Tiaan Coertzen

Elon musk said AI agents are limited by memory. But, the main limiter is actually the network card (NIC) that moves packets between machines. Let me explain my full thesis and the stocks I'm buying: In traditional AI (especially large scale training or single model inference), performance is primarily limited by compute (FLOPs) and on chip/high bandwidth memory (HBM). Agentic AI is different though. Agentic AI is the systems that autonomously plan, call tools/APIs, retrieve external data (e.g., RAG), maintain state across steps, collaborate with other agents, and execute multi-step workflows, etc.. This is obviously the future. But the dominant bottleneck shifts to the network interface card (NIC) and the broader interconnect fabric. Super fast GPUs and memory often times sit idle while waiting for packets to move between machines, services, or agents. Creating, a huge bottleneck, and giving us some great investment opportunities. For large caps (more diversified, but still having some exposure): 1) $NVDA 2) $AMD 3) $AVGO 4) $ANET Smaller pure plays: 1) $ALAB 2) $CRDO 3) $MRVL 4) $NAPA My main bet: 1) $SILC Bookmark this post and come back when you realize I was right. $SILC at $50 is a dream. Don't fade my call.