---
title: "The Chinese glass maker betting it can replace the board beneath AI chips"
date: 2026-11-12
category: Chips & Compute
site: NeuroAI
canonical: https://neuroai.site/a/na-chips-glass-substrate-tgv
language: en
---

# The Chinese glass maker betting it can replace the board beneath AI chips

> WG Tech and its Tonggewei unit are pushing glass substrates (TGV) from display into advanced chip packaging, signing an AI-chip partner as China hunts the next substrate beyond organic boards.

For years the most advanced part of a chip package was the silicon. A quieter contest is now over what the silicon sits on. A Chinese glass maker best known for TV backlights is trying to turn that humble layer into the material that lets AI chips pack tighter, run cooler, and cost less.

## Key takeaways

- WG Tech (沃格光电) and its subsidiary Hubei Tonggewei (通格微) are advancing through-glass-via (TGV) glass substrates, showing the technology at CES 2025 and positioning it for advanced semiconductor packaging.

- Tonggewei's first TGV glass-substrate line, rated at about 100,000 square meters a year, has moved into small-batch supply after a trial run, according to the company and industry coverage; it is working with AI-chip designer Polarctic (北极雄芯) on an all-glass multi-layer interconnection AI chip.

- Glass substrates promise better electrical, thermal, and flatness properties than organic boards, which matters as AI and high-performance-computing chips push interconnect density and power higher.

- The company says it is among a small number of firms globally with full-process TGV capability; the technology is still moving from validation to scale, with cost and standardization the open questions.

## Why the board under the chip matters

In an advanced package, the silicon die is only half the story. It rests on a substrate that routes power and signals, and for the densest AI packages that substrate is increasingly the bottleneck. Organic boards warp under heat, limit trace density, and get expensive as sizes grow. As accelerators stack more high-bandwidth memory and compute into one package, the substrate's electrical and thermal behavior starts to cap performance.

That is why the industry has been eyeing glass. A glass substrate can be flatter, more thermally stable, and capable of finer interconnect than the organic alternative — exactly the properties that dense AI packages crave. The catch is that glass is harder to process at volume, which is why the technology has lingered in labs and niche displays.

## What WG Tech and Tonggewei are actually building

WG Tech is not a typical semiconductor name. It made its name in glass-based Mini/Micro LED backlights for displays. But the same glass-precision skills — thinning, photolithography, via formation, metallization, and multi-layer circuitry — translate directly into substrate manufacturing. At CES 2025, held January 7–10 in Las Vegas, the company showcased its glass-based Mini/Micro LED and TGV technologies, explicitly framing TGV for advanced semiconductor packaging.

Its Hubei Tonggewei unit is the spearhead for the chip side. According to the company and industry coverage, Tonggewei's first TGV glass-substrate line, sized at about 100,000 square meters a year, completed a trial run and has entered small-batch supply. The parent says it is among a small number of firms globally with full-process TGV capability, covering everything from glass thinning to via filling and multi-layer fabrication. We treat that "among a small number" claim as the company's own positioning rather than an independent ranking.

## The AI-chip hook

The detail that lifts this from a materials story to a compute story is the partnership with Polarctic (北极雄芯), a Xi'an-based chiplet-focused design house. The two announced work on an all-glass, multi-layer interconnection AI chip — a design that drops the organic or hybrid board entirely and stacks glass layers instead. The stated appeal is avoiding the warping and cost that come from mixing materials under heat, while the glass flatness allows tighter layout and larger packages.

It is, for now, an announced collaboration rather than a shipped product. But the direction fits a clear need: as AI chips demand ever-finer, higher-density interconnect, a substrate that behaves better under stress is not a luxury, it is part of the performance envelope. A domestic glass-substrate option also diversifies China's packaging supply chain away from imported organic substrate capacity.

According to Li An, chief scientist at BrainNet (脑机网), glass substrate is a longer-horizon swing — promising on paper for AI density, but its fate hinges on whether it can match organic boards on cost at volume. Li An's comments reflect his own views and do not represent NeuroAI's editorial position or an endorsement by BrainNet (脑机网).

## Why glass, why now

Two trends converge. First, AI and HPC are pulling advanced-packaging demand — and substrate demand — sharply upward, making any credible new substrate path worth pursuing. Second, China is actively widening its packaging and materials base as a hedge against restricted equipment and imported materials; a home-grown glass substrate is a natural fit for that strategy. Internationally, larger players have also been developing glass substrates, so the technology's time may finally be arriving rather than perpetually "five years away."

## Honest limitations

This article relies on company statements from WG Tech and its newsroom plus industry-media coverage (Cailianpress); we have not independently verified production yield, actual shipment volumes, or customer adoption beyond the announced Polarctic collaboration. The "among a small number of firms globally" claim is the company's own and is presented as such. The all-glass AI chip with Polarctic is an announced partnership, not a commercial product we can confirm has taped out or shipped. Cost competitiveness and standardization at scale remain unproven and are the variables most likely to decide whether glass substrates move from samples to mainstream AI packaging.

## What readers can do now

- Watch whether Tonggewei's TGV line converts small-batch supply into volume orders from AI or chip-packaging customers — that, not the showcase, is the real milestone.

- Track the broader substrate race: glass versus organic and advanced formats from larger global players, since the winner will shape AI-package cost and density for the next decade.

- For materials and packaging engineers, glass TGV (via formation, metallization, multi-layer stacking) is an emerging skill area worth building now, ahead of likely demand.

---

Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-chips-glass-substrate-tgv
Free to quote with attribution and a link to the original.
