Custom circuit boards, designed and manufactured with AI - turning a months-long hardware project into days of work.
Diode Computers, Inc. - New York, NY. The founding bet: a circuit board is not a drawing, it is code. Photographed here is the company's diode-symbol mark, the emblem of a startup rebuilding electronic design automation from the compiler up.
For decades, designing a printed circuit board has looked more like drafting than programming. Engineers push shapes around a canvas, export files by hand, mail them to a fabricator, wait, and revise - a loop that can stretch a single board to a full quarter. Diode Computers, a New York startup founded in 2024, argues the whole ritual is a tooling problem in disguise.
The company's answer is to treat a schematic as code. In Diode's world, a board is described in a compact, Python-readable language, compiled by a Rust engine in milliseconds, checked against manufacturing constraints, and handed off to fabrication - all inside a workflow that an AI agent can read and iterate on while a human engineer signs off. The pitch is direct: "Hardware can move at the speed of software. We're writing the compiler that makes it so."
That thesis attracted serious money. In 2025, Diode announced an $11.4 million Series A led by Andreessen Horowitz, with Y Combinator, Caffeinated Capital, and BoxGroup joining. The company had already passed through YC's Summer 2024 batch, and its early customer list - physical-AI lab Physical Intelligence and defense maker Saronic among them - suggested the demand was real, not theoretical.
Legacy EDA workflows are fragmented: manual file exports, slow fabricator back-and-forth, and long revision cycles that cap most teams at roughly one board per quarter. Diode collapses that loop.
Where incumbents lock designs inside proprietary file formats, Diode is open-source and export-friendly - so a design is never trapped.
Finish a months-long design project in days, and then ship working boards straight to your door.
An MIT-licensed, Rust-based CLI that compiles schematics from code in milliseconds, manages dependencies, runs deep validation, and generates KiCad layouts.
A Starlark/Python-flavored language for describing PCB schematics as code - legible to both large language models and human engineers.
Compiles and exports to KiCad, Altium, Cadence, BRD, ODB++, and IPC-2581, so teams keep full portability and avoid vendor lock-in.
Matches every component to distributor and manufacturer catalog items with real-time cost and availability data.
The pcb.new / pcb.store service takes a spec to a designed, verified, and manufactured board shipped to your door - with expert human review.
Open-source tooling builds the community and standard; paid, per-project design-and-build engagements (roughly $10K-$50K plus fab costs) monetize the work.
The 2025 Series A was led by Andreessen Horowitz, with participation from Y Combinator, Caffeinated Capital, and BoxGroup. Funds go toward AI design algorithms, validation, and growing the engineering team.
Early teams report cutting design revisions from five iterations down to one.
Six years building hardware at big tech and medtech - including robotics at Apple's Special Projects Group, plus Butterfly Network and Chromatic - before founding Diode.
Co-founder and CTO, leading the compiler and platform engineering behind Diode's code-first EDA stack.
Rust for the compiler, Python and TypeScript across the toolchain, Next.js on the web, and AI agents for design generation - all with human-in-the-loop review.
Electronic design automation is dominated by heavyweight suites - Cadence Allegro, Altium Designer, Siemens' Mentor tools - alongside the open-source stalwart KiCad. Those tools are powerful and entrenched, but they were built for a pre-AI era of manual, GUI-driven design.
Diode sits with a newer cohort of code-native EDA projects such as atopile and tscircuit, which also describe circuits as text. Its differentiator is the full loop: not just a language, but a millisecond Rust compiler, manufacturing-aware constraints, live component sourcing, and an actual manufacturing service that ships boards. In the emerging "physical AI" era - robots, drones, autonomous systems - the bottleneck is rarely compute; it is the hardware iteration loop. Diode is betting the company on unblocking exactly that.
Davide Asnaghi and Lenny Khazan launch Diode in New York to automate circuit board design with AI.
Diode joins YC's S24 batch and releases its open-source, code-first PCB toolchain and the Zener language.
Andreessen Horowitz leads an $11.4M Series A to scale the AI design platform and grow the engineering team.
The company logo is a stylized diode circuit symbol - the name is the mark.
Its schematic language, Zener, is named after the Zener diode and built on Starlark, the same family Google's Bazel uses.
CEO Davide Asnaghi worked on robotics inside Apple's secretive Special Projects Group.
Diode openly credits earlier code-native EDA projects like atopile and tscircuit as inspiration.
The compiler recompiles a schematic in the time it takes to blink - milliseconds.
Diode builds an AI-powered, code-first platform for designing and manufacturing custom printed circuit boards, taking projects from spec to delivered board in days instead of months.
It was founded in 2024 by Davide Asnaghi (CEO), previously a robotics engineer at Apple's Special Projects Group, and Lenny Khazan (CTO).
Diode raised an $11.4M Series A led by Andreessen Horowitz, with participation from Y Combinator, Caffeinated Capital, and BoxGroup.
Yes. Diode's pcb toolchain and Zener language are open source under the MIT License, and designs export to KiCad, Altium, and Cadence to avoid vendor lock-in.
Hardware teams in physical AI, robotics, aerospace, and defense - including named customers Physical Intelligence and Saronic - use Diode for fast, verified board design and manufacturing.