Skip to content
Welcome to our new website — take a look around and enjoy. Meet us at LANE 2026 in Erlangen · 21–25 September Conference details →
Use cases/White paper/Through-Glass Vias for AI packaging
White paper · 3D Beam Shaping

Through-Glass Vias: the optics-ready foundation for next-generation AI packaging.

Glass interposers with Through-Glass Vias are the platform the >$100B HBM market is converging on. But they only work if you can drill TGVs at production yield. Conventional optics were never designed for it. Our white paper makes the case for 3D beam shaping as the drop-in, scanner-compatible, fab-ready answer.

Market: AI / HBM packaging Process: TGV laser drilling Products: 3D Gaussian · Ring-Core · Top-Hat
The TGV bottleneck
  • Glass is the next interposer: larger panels (500–700 mm), ~50% less warpage, better signal integrity at lower cost than silicon
  • The enabling process: copper-filled vias 20–50 µm in diameter, aspect ratios 1:4 to >1:50, at yield across full panels
  • On a 10,000-via panel a 1% defect rate can scrap the unit: advanced beam shaping reports 1–5% absolute yield gains and 50–200% higher throughput
Why conventional optics fall short
  • Gaussian: ~40 µm depth of focus for a 5 µm spot; drilling 300–500 µm glass gives 5–15° taper and focus-drift yield loss
  • Bessel / axicon: extended depth, but alignment-sensitive, not scanner-compatible (10–50× slower), lab-only
  • The missing piece is a beam profile that stays consistent on the production floor
The beam physics that matter (extended DOF, flat fluence, plasma venting) are all achievable with stable, drop-in reflective optics that keep full scanner throughput.
One integration principle, three profiles: 3D Gaussian (500% DOF), 3D Ring-Core (100% DOF) and 3D Top-Hat (60% DOF)Midel's 3D beam-shaping line-up for TGV drilling: the 3D Gaussian with 500% Gaussian depth of focus, the 3D Ring-Core with 100%, and the 3D Top-Hat with 60% — each shown as a simulated caustic with measured profiles
One architecture, three profiles

Matched to the challenge you lead with

All three share one integration principle: a single drop-in reflective DOE in the collimated beam path, before the focusing lens. No changes to laser, scanner, software or alignment. Scanner-compatible at full production speed. Figures from the white paper, confirmed per configuration.

Full white paper

Get the complete technical case

Read the full analysis: the AI-memory imperative, the glass-interposer bottleneck, and a side-by-side approach comparison. It also covers how each of our three shapers maps to the TGV challenges of depth of focus, taper, plasma evacuation and optical efficiency.

  • The $100B HBM driver and why glass wins
  • Where Gaussian, Bessel and axicon approaches break down
  • A full approach-comparison matrix: conventional vs. 3D
  • Drop-in integration: from lab to production floor

Ready to solve TGV drilling at yield?

Tell us your panel, laser and via geometry. We’ll map it to the right 3D profile and scope a first project.

Discuss your challenge