The Third Dimension of Compute

For more than half a century, the semiconductor industry has pulled off one of the great engineering feats in human history: making things smaller. Transistors shrank, chips got faster, computing got cheaper, and entire industries were built on the assumption that this process would continue.

Now the industry is running out of room.

The most advanced chips in the world are manufactured at dimensions measured in nanometers, using some of the most sophisticated machines ever built. An advanced lithography system can cost hundreds of millions of dollars, and the fabs surrounding those machines cost tens of billions. Each additional improvement is getting harder, slower, and more expensive.

So chipmakers have increasingly started moving in another direction: up.

Modern processors are becoming three-dimensional systems, stacking memory, compute, interconnects, and other components to pack more performance into the same footprint. But there is an awkward mismatch at the center of this transition. The chips are becoming three-dimensional while much of the machinery used to manufacture them is still built around two-dimensional processes.

Atum Works was founded around a simple question: what if the manufacturing process went 3D too?

A Different Way to Build

Atum is developing a new form of nanoscale manufacturing that can fabricate complex structures directly in three dimensions. The company is building what it describes as a scalable 3D lithography platform capable of creating structures at microscopic scales.

The easiest shorthand is audacious: 3D ASML. It’s a description Atum itself has embraced.

Traditional semiconductor manufacturing creates complexity through a long sequence of deposition, lithography, etching, and other steps repeated layer after layer. As chips become more three-dimensional, that process becomes increasingly cumbersome. Atum starts with a different premise: treat manufacturing as a volumetric problem from the beginning and create microscopic structures directly in three dimensions.

If it works at scale, engineers could produce geometries that today require enormous numbers of process steps, or in some cases simply cannot be manufactured at all.

It sounds like the kind of idea that should come out of a national lab, and in a way, it did.

Built for the Problem

Lucas Pabarcius, Matteo Kimura, and Malcolm Tisdale came together with backgrounds spanning Caltech, NASA, robotics, physics, and some unusually ambitious engineering projects.

Lucas studied physics at Caltech and spent time at NASA Goddard working across optics, propulsion, and spacecraft control systems. His work has consistently lived at the intersection of fundamental physics and things that actually have to work in the physical world.

Malcolm studied mechanical engineering at Caltech and worked at NASA’s Jet Propulsion Laboratory, where he designed and tested robotic systems for future Mars helicopters. At Caltech, he also helped lead teams developing technologies for extreme lunar environments through NASA’s BIG Idea Challenge.

Matteo came out of Caltech as a computational mathematician, applying mathematics and algorithms to controls and difficult engineering problems, including lunar engineering projects.

There is a useful pattern in those backgrounds. These are people accustomed to working on physical systems where advances in computation, physics, controls, and manufacturing all have to come together.

Atum requires exactly that kind of range. The problem sits at the intersection of physics, optics, materials science, computation, precision engineering, and manufacturing. A breakthrough in one discipline is useless if the rest of the system cannot keep up.

For a company attempting something this technically ambitious, the team feels unusually well matched to the problem.

Compute Becomes a Manufacturing Problem

AI has made the limitations of today’s hardware impossible to ignore. The industry needs more compute, more memory bandwidth, faster interconnects, better packaging, more efficient cooling, and new materials. Chip designers can imagine increasingly sophisticated architectures, but eventually someone has to manufacture them.

That makes manufacturing itself one of the important frontiers in computing.

Atum is starting with semiconductors because the need is immediate and the value of solving it is enormous. But the underlying capability is broader. Precise three-dimensional manufacturing at microscopic scales could eventually matter anywhere engineers can design structures more complex than today’s manufacturing systems can economically produce.

This is what makes the company particularly interesting to us. Atum is not tied to a single chip architecture or one generation of semiconductor design. It is working further down the stack, on the machinery that determines what can be made in the first place.

The ASML Question

Comparing a startup to ASML should make any investor uncomfortable.

Atum is not building an EUV machine, so the comparison shouldn’t be taken literally. What makes ASML interesting is the position it occupies. When the semiconductor industry needed an entirely new way to manufacture the next generation of chips, the company capable of making that generation possible became one of the most consequential businesses in technology.

Atum is making a similar kind of bet on three dimensions.

If computing continues moving vertically, manufacturing systems designed from the beginning to operate in three dimensions could become increasingly important. And if Atum can build one of those systems, the company may end up enabling products its founders haven’t imagined yet.

Why We Invested

At Ritual, we are drawn to companies where the ambition initially feels slightly unreasonable, but the people pursuing it make you reconsider what might actually be possible. Atum was one of those companies.

Lucas, Matteo, and Malcolm have spent much of their early careers building at the edge of physics and engineering. They understand the underlying science, know what it takes to make difficult physical systems actually work, and chose a problem where progress could unlock entirely new ways of building computing hardware.

But what ultimately convinced us was the shape of the opportunity. For decades, improvements in computing were driven by our ability to manufacture smaller and smaller structures. The next era may require learning how to manufacture entirely different ones.

If Atum succeeds, it could give engineers a new set of building blocks for that era, and history suggests that when you expand what engineers are capable of making, the most interesting applications are often the ones nobody saw coming.

Visit Atum Works at https://atum.works/.