AI Server Platforms Drive Ultra-Low-Loss Adoption
Nvidia's latest accelerator platforms and competing AI server designs increasingly specify M7 and M8 grade laminates capable of supporting 224 gigabit SerDes lanes without excessive signal attenuation, a requirement conventional FR4 materials cannot meet at these frequencies. Taiwan Union Technology and Elite Material both expanded ultra-low-loss production capacity meaningfully across 2025 as hyperscaler orders accelerated ahead of next-generation platform launches. This shift is pulling laminate demand into a specification tier that barely existed commercially five years ago, forcing suppliers without ultra-low-loss capability to either invest quickly or risk losing approved-vendor-list positions to competitors who moved earlier.
Market Impact: Data center capex rose 27 percent








