In a tiny, windowless conference room at the R&D headquarters of Intel, the world's dominant microprocessor and semiconductor manufacturer, Mark Bohr, the company's director of process architecture and integration, is coolly explaining how Moore's law, as it is commonly understood, is dead—and has been for some time. This might seem surprising, given that Bohr is literally in the Moore's law business: his job is to figure out how to make Intel's current 14-nanometer-wide transistors twice as small within the decade. But behind his roundrimmed glasses, Bohr does not even blink: "You have to understand that the era of traditional transistor scaling, where you take the same basic structure and materials and make it smaller—that ended about 10 years ago."

In 1965 Gordon Moore, then director of R&D at Fairchild Semiconductor, published the bluntly entitled document "Cramming More Components onto Integrated Circuits." Moore predicted that the number of transistors that could be built into a chip at optimal cost would double every year. A decade later he revised his prediction into what became known as Moore's law: every two years the number of transistors on a computer chip will double. ...