By Christopher Mims | Photography by Justin T. Gellerson for WSJ
Sometime in the next 15 years, the same technology that produced the world's first photograph will allow us to make the most powerful, densely packed silicon microchips allowed by physics. It will be a triumph of engineering -- and the final step in the march of chip-industry progress known as Moore's Law.
Getting there is the multitrillion-dollar challenge the entire semiconductor industry now faces. It's so difficult that experts believe it won't happen commercially until 2040. Without this final step, electronics will only gain capability by bulking up.
The most valuable companies on earth are all watching this space. And that's why an innovation currently being tested at Johns Hopkins University is so interesting to them all.
Moore's end
Moore's Law says that every two years, the number of transistors packed into a microchip doubles. That can only happen so many times before there's no physical way to shrink transistors down any further -- at least not on silicon wafers, which is what microchips have been made from since their inception.
The technology that will get us to this endpoint, photolithography, is familiar to anyone who remembers how film used to be developed: Light shines through a negative and onto a sheet of paper coated with a light-sensitive chemical, re-creating the image.
With chips, the precision of that "image" is measured in tens of atoms. A light source with a very tight wavelength, somewhere between ultraviolet light and X-rays, shines through a chip-shaped stencil and onto the silicon wafer. On the wafer's surface, a chemical substrate called a photoresist reacts to the light, and etches the stencil pattern into the silicon.
The most advanced version of this tech currently in use today -- building the chips that go into Nvidia-powered AI supercomputers and potentially Apple's next-generation iPhones -- gets to a precision of 10 nanometers, or roughly 60 silicon atoms. The final frontier of silicon etching aims to cut that down to about 5 nanometers. If it's any smaller than that, the electric current flowing through the chip just won't stay in its lanes.
More transistors packed into a smaller space means you get the same processing power with less energy demand -- or it means for the same energy, you get more processing power. So Moore's Law has fueled both the mobile-computing revolution and the artificial-intelligence datacenter boom.
From camera obscura to X-ray lithography
The problem with etching patterns with atomically precise beams of light is that the material you shine it on also has to be atomically precise, or the pattern comes out blurry.
Back in the 1990s, chemists invented a weird substance called a metal-organic framework. It's a nearly perfect pattern of alternating shapes, where metal atoms hold longer carbon-based molecules in place in a sort of lattice. Best of all, it's self-organizing: Under the right conditions, it automatically forms a regular, crystalline pattern. The chemists who discovered it just won the Nobel Prize for their invention.
These properties make metal-organic frameworks an ideal photoresist -- that chemical layer on a silicon wafer that absorbs light to create the etched pattern of transistors and connections. Another advantage of metal-organic frameworks: You can create them using a huge variety of metals and organic molecules.
The key to designing a photoresist? "Making sure you have some type of regular structure that is very, very small," says Kayley Waltz, a fifth-year Ph.D. student at Johns Hopkins. Recently, I visited the labs of Waltz's adviser, chemical engineer Michael Tsapatsis, and Howard Fairbrother, a chemist. Together, their students and postdocs collaborated on a string of papers pioneering chip-making techniques involving metal-organic frameworks.
On a screen, Waltz showed me a super-magified image of a blue jay -- the Johns Hopkins mascot. She etched it into silicon using the metal-organic photoresist invented by the team she led. (Waltz's success at the university landed her a job at Micron, one of the top makers of memory chips.)
To take advantage of the atomic precision of this new photoresist, the light beam needs to be sharpened as well. Today's cutting-edge chips with 10nm features are created using extreme ultraviolet light sources from machines made by Dutch company ASML. Each can cost up to $400 million, and is as big as a city bus. To etch features any smaller, ASML or one of its nascent competitors will have to find a way to go from extreme UV light to X-rays, which have an even tighter wavelength.
Barriers to commercialization
I spoke to three additional researchers who lauded the Johns Hopkins team for its pathbreaking work. But all of them have explored alternative photoresist chemicals that will compete to be the one semiconductor companies settle on.
Whichever substance will win must be compatible with existing microchip fabrication processes, because so many billions have already been invested in them, says Kaiying Wang of the University of South-Eastern Norway. "How to adopt this process into the current, mature semiconductor industry is a challenge," he adds.
Chang-Yong Nam, a senior scientist at Brookhaven National Laboratory, says he thinks the properties of metal-organic frameworks make them an alluring candidate, though he himself is currently at work on a photoresist option that builds on current approaches.
Given the enormous engineering challenge of integrating metal-organic frameworks into commercial chip-making, we probably won't see it in fabs until 2040 at the earliest, says Tsapatsis of Johns Hopkins.
Even so, today's technological momentum is probably enough to get us there. Assuming Apple stays on its yearly cadence of device renewal, we will be on the iPhone 32 by then. Makers of smartglasses need such more powerful yet less power hungry chips. Same goes for all the AI companies, which seek computational power that doesn't add to the energy demands already threatening to drive our electricity bills higher than our mortgages.
Beyond 2040, if we want features on chips to become any smaller and gadgets to become more capable, we will probably have to abandon silicon altogether, according to every expert I spoke with. The good news for the Johns Hopkins team: The metal-organic frameworks they've pioneered are potentially so precise that future microchips made of graphene or other exotic materials might be patterned using them.
What comes after silicon microchips -- pocket quantum computers, say, or transistors only an atom thick -- could be as strange to people living today as microchips were in the 1960s, when Gordon Moore started laying down the law.
Write to Christopher Mims at christopher.mims@wsj.com
(END) Dow Jones Newswires
February 27, 2026 14:00 ET (19:00 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments