ABSTRACT

In 2005, the computing industry set down a path many believed was destined for disaster. Facing the near-term physical and atomic limits of cramming over 592 million transistors onto a single microprocessor chip, AMD and Intel Corporation switched from increasing the number of transistors per microprocessor to adding more and more processors—cores—per chip. While in the short term this switch to a multicore paradigm avoided the physical limitations associated with the previous thirty-year trajectory, unprecedented technical land mines lay on the immediate horizon. Unfortunately the computing industry did not seem ready to tackle what was likely to be its greatest challenge in over thirty years. In the 1950s and 1960s, vertically integrated computing firms were home to research and development, but more recently large firms have outsourced their innovation needs to small and medium-size enterprises, universities, and government labs, which are more flexible and innovative. This disintegrated innovation model, however, created challenges, both in coordinating technology development incentives across the multitude of players and in supporting long-term research. In addition, with the primary source of demand in computing having moved from the military to the commercial sector, the long-term supporter of the computing industry—the defense industry—was struggling to maintain its power as well as its technological edge. And yet suddenly there in the fray, despite being proclaimed "dead" by the very computing academics with whom it had once founded the industry, was the computing industry's long-time supporter, the Defense Advanced Research Projects Agency (DARPA) funding a technology that could help save the day. This chapter tells the story of the crisis in Moore's Law, the dramatic change in both the computing industry and the military-industrial complex at the start of this crisis, and the controversial emergence of a new role for government in the form of the computing industry's savvy, beloved, and controversial supporter—DARPA.