NVIDIA is working with companies across the supply chain to advance the adoption of glass substrate technology, aiming to address the growing physical limits of AI chip packaging in terms of size, heat dissipation, and interconnection. If the technology breaks through, glass substrates could support larger-scale HBM integration, boosting packaging density and data transmission capabilities for next-generation GPUs.
According to reports, German equipment maker SCHMID disclosed during a recent first-half earnings briefing that it is collaborating with key companies in the respective supply chains of Intel, NVIDIA, and AMD to jointly develop glass core substrate equipment. SCHMID Chief Strategy Officer Roland Rettenmayer said that technical obstacles remain in the through-glass via (TGV) metallization process, and end-customer certification has not yet been completed.
NVIDIA CEO Jensen Huang has previously engaged with multiple parties on this technology roadmap. In a meeting with SK Group Chairman Chey Tae-won, he discussed next-generation semiconductor cooperation including glass substrates, and actively pushed TSMC to accelerate related technology development. This indicates that NVIDIA is evaluating the feasibility of introducing glass substrates into next-generation GPUs.
Glass Substrates Target AI Chip Packaging Bottlenecks, Poised to Expand HBM Integration
As AI chip performance and power consumption continue to rise, the limitations of traditional plastic substrates in heat resistance and warpage control have become more pronounced. Glass substrates offer higher flatness and strength, a lower coefficient of thermal expansion, less deformation under high-temperature environments, and also possess favorable dielectric properties and lower signal loss, making them a potential solution for next-generation high-performance chip packaging.
Another major advantage of glass substrates lies in their ability to scale packaging size. Current mainstream AI chips typically use silicon interposers to integrate GPUs and HBM side by side, but silicon interposers face constraints in size scaling and manufacturing costs. Glass substrates could partially or even completely replace silicon interposers, accommodating more chips and HBM within the same package, thereby increasing packaging density and shortening data transmission paths.
For AI chips, this means larger-scale HBM integration and higher data throughput. HBM enhances bandwidth by vertically stacking multiple layers of DRAM and is a critical memory component in current AI accelerators. As model sizes and computing power demands continue to grow, how much HBM a package can accommodate has become an important factor limiting AI chip performance.
TGV Metallization Remains Key to Mass Production, Commercialization Not Yet Complete
However, glass substrates still need to overcome technical and certification hurdles before large-scale commercialization. SCHMID explicitly pointed out that the TGV metallization process remains one of the main technical obstacles. Glass itself is non-conductive, and how to form stable, reliable through-vias with high-density interconnection capability inside the substrate is key to achieving mass production.
At the same time, the end-customer certification process has not yet been completed, meaning the related technology is still in the industrialization advancement stage. SCHMID is currently developing the equipment and solutions needed for glass substrate production, while simultaneously expanding its AI server motherboard and next-generation packaging equipment businesses. For NVIDIA and its supply chain, whether glass substrates can ultimately enter large-scale GPU production still depends on whether manufacturing processes, reliability, and customer certification can further mature.
Comments