For more than half a century, the computer industry has followed an almost magical rule. Every few years the transistors, the tiny switches that make up the brain of any computer, the “digital neurons”, They were getting smaller, faster and more efficient. Thanks to this, our mobile phones today are millions of times more powerful than the computers that took humans to the Moon. But that race had been approaching a wall for some time. Not because there was a lack of ideas, but because Transistors are already being manufactured at a scale where there are only a few atoms left between one electric current and another. Reducing them even further seemed to push the limits of physics.
Now IBM claims to have found a new way out. The company has presented the first chip technology below the nanometer, with transistors of just 0.7 nanometers and a new architecture called Nanostack, capable of housing nearly 100 billion transistors on a chip about the size of a fingernail. Let’s go in parts. Regarding the number of transistors, we are talking about up to 5 times more than the most advanced processors of a current mobile phone and in a space similar to that occupied by a Snapdragon 8 Elite or an Apple A18 Pro.
And then there is the nanometer boundary, the distance that separates each of the transistors. What does 0.7 nanometers really mean? A nanometer is one billionth of a meter and, to be more precise, A human hair is approximately 80,000 nanometers thick and a red blood cell is around 7,000 nanometers. The IBM chip’s transistors are spaced less than a nanometer apart, a nothingness of atoms speaking ill and soon.
Those responsible for the advance discovered that the real innovation was not making the transistors smaller, but rather stopping placing them next to each other. Until now, The chips were similar to a city that could only grow when the land was exhausted, there was hardly any room to add new buildings. IBM’s proposal was to build skyscrapers. Its Nanostack architecture stacks transistors in three dimensions, one on top of the other, instead of only distributing them on a flat surface.
Thanks to this, they have managed to practically double the density compared to the experimental 2 nanometer chip that IBM presented in 2021. Why is more better? Each transistor performs extremely simple operations. But when a chip contains tens of billions running at the same timethis enormous multitude allows the execution of artificial intelligence models, scientific simulations or video games of enormous complexity.
According to IBM, this new architecture could offer up to 50% more performance or reduce power consumption by up to 70% compared to its previous 2 nanometer technology. At a time when data centers dedicated to artificial intelligence consume increasing amounts of electricity, any such improvement is especially relevant. Furthermore, the company claims to have Reduced the size of SRAM memory by 40%, one of the essential components for feeding data to modern AI processors. It should be noted that, for now, the new microchip is a research prototype, not a processor ready to reach personal computers.
The company itself estimates that it will take around five years before this technology can begin to be manufactured commercially, provided that industrial processes manage to reliably reproduce such tiny structures. It wouldn’t be the first time it happens. His experimental 2-nanometer chip, announced in 2021ended up serving as a basis for subsequent developments by manufacturers such as Samsung or Rapidus.