Atomic Machines, a company based in Emeryville, California, emerged on 7 October 2026 from six years in stealth and said it has raised $250 million to date. Its release names investors including OneIM, Gigafund, XTX Ventures, KAS Venture Partners, Valor Equity Partners, Tru Arrow Partners, Construct Capital, Sozo Ventures and the Regents of the University of California.
The company introduced the Matter Compiler, which its release calls an AI-native, all-digital manufacturing system that builds working machines directly from code. It also announced PrimeSwitch, a power relay that its release pitches at AI data centers moving to 800-volt DC distribution.
Business Pill · OPERATING LEVERAGE
A short explainer of operating leverage: when costs are mostly fixed, each extra sale adds more to profit. It relates to the release’s claim that one fab’s capital is amortized across every product run on it. It teaches the general idea only and says nothing about any company in this story.
The key insight: As we read it, the company is selling two things at once: a manufacturing system it says turns code into micro-machines, and a first product aimed at the power hardware of AI data centers. The release ties PrimeSwitch to the move to 800-volt DC distribution.
The release says the Matter Compiler is aimed at micro-machines: “motors and gears the size of a grain of sand, robots small enough to work inside the body, cooling built directly into the chips that run AI”.
The company was founded by Jeff Holden. The release says he joined Amazon as its tenth engineer, built Amazon Prime and rose to senior vice president, then became Uber’s first chief product officer and founded its self-driving, AI labs and air-taxi divisions.
The release says PrimeSwitch will be shown publicly for the first time at the Open Compute Project (OCP) Global Summit in San Jose, October 12–15.
Atomic Machines says it has raised $250 million to date; the investors listed are among those its release of 7 October 2026 names. The release gives no valuation or split by round.
Where the AI Sits
The release describes two claimed breakthroughs. The first is a general-purpose constructor in which, it says, code alone determines the device produced, with features down to single-digit microns and no hard tooling of any kind.
The second, in the release’s words: “the entire physical manufacturing system sits inside the AI’s loop: every operation is measured with high-precision metrology and corrected in real time, and the AI runs its own experiments on its own hardware to test what simulation proposes.”
The company page puts the reason this way: “Generative AI designneeds more than a model: it needs physical systems that give the AI granular, closed-loop control over the manufacturing process and feedback from it.”
In his origin-story essay, Holden calls the destination prompt-to-product manufacturing: “an experience like working with a modern AI coding agent, except that some of the output tokens are physical.” The Matter Compiler page lists one-shot Prompt-to-Product as the company’s aim.
The 50-microsecond opening time, the ‘about a thousand times faster’ comparison, the ratings and the availability are from Atomic Machines’ release of 7 October 2026; we did not test them.
A Switch Built for 800-Volt Data Centers
The release says PrimeSwitch PS-150 is a bistable electromechanical relay that opens in 50 microseconds, with on-resistance of 200 micro-ohms, 1,500 volts of galvanic isolation and a continuous current rating of 150 amps. It says the package is 9.5 millimeters across and 3 millimeters thick, and that the relay holds its state with zero power.
The release puts the opening speed at about a thousand times faster than a conventional contactor. It ties the product to AI data centers moving to 800-volt DC, where, it says, “a fault delivers destructive energy in microseconds and burns as an arc that direct current will not extinguish on its own.”
In the hybrid breaker the release describes, a small solid-state bypass carries the current for the microseconds the contact is opening. Its summary: “The silicon interrupts; the metal carries and isolates; the breaker never arcs.”
The release says the PS-150 is currently shipping to early-access customers for evaluation. The PrimeSwitch page lists availability as sampling now.
It says the Matter Compiler needs neither tooling nor a fixed process, so “A lot size of one carries no engineering charge and no process development.” It adds that high-mix and high-volume production happen in the same fab, with capital amortized across every product run on it.
Holden says in the release: “When the marginal cost of a new design is the cost of the design file, you get an innovation explosion, and micro-machines are where that explosion has been waiting to happen.”
Luke Nosek, Managing Partner at Gigafund, says in the release: “If vibe coding is how software gets made now, this is the start of vibe manufacturing.”
The Structural Read
The release locates the AI in the factory, not in the product. It says the entire physical manufacturing system sits inside the AI’s loop, with every operation measured and corrected in real time.
The first product targets a power problem. The release says conventional mechanical protection opens in milliseconds, and that the PS-150 opens in 50 microseconds while keeping metal-to-metal conduction.
The economic claim is about tooling. The release says a lot size of one carries no engineering charge and no process development, and that the same fab handles high-mix and high-volume production.
PROOF POINT The release says PrimeSwitch will be shown publicly for the first time at the OCP Global Summit in San Jose, October 12–15.
AVAILABILITY The release says the PS-150 is currently shipping to early-access customers for evaluation; the product page lists it as sampling now.
NEXT DEVICES The company page says PrimeSwitch is the first of what will be many new micro-machines, and the Matter Compiler the first of several editions.
The framework’s starting point: “The structural bottleneck isn’t computing, it’s electrons.”
As we read it, PrimeSwitch is pitched at one piece of that problem: protecting the 800-volt DC power distribution the release says AI data centers are moving to, a layer below the chips themselves.
What Is Not Established
We read Atomic Machines’ release of 7 October 2026 in full as reprinted by Yahoo Finance, because businesswire.com refused our request. We also read the company’s Matter Compiler, PrimeSwitch and company pages and its two blog posts of 7 October in full.
The documents we read give $250 million as the total raised to date. They do not give a valuation, how the total splits across rounds or investors, revenue, PrimeSwitch pricing or the names of the early-access customers. The PS-150 figures are the company’s own, and we did not test them. We did not read press coverage of the launch, and we did not contact Atomic Machines or its investors.
Business Engineer Framework
The State of AI Data Centers
A Business Engineer framework on the physical build-out behind AI, from gigawatt campuses to the power that feeds them.
Atomic Machines says it has raised $250 million to date and built the Matter Compiler, a manufacturing system it describes as AI-native and run from code, after six years in stealth. Its first product, PrimeSwitch PS-150, is a 150-amp relay the company says opens in 50 microseconds and is aimed at the 800-volt DC power systems AI data centers are moving to. The company says the PS-150 is shipping to early-access customers for evaluation.
A note on sourcing. We read Atomic Machines’ release of 7 October 2026 in full as reprinted by Yahoo Finance, and the company’s Matter Compiler, PrimeSwitch and company pages and two blog posts in full; the figures and statements are the company’s and its named investor’s own. We did not read press coverage of the launch, and we did not contact Atomic Machines or its investors. Nothing here is a forecast, and nothing here is financial or investment advice.
Gennaro Cuofano is a business analyst with an International MBA in corporate finance and business strategy (2012). He leads research at FourWeekMBA, which he created, and at The Business Engineer (businessengineer.ai), his newsletter on AI and business strategy, with over 90,000 subscribers and more than 660 published analyses. His work breaks down how companies make money, read from their own filings: revenue and profit structure, capital allocation and ownership. FourWeekMBA reached about four million business people, including C-level executives, investors, analysts and product managers, in 2022 alone. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
