Organic build-up substrates carried advanced packaging this far. Large AI packages now push warpage, flatness, and routing density into a materials problem organic cores struggle to solve. Glass-core substrates—with through-glass vias (TGV)—are the industry’s bet for the next carrier: stiffer, flatter, and better matched to the thermal and electrical demands of big chiplets.
Mid-2026 turned that bet into CapEx. Equipment makers pulled pilot timelines forward; OSATs and substrate houses announced TGV lines and glass-core JVs aimed at qualification windows in 2026–27, with volume still gated by yield.
What glass is meant to buy
- Dimensional stability — Less warpage on large panels and packages.
- Routing headroom — Denser, cleaner vias for high-speed AI fabrics.
- A path beyond organic limits — When package size outruns conventional cores.
What still decides volume
TGV hole quality, metallization, and crack control dominate yield. Inspection and handling of brittle glass are not free. Panel-level flows (including CoPoS-class platforms) and glass cores are related but not identical races—glass can fail even when the panel format succeeds. Customer qualification, not conference demos, sets the clock.
What to watch next
- Pilot-line yield data on TGV and glass-core stacks from major OSATs and substrate JVs.
- First AI-package customer qualifications that name glass core as the production carrier.
- Whether equipment lead times for drill, etch, and inspection become the bottleneck.
HBM fills the stack. GAA builds the logic. Glass tries to carry both without warping the package.