Nvidia CEO Jensen Huang stated that advanced reasoning and agentic AI systems may require up to 1 billion times more computing power than earlier language models. This shift in demand is driven by the transition from simple word prediction to multi-step problem solving and autonomous task execution.
The company is focusing on its next-generation Vera Rubin platform, which integrates GPUs, custom CPUs, networking, and storage. Nvidia targets a 10-fold increase in inference efficiency with this system, aiming to maximize performance per watt in an environment where power is a critical constraint.
Nvidia projects at least $1 trillion in revenue from its Blackwell and Rubin platforms by 2027. Huang also forecasts trillions of dollars in total spending across the broader AI infrastructure ecosystem, though he notes that physical bottlenecks could impact growth rates.
Huang identified power availability, land use, and facility construction as primary factors that may limit AI infrastructure development through 2027. While demand remains high, the actual pace of deployment depends on how quickly utilities and builders can deliver necessary resources.
