Huawei Plans 2027 Mass Production of Glass-Substrate AI Chips, Partners with BOE

Huawei plans to adopt glass substrate technology, targeting mass production by 2027. The company has collaborated with domestic supply chain companies to map out a glass substrate mass production roadmap, with participants including BOE, Visionox and other local manufacturers. At the same time, Huawei will import some key materials from South Korean suppliers to ensure process maturity and supply stability.

Glass substrates offer several technical advantages over traditional organic packaging substrates. Their extremely high flatness facilitates higher-density three-dimensional stacking packaging, and they are less prone to warpage under high-temperature processes, thereby improving yield. The insulating properties of glass help reduce current leakage, improve power consumption, and support higher data transfer rates, providing stronger computational support for large-scale parameter models.

Against the backdrop of long-term restrictions on EUV lithography machines, Huawei views glass substrates as a way to compensate for its process disadvantage through advanced packaging and substrate innovation. By increasing compute density and energy efficiency, Huawei hopes its AI chips will achieve more competitive performance metrics in high-load scenarios such as training and inference of large language models.

Huawei's overseas expansion still faces realistic obstacles such as U. S. sanctions.

With Nvidia largely withdrawing from the Chinese market due to regulatory and geopolitical factors, Huawei has captured about 60 percent of the domestic AI chip market. This advantage mainly stems from the absence of competitors. As global data centers expand AI computing power, they are increasingly focusing on energy consumption, actively seeking hardware and packaging forms that reduce power consumption under equivalent compute. The market has shown a shift toward new module forms such as SOCAMM2, which integrate faster and lower-power LPDDR5X DRAM chips to reduce overall energy consumption and improve bandwidth.