Theses›Semiconductor manufacturing›Known-good-die test

Known-good-die test

2.5D/3D integration makes known-good-die testing mandatory (one bad die scraps a multi-die package worth thousands) and AI-die power forces liquid-cooled / high-current burn-in.

  • Bear
  • Closing
  • $7.7B by 2035

The callStacked AI packages make bare-die test mandatory, and the four incumbents are already shipping the liquid-cooled burn-in that looked like the opening for a newcomer.

D1D2R1R2R3C1foragainst D1D2R1R2R3C1foragainst
Fig. 1The balance. Drivers and catalysts against restraints, each block sized by its weight; catalysts hang dashed because they have not happened yet, sized by their odds. It tips toward the restraints.
  1. D1An AI package with several logic dies and HBM is worth thousands of dollars, so one bad die scraps the lot and every die must be tested first.
  2. D2Today's testers and burn-in ovens were built for packaged chips, not bare AI dies that run hot, draw high current and need liquid cooling and die-level contact.
  3. R1The equipment layer is incumbent-captured and the incumbents already ship the opening: Aehr's Sonoma is liquid-cooled wafer-level burn-in.
  4. R2No entrant has got into the top four in 30 years, and the only seed-shaped opening is a sovereignty-contingent European entrant that doesn't exist yet.
  5. R3M&A actuals are modest: Cascade at 2.4x, Harbor 1.0x, Xcerra 0.8x revenue.
  6. C1Multi-vendor chiplet sourcing reaches volume, so known-good-die certification becomes a separately specified and purchased test step rather than part of the OSAT's flow.2030
Rests onthe beliefs underneath
Supported by · confidence
What would prove itdated tests, open and resolved
20272028202920302031today12345predictionswatch 20272028202920302031today12345predictionswatch
Fig. 2Each ring is a dated test: dashed while open, struck through when overdue, ticked or crossed once resolved.
  1. 1By end-2027 a tier-1 OSAT or memory maker publicly names known-good-die wafer-level test / burn-in as a capacity bottleneck or a dedicated capex line for AI 2.5D/3D packages.Dec 2027 · open
  2. 2A tier-1 OSAT or memory maker publicly naming KGD wafer-level test or burn-in as a bottleneck or a capex line.Dec 2027 · open
  3. 3KGD / wafer-level burn-in becomes a named gating requirement for AI-die 2.5D/3D packages by 2028, with the AI/HBM-driven burn-in sub-line growing faster than the ~13-15% segment trend.Dec 2028 · open
  4. 4The AI burn-in sub-line outgrowing the segment.Dec 2028 · open
  5. 5Multi-vendor chiplet sourcing at volume, with KGD certification bought as its own step.Dec 2030 · open
Expressionslisted, private, technologies
Private · 3 private companies, 2 of them in Germany.

TYLsemi · LPKF Laser & Electronics · Keystone Photonics

1 more private company tracked, not shown

Technologies
Who has already bet on itprivate rounds, from public sources

raised by 1 company, $51M of it in the last three years. Wave building: Money is arriving: recent rounds, but no runaway leader yet.

  • ·Keystone Photonics Rounds not itemised
What changedthe record, newest first
  • 23 Jun 2026The bottleneck in AI chips moved to packaging and high-power burn-in, and the testing step that catches a bad die before it scraps a thousand-dollar stacked package is now a named yield gate, not an afterthought. The catch for an investor is that the obvious pure-play already exists and is listed, shipping the first production burn-in systems for AI processors, so the clean way to express this is the public name, not a hunt for a startup.