Microsoft deploying multiple gigawatts of homegrown Maia 300 chips across data centers to reduce Nvidia GPU dependence

Third-generation Maia designed as primary inference system; rollout begins next year as Microsoft pursues vertical integration in AI hardware.