Une puce sur l’épaule

Last year, the chip industry produced more transistors than the combined quantity of all goods produced by all other companies, in all other industries, in all human history. Nothing else comes close.

Conway and Mead eventually drew up a set of mathematical “design rules,” paving the way for computer programs to automate chip design. With Conway and Mead’s method, designers didn’t have to sketch out the location of each transistor but could draw from a library of “interchangeable parts” that their technique made possible. Mead liked to think of himself as Johannes Gutenberg, whose mechanization of book production had let writers focus on writing and printers on printing. Conway was soon invited by MIT to teach a course on this chip design methodology. Each of her students designed their own chips, then shipped the design to a fabrication facility for manufacturing. Six weeks later, having never stepped foot in a fab, Conway’s students received fully functioning chips in the mail. The Gutenberg moment had arrived.

ASML’s history of being spun out of Philips helped in a surprising way, too, facilitating a deep relationship with Taiwan’s TSMC. Philips had been the cornerstone investor in TSMC, transferring its manufacturing process technology and intellectual property to the young foundry. This gave ASML a built-in market, because TSMC’s fabs were designed around Philips’s manufacturing processes. An accidental fire in TSMC’s fab in 1989 helped, too, causing TSMC to buy an additional nineteen new lithography machines, paid for by the fire insurance. Both ASML and TSMC started as small firms on the periphery of the chip industry, but they grew together, forming a partnership without which advances in computing today would have ground to a halt.

Yet DARPA and the U.S. government have found it harder than ever to shape the future of the chip industry. DARPA’s budget is a couple billion dollars per year, less than the R&D budgets of most of the industry’s biggest firms. Of course, DARPA spends a lot more on far-out research ideas, whereas companies like Intel and Qualcomm spend most of their money on projects that are only a couple years from fruition. However, the U.S. government in general buys a smaller share of the world’s chips than ever before. The U.S. government bought almost all the early integrated circuits that Fairchild and Texas Instruments produced in the early 1960s. By the 1970s, that number had fallen to 10−15 percent. Now it’s around 2 percent of the U.S. chip market. As a buyer of chips, Apple CEO Tim Cook has more influence on the industry than any Pentagon official today.

However, “globalization” of chip fabrication hadn’t occurred; “Taiwanization” had. Technology hadn’t diffused. It was monopolized by a handful of irreplaceable companies. American tech policy was held hostage to banalities about globalization that were easily seen to be false. America’s technological lead in fabrication, lithography, and other fields had dissipated because Washington convinced itself that companies should compete but that governments should simply provide a level playing field. A laissez-faire system works if every country agrees to it. Many governments, especially in Asia, were deeply involved in supporting their chip industries. However, U.S. officials found it easier to ignore other countries’ efforts to grab valuable chunks of the chip industry, instead choosing to parrot platitudes about free trade and open competition. Meanwhile, America’s position was eroding.

Jim Keller, the star semiconductor designer who’s widely credited for transformative work on chips at Apple, Tesla, AMD, and Intel, has said he sees a clear path toward a fifty times increase in the density with which transistors can be packed on chips. First, he argues, existing fin-shaped transistors can be printed thinner to allow three times as many to be packed together. Next, fin-shaped transistors will be replaced by new tube-shaped transistors, often called “gate-all-around.” These are wire-shaped tubes that let an electric field be applied from all directions—top, sides, and bottom—providing better control of the “switch” to cope with challenges as transistors shrink. These tiny wires will double the density at which transistors can be packed, Keller argues. Stacking these wires on top of each other can increase density eight times further, he predicts. This adds up to a roughly fifty times increase in the number of transistors that can fit on a chip. “We’re not running out of atoms,” Keller has said. “We know how to print single layers of atoms.”