Atomic Machines exits stealth with $250 million and a system for building tiny machines

Atomic Machines has emerged from six years in stealth with a manufacturing system it says turns digital designs into working miniature machines. In a TBPN interview published October 13, CEO Jeff Holden disclosed $250 million raised and described its first device: a compact power relay. The larger ambition is to let others invent what the system builds.

A manufacturing language, not just a design tool

The platform, called the Matter Compiler, addresses a gap Holden sees between describing a machine and knowing how to manufacture it. An AI might help someone design a device, he argued, but that does not supply a production process. Manufacturers still have to work out how to make the thing, often building and refining a line for a particular device or class of devices over years.

Atomic Machines’ answer is a formal manufacturing language. Holden calls its structured instructions a process graph: a way to express what the system should build in a form it can execute. He says a correctly expressed design can produce a working micro-machine. His comparison with software is specific: an AI needs an underlying language and a compiler that can translate instructions into an output.

Here, that output is physical. Holden distinguishes the system from a 3D printer that makes shapes in material: his stated aim is complete machines with moving parts. Despite the company’s name, it is not yet manufacturing at atomic scale. It works with larger-scale materials to build millimeter-scale devices with features measured in microns, or thousandths of a millimeter.

PrimeSwitch gives the platform a first device

PrimeSwitch is a power relay, a device for switching an electrical circuit. Holden said it measures 9.5 millimeters across and 3 millimeters thick, can carry 150 amps, and can open in 50 microseconds. Those performance figures are company claims. The E’ville Eye describes the relay as designed for 800-volt power systems increasingly used in AI data centers.

Holden’s explanation for starting small is a manufacturing constraint, not simply a preference for compact products. He argues that conventional manufacturing becomes difficult at low-micron dimensions. Semiconductor fabrication can work at that scale, but was developed around transistors and related circuits. Atomic Machines is targeting the space for other kinds of miniature machines.

The next ambition is a foundry for outside inventors

Holden said Atomic Machines intends to offer the Matter Compiler as a foundry—a manufacturing service that builds devices designed by others. That would leave outside inventors to choose many of the applications, rather than requiring the company to develop every finished product itself. His longer-term goal is an AI interface that lets people describe a machine conversationally and receive physical output.

The applications he discussed range well beyond electrical switching. They included tiny drug-dosing pumps, medical robots that could move through the body, and cooling loops bonded to a GPU and placed inside its package. Holden presented these as potential uses of the manufacturing approach, not as an accompanying lineup of released products.

The company is also preparing to grow its physical operation. The E’ville Eye reported plans to hire for dozens of positions at a new East Bay facility. Holden disclosed the $250 million raised while discussing the exit from stealth, but did not identify investors, a valuation or a funding stage in the interview.