💰 Nvidia CEO Jensen Huang projects $3T-$4T by 2030 for AI infrastructure, saying demand is accelerating.
i.e. GPUs, memory, storage, ultra-fast networks, power and cooling upgrades, buildings, and software that keeps the cluster busy.
Nvidia has indeed seen data center revenue rise by 10x in the past 2 years.
Earlier in the year, Nvidia unveiled its AI roadmap, including next-gen Vera Rubin architecture in the second half of 2026, and beyond that, Feynman. During that keynote, Huang also predicted data center builds will surpass $1 trillion in expenses by 2027.
Two forces drive spend, training that needs more memory and tighter interconnects, and inference that scales with users and tokens, with inference often dominating the bill once products launch.
Large buyers anchor the curve since large cloud providers like Microsoft and Amazon are flagged for about $600B of data center capital spending, which sets a high base of orders.
---
pcguide .com/news/nvidia-ceo-jensen-huang-expects-3-trillion-to-4-trillion-spend-on-ai-infrastructure-by-2030/