One of the most overlooked parts of the AI buildout is that Ethernet switching is growing 49% annually versus 36% for AI servers as larger GPU clusters require increasingly powerful networking.
That’s great for $MRVL , ALAB, $CRDO and $AVGO because every new generation of AI chips needs faster connections to move data between GPUs, memory and entire server racks.
ANET is also interesting because it holds ~25% of AI Ethernet switching as hyperscalers are adopting Ethernet to connect massive AI clusters.
That’s great for $MRVL , ALAB, $CRDO and $AVGO because every new generation of AI chips needs faster connections to move data between GPUs, memory and entire server racks.
ANET is also interesting because it holds ~25% of AI Ethernet switching as hyperscalers are adopting Ethernet to connect massive AI clusters.