionized the world, especially in the worlds of communication and commerce, and it has done so by the efforts of thousands of engineers constantly chipping away at the physical barriers of what we can do. These efforts have allowed the fulfillment of "Moore's Law," Gordon Moore's prescient observation that electronic advances allow the number of transistors to be placed on an integrated circuit (and therefore computing power) to double every 18-24 months even as costs fall by one half in the same period. Stuart told us that Intel can see its way to the next 3 generations of this progress, but after that, physical limits seem insuperable without another major conceptual breakthrough.
My favorite spot in the museum was an electronic copy of my father's scrapbook from age 12. In it he collected articles from Popular Mechanics, and on the first page he wrote that these were notes about his hobby, building things, which allowed m\him to make useful things and Christmas presents. We also went up to visit the VP and Presidential offices, each of which is a simple cubicle the same size as everybody else's: Intel still retains the egalitarian, meritocratic culture that has been its hallmark since the beginning.
On our way down the elevator, Damian whispered to me that he'd like to work at Intel, so I asked Stuart Pann what educational path someone should follow to be hired at Intel. What's needed, he replied, is a degree in engineering or computer science from a top twenty school. So later, at lunch, we looked up the top twenty schools on our iPhones-specialized knowledge at the flick of a finger, thanks to fast, cheap, plentiful and complex integrated circuits.