Evidence-first AI semiconductor analysis: what new silicon claims prove, where bottlenecks move, and whether gains survive at system and economic scale.
I’m starting a closer look at AI power semiconductors, with $NVTS as a case study. The question that caught my attention: what does a new foundry partner actually de-risk?
GF isn’t starting from scratch: its 2024 Tagore acquisition brought Power GaN IP and engineering talent.
That gives Navitas a relevant manufacturing foundation. What I’m watching next is evidence of qualification and repeatable production for the new GF-made products.
I keep coming back to the power side of AI because there are still companies down here that barely get discussed.
Navitas $NVTS is starting the first shipments of its US-made Gen 5 GaNFast chips this month, manufactured on GlobalFoundries’ 200mm GaN-on-silicon line in Vermont.
Internal samples come in October, strategic customer samples before year-end.
I’m not saying this suddenly makes NVTS a must-own.
But AI racks keep getting denser, power conversion keeps getting harder, and Navitas is slowly moving from “cool GaN story” toward actually having domestic production behind it.
One I’m watching pretty closely.