Place-and-Bond
Die-attach and hybrid-bonding placement control.
Semiconductor Packaging
Chipira
Autonomous operations layer for semiconductor advanced packaging (OSAT) — agentic die-attach and hybrid bonding, AI warpage and void inspection, HBM stacking.
Overview
What Chipira does.
Chipira closes the loop on semiconductor advanced packaging: die-attach and hybrid bonding, warpage and void inspection, HBM die-stacking, yield and ramp intelligence and the package and line digital twin.
Every line deployment produces labelled bond and inspection traces that train the next model, tightening the loop from die to finished package.
| Detail | Value |
|---|---|
| Stage | Seed |
| Category | Physical AI |
| Year | 2026 |
| Website | chipira.com |
| Headquarters | Not disclosed |
| Raised | $1.5M |
| Investor | 10xC |
| Model | Per-line software subscription with a yield-outcome component. |
| Measured by | Stack yield, void and warpage false-call rate, bond placement accuracy, lines expanded per customer. |
Products
One loop, module by module.
Every deployment on chipira.com writes labelled traces back into the model, so the system generalises across plants, alloys, grades and geometries.
Warpage & Void Inspection
AI warpage and void inspection.
HBM Die-Stacking
HBM die-stacking control.
Yield & Ramp
Yield and ramp intelligence.
Package & Line Twin
Package and line digital twin.
Why we invested
The thesis behind Chipira.
Advanced packaging is the growth layer of the chip supply chain, and it is a yield business measured in fractions of a percent. Warpage and voids are found after bonding, when a die stack is already committed.
Chipira moves the read before the bond and the control into software. The buyer is an OSAT or IDM paying in hard currency; the wedge is one inspection or bonding step, and stack yield is the number that opens the next module.
Who is building it
Co-Founder & CEO
Co-Founder & CEOOwns the OSAT relationships, the pilot programme and the commercial model.
Co-Founder & CTO
Co-Founder & CTOOwns the vision and bond models, the edge runtime and the fab-grade audit architecture.
Building something similar?
Building in Physical AI? We want to hear about it.