Citation and evidence

Atomic Machines

4 min full readUpdated 7 references

This article's verification

Report a problem with this article

More

Use this article

Raw MarkdownExplore connections

Improve this page

Suggest editRevision history

Topics

AI CompaniesAI HardwarePhysical AI

Cite this article

Atomic Machines is a California company developing the Matter Compiler, an AI manufacturing system for making functional micro-machines from digital instructions. It introduced the platform and its first disclosed product, the PrimeSwitch PS-150 electromechanical relay, on October 7, 2026, after six years of development in stealth.[1] The company describes the platform as a fab that combines design, fabrication, measurement, and learning.[2]

Company and launch

Jeff Holden founded Atomic Machines and is its chief executive. The company has offices in Emeryville and Santa Clara in the San Francisco Bay Area. At its launch, it reported $250 million raised to date. This was a cumulative funding disclosure, not an announcement of a single new $250 million financing round.[1]

Its leadership includes chief technical officer Klaus Zietlow, vice president of Device and Matter Compiler Engineering James Stölken, and director of AI, Modeling and Simulation Marta D'Elia. The company lists Construct Capital, Gigafund, OneIM, Sozo Ventures, Valor Equity Partners, XTX Ventures, and the University of California among its investors.[3]

Matter Compiler

The Matter Compiler uses software-defined operations to construct devices with moving parts and multiple materials. In Holden's description, reusable manufacturing operations replace product-specific molds, masks, and fixtures. Measurements during fabrication let the system detect and correct errors rather than carry them through subsequent operations. The name refers to the manufacturing machines in Neal Stephenson's novel The Diamond Age; Holden describes Atomic Machines' work as a bounded microscale implementation of that idea.[4]

Construct Capital, an investor in the company, describes the manufacturing system as a collection of automated machines called Nodes. Each Node performs a class of operations, including laser machining, fluid deposition in picoliter droplets, electrochemical plating and polishing, or robotic handling and assembly. The investor identifies metals, polymers, glasses, and ceramics as materials in the platform's library.[5]

The same account describes a learned world model of how manufacturing operations change material. Measurements before and after operations provide feedback for corrections and training data for reinforcement learning. These are descriptions from an investor of the company's architecture, rather than independent performance measurements.[5]

The company's proposed workflow starts with a description of the desired device, followed by design, simulation, process planning, and approval for production. Its platform page labels one-shot prompt-to-product manufacturing as an aim. It should therefore be distinguished from a demonstrated service that can manufacture any requested object.[2]

PrimeSwitch PS-150

PrimeSwitch PS-150 is the first micro-machine Atomic Machines disclosed from the Matter Compiler. It is a bistable electromechanical relay: it retains its switching state without continuous holding power. The product page lists the following manufacturer specifications.[6]

PropertyManufacturer specification
Continuous current150 A, AC or DC; no active cooling required
Opening time50 microseconds to 1,500 V galvanic isolation
On-resistance200 microohms
Dimensions9.5 mm diameter x 3 mm
Holding powerZero, bistable operation
ConfigurationSPST
PackageSMD
AvailabilitySampling Now

Expanded article table

Atomic Machines targets the relay at power protection for 800 V DC systems in AI data centers. Its launch announcement describes using the relay in a hybrid breaker with a solid-state bypass that carries current while the mechanical contact opens. It reported shipping devices to early-access customers for evaluation, rather than general retail availability.[1] The product specifications describe the relay, not the performance of every breaker assembled around it.[6]

Earlier technical presentation

A Berkeley Sensor & Actuator Center seminar on February 25, 2025, listed Stölken as a speaker from Atomic Machines. Its abstract described the Matter Compiler as a modular manufacturing platform with extensible materials and process libraries and identified a micro-power relay as a device example.[7] The seminar listing documents an earlier public technical presentation; it is not a peer-reviewed validation of the October 2026 product specifications.[7]

References

  1. ^1 ^2 ^3Atomic Machines. Atomic Machines Emerges from Stealth with $250 Million to Launch the Matter Compiler, an AI-Native Manufacturing System That Builds Micro-Machines from Code. Business Wire, October 7, 2026.
  2. ^1 ^2Atomic Machines. The Matter Compiler. Accessed October 10, 2026.
  3. ^Atomic Machines. Company. Accessed October 10, 2026.
  4. ^Jeff Holden. Atomic Machines Origin Story. Atomic Machines, October 7, 2026.
  5. ^1 ^2Construct Capital. Why We Invested in Atomic Machines: Prompting Matter. October 7, 2026.
  6. ^1 ^2Atomic Machines. PrimeSwitch. Product specifications. Accessed October 10, 2026.
  7. ^1 ^2Berkeley Sensor & Actuator Center. BSAC Technology Seminar: Revolutionizing Micromachine Manufacturing: AI-Driven Design and MEMS 2.0 with the Matter Compiler. February 25, 2025.

Improve this article

Add missing citations, update stale details, or suggest a clearer explanation. Every suggestion is reviewed for sourcing before it goes live.

1 revision · v1 · 749 words · full history

Fact-checks are independent of edits: a reviewer re-verifies the article against its sources and stamps the date. How we verify

Research and drafting on this wiki are AI-assisted, under named human editorial standards. How AI is used here

Reviewer note: Independent full-article review against 7 cited primary and academic sources, October 10, 2026. Checked subject identity, specifications, availability, benchmark conditions and limitations.

Cite this page: AI Wiki. "Atomic Machines." aiwiki.ai, updated 10 Oct 2026, fact-checked 10 Oct 2026. CC BY 4.0. https://aiwiki.ai/wiki/atomic_machines

Suggest edit