The analyst contends that hyperscalers like Google, Amazon, and Meta are heavily investing in their own custom AI accelerators (TPUs, Trainium, Meta's ASICs) and interconnects to reduce reliance on a single supplier. This vertical integration strategy is driven by engineering and cost optimization at the massive scale of their AI spending.
Hyperscalers' increasing vertical integration into custom AI silicon will shift compute supply away from NVIDIA, creating new competition for GPU market share.