GaN is winning the AI power argument, not yet the cost battle

Gallium nitride (GaN) is gaining traction in AI data centers, yet cost advantages remain elusive. Technologically, GaN's high power density and efficiency are pushing server power architectures toward higher voltages and compact designs, compelling upstream suppliers of substrates and epitaxial materials to boost yields and capacity. However, silicon-based solutions still dominate due to cost and maturity. Geopolitically, tightening export controls on semiconductors from Taiwan, China and the U.S. are complicating GaN supply chains, forcing companies to seek alternative sourcing strategies. In competitive dynamics, established players like Infineon and STMicroelectronics are accelerating their GaN rollouts, while startups are targeting niche segments. Over the next 12–24 months, GaN will gradually penetrate high-end AI servers, but widespread adoption hinges on further cost reductions and technological maturation.