Nvidia CEO Jensen Huang stated that advanced reasoning and agentic AI systems may eventually require up to 1 billion times more computing power than earlier models. This shift is driven by the transition from simple next-word prediction to complex problem-solving that involves multi-step reasoning and self-verification.
The company is developing the Vera Rubin platform, which integrates GPUs, custom CPUs, networking, and storage. Nvidia targets a 10x efficiency gain in inference for this system, aiming to maximize performance per watt as power availability becomes a critical constraint for data center development.
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 noted that physical bottlenecks like power, land, and construction may limit growth through 2027.
While demand for AI compute remains high, Nvidia acknowledges that the actual pace of infrastructure expansion depends on the speed at which utilities and builders can deliver necessary resources. Investors are advised to monitor deployment timelines and real-world efficiency metrics.
