AMD took AI GPU share from Nvidia. Can Marvell Technology do the same to Broadcom in custom ASICs & networking?

We’ve seen AMD establish itself as the primary alternative to Nvidia in the AI chip space. Looking at custom silicon and interconnects, Broadcom currently holds a dominant position with major hyperscalers. ​However, Marvell Technology seems to be positioning itself in a very similar way, gaining serious traction in custom ASICs (working with AWS, Google, Microsoft) and electro-optics/DSPs.​ Beyond data centers, custom ASICs and specialized low-latency networking are also becoming essential for the robotics and edge AI boom, where autonomous systems rely heavily on real-time sensor fusion, microsecond-level control, and high power efficiency.