“I’ve committed the whole company to 18A,” Intel CEO Pat Gelsinger told an interviewer earlier this year: It’s a disgusting statement to non-techies, but the story behind it is understandable to anyone and worth knowing. It is an awesome reminder of the endless reach of human ingenuity, relevant to all kinds of challenges.
What Gelsinger was talking about: Intel is one of the world’s leading designers and manufacturers of computer chips, and the 18A is the chip Intel plans to start making in volume next year. It will be the most advanced chip ever, though Taiwanese semiconductor manufacturing The company (TSMC), the world’s largest chip maker, will start making a chip next year that matches at least 18A in volume. This designation, 18A, indicates the density of the chip’s transistors, among other things, and the lower it is, the higher the density. This is still the lowest number of chips; 18A stands for 18 angstroms, one angstrom being ten billionths of a meter. The rest of the industry labels chips by nanometers; TSMC’s new chip is 2nm, by that measure the 18A would be 1.8nm. Even without understanding the production process, we are amazed by its infinitesimal scale. Chipmakers today work at the atomic level. A silicon atom is 0.21 nanometers wide, for example.
That’s enough atomic physics. It’s important to note that these new chips are more than amazing. They were supposed to be impossible. Moore’s famous law states that the number of transistors on a chip—essentially the transistor density—doubles every two years or so. It was accurate for decades, but even Gordon Moore himself admitted: “You can’t go on forever.” The laws of physics would allow transistors to be so small. The crucial question was how small.
Experts have been predicting the answer with certainty and wrong for years. Some said 2010 would be the end, when the leading chip was 28nm. After chipmakers passed that threshold, a new set of forecasters saw 2020 as the frontier, when the leading edge was 5nm. However, even now Intel’s 1.8nm chip is not the end of the line. The company’s next chip — 1.4 nm — is under development.
How can this be? The answer lies in two parts, both of which generally belong to human ingenuity.
· Chipmakers and their suppliers continued to find innovative new ways to extend Moore’s Law life that prognosticators could not predict. For example, chipmakers figured out how to stack transistor components on top of each other, packing more computing power into a given space. Another example: Chipmaking involves shining light through a mask onto a silicon wafer, but as transistors became increasingly microscopic, it was difficult to use light at short enough wavelengths to print sharp patterns. ASML, a Dutch manufacturer of chip-making equipment, developed machines to handle the necessary extreme ultraviolet light, and today’s cutting-edge chips cannot be made without these machines. To produce the 18A, Intel CEO Gelsinger negotiated with ASML to be the first to receive the company’s latest model.
· Companies ran around Moore’s Law. Adding more transistors isn’t the only way to get more value out of a chip. Advanced algorithms and software now help users get better performance from their computer chips. Chipmakers are also producing chips designed for specific applications, such as AI. Those chips aren’t the best at everything, but they’re perfect for specific tasks. Broadcomtechnology company quietly become one of the most valuable companies in the worlddesigns many such chips.
Strictly speaking, Moore’s Law does not apply. As transistors approach the atomic scale, the number of transistors on a chip has not doubled every two years. But what? Computer science continues to advance at a rapid pace—quantum computing and other wonders on the horizon—and that’s what counts.
The saga of Moore’s Law is an example of how we almost always underestimate human ingenuity. Early computer language translation researchers were pessimistic that it could progress from its near-useless status quo in the mid-1960s, but today your phone has a free app that can translate 100 languages and more, and it’s getting better all the time. . MIT professor Hubert Dreyfus, in a 1972 book called What computers can’t dohe saw little hope that computers could make significant progress beyond the mediocrity achieved in playing chess. However one IBM computer beat world champion Garry Kasparov in 1997, and Kasparov told me today that the free chess app on your phone is more powerful than the computer that beat him.
There is no telling what will happen to Pat Gelsinger’s Intel bailout plan or individual search. But in general, over time, we can be sure that at least two things will be infinite: human desires and human ingenuity. We will always be satisfied with problems to solve, and we will always find surprising, unexpected solutions, at least to some of them. In this troubled world, that’s not a bad basis for optimism.
