Solutions
By Design Type
By Industry
Hardware-Rich Development
Designed by
x
Manufactured by
%201.svg)
Live Webinar
What It Took to Manufacture the First AI-Designed Computer
Thursday, June 25, 2026
10:00 AM PT / 1:00 PM ET
Live with Q&A
Project Speedrun proved what physics-driven AI can do: a complete, Linux-capable motherboard, fully validated with DDR4, Ethernet, USB, HDMI, and graphics support, designed in a fraction of the time a human team would need.
But a layout that passes in software still has to survive a real manufacturing floor.
So we sent two of those boards — an i.MX 8M Mini System on Module and its carrier BaseBoard — to Sierra Circuits. They found 58 issues across the two boards: fab capability gaps, documentation holes, footprint mismatches, BOM errors, and a 2mil trace challenge that pushed a top fab's finest process to its limit.
This session will walk you through the real handoff process between design and manufacturing, told through real boards and a real fab.
What You Will Learn
01
Published capabilities vs. DFM in practice
Why a fab's spec sheet says 2mil trace and 4mil laser drill while the DFM team flags below 3mil and asks for 7mil drill-to-copper, and how to close that gap before you order instead of after.
02
How fine-pitch boards actually get built
Laser-drilled and copper-filled vias, hybrid drilling, and the mSAP process that make 2mil routing for an i.MX 8M Mini BGA fanout manufacturable, plus the impedance tradeoffs that come with it.
03
Why documentation is the silent killer
18 of 58 issues, 31% of the total, were documentation gaps: missing impedance tables, unclear stack-up, ambiguous via-fill specs. We show how automating fab note generation eliminates them.
04
Where footprint and BOM mistakes hide
The most common assembly issues are footprint and BOM mismatches. See the real cases Sierra's team caught, from a single part number with two footprints to a wrong-spec alternate, and how library validation stops them before they cost a build.
05
What "manufacturable" really means for AI-designed hardware.
The honest checklist for taking an AI-designed layout from done in software to assembled and working in the lab.
Who Should Attend
PCB design and layout engineers
DFM and DFA engineers
Manufacturing and NPI engineers
Hardware engineering leaders evaluating AI for layout
Anyone who has ever lost a build to a documentation gap or a footprint mismatch
Presenters
Ben Jordan
Staff Electrical Engineer, Quilter
Ben brings more than 20 years in EDA to Quilter, where he designs real boards with the use of physics-driven AI, among other things. Ben led Project Speedrun, the first AI-designed computer, and will walk you through what it took to get it manufactured.
Amit Bahl
Chief Revenue Officer, Sierra Circuits
Known across the industry as "The PCB Guy," Amit has spent more than 15 years helping designers and engineers master design for manufacturing. He brings the fab's point of view: what Sierra's CAM and DFM teams flagged, why, and how to design so they never have to.
© 2026 Quilter AI. All rights reserved.