Interview

Diode Computers raises a Andreessen Horowitz-led Series A to use AI to design and manufacture PCBs in the US

Jul 23, 2025 with Davide Asnaghi

Key Points

  • Diode Computers closes Andreessen Horowitz-led Series A roughly one year after founding, using LLMs to automate PCB design via code rather than traditional GUI workflows.
  • The company trains proprietary models on synthesized circuit data and SPICE simulations, retaining IP rights to generated components while assigning full board designs to customers.
  • Diode operates small-batch assembly at Brooklyn Navy Yard and plans expansion to San Francisco, Austin, and a consolidated facility in Arizona or Ohio to undercut Chinese PCB pricing.
Diode Computers raises a Andreessen Horowitz-led Series A to use AI to design and manufacture PCBs in the US

Summary

Diode Computers has closed an Andreessen Horowitz-led Series A roughly one year after founding, positioning itself at the intersection of AI-generated PCB design and domestic US manufacturing. The company, co-founded by a former Apple custom silicon engineer and Lenny (co-founder), operates out of the Brooklyn Navy Yard and uses large language models trained on proprietary datasets to generate circuit board schematics via code rather than the traditional visual, GUI-based workflow used by incumbents like Altium and Cadence.

The Core Problem

PCB design for high-complexity applications in aerospace, robotics, and medical devices traditionally takes months. The engineers capable of working at that level are concentrated at Tesla, SpaceX, Apple, and Meta, making that talent effectively inaccessible to startups and mid-market hardware companies. Diode's thesis is that automating schematic generation via LLMs democratizes a workflow previously available only to the largest players.

Data Strategy

Unlike software, where GitHub provides an open training corpus, no equivalent public dataset exists for PCB design. Diode is rebuilding its training data from scratch through annotation, synthesis, and simulation, using SPICE-based electrical simulation to bootstrap the dataset. The company is set to announce a partnership with a major open-source hardware project within days of this recording. Diode assigns full IP of customer board designs to clients but retains rights to the individual generated components, using those as a proprietary training asset. The stated competitive moat is not building new foundation models but continuously fine-tuning existing frontier models on higher-quality domain-specific data, including a dedicated error-detection model to distinguish valid from flawed generated designs.

Manufacturing Footprint

Diode currently handles small-batch assembly in-house at Brooklyn Navy Yard and routes high-volume orders to third-party manufacturing houses. Expansion plans include assembly shops in San Francisco and Austin, with a larger consolidated facility targeting Arizona or Ohio for scaled production. The long-term goal is vertical integration of manufacturing, with the explicit aim of making domestic US PCB ordering cost-competitive with China, describing current pricing parity on process simplicity but acknowledging the cost gap has not yet closed.