Jensen Huang Warns Agentic AI Will Lift Compute Demand 1,000% in 2 Years
Updated
Updated · 24/7 Wall St. · May 16
Jensen Huang Warns Agentic AI Will Lift Compute Demand 1,000% in 2 Years
3 articles · Updated · 24/7 Wall St. · May 16
Huang said at ServiceNow’s Knowledge 2026 conference that agentic AI will require at least 1,000% more compute than generative AI within two years, as autonomous systems run multi-step tasks for minutes or hours.
That jump lands on a grid already under pressure: U.S. data centers draw about 41 gigawatts, up 150% in five years, and could consume 325 to 580 terawatt-hours by 2028—up to 12% of U.S. electricity use.
Utilities and developers are already feeling the strain. Dominion won a 2026 household rate increase of $11.24 a month, while some AI projects are being delayed or turning to on-site natural gas generation when grid connections lag.
Tech companies are locking in power supplies as capex surges. The IEA said conditional data-center deals tied to small modular reactors grew from 25 gigawatts at end-2024 to 45 gigawatts by April, while the four biggest cloud groups plan more than $710 billion in 2026 AI infrastructure spending.
The broader takeaway is that efficiency gains are not offsetting demand growth: global data-center electricity use is projected to rise from 485 terawatt-hours in 2025 to 950 terawatt-hours by 2030, pushing AI’s impact beyond chips into utilities, nuclear and power markets.
AI promises a digital revolution, but will its massive physical footprint drain our water and power resources dry?
The U.S. and China are racing for AI dominance. Who will win the high-stakes energy game that powers it?
The 1,000% Compute Boom: How Agentic AI Is Reshaping Infrastructure, Industry, and Society
Overview
The report highlights a major transformation in the AI landscape, marked by the shift from generative AI to agentic AI. Unlike previous models that simply generate content, agentic AI can act autonomously, reason, plan, and execute complex tasks, requiring multiple steps and iterations to achieve goals. This evolution is driving a massive surge in computational demand—up to 10 times more than before—according to Nvidia’s CEO. As each decision and learning cycle by agentic AI adds to the computational load, the industry is responding with large-scale infrastructure investments, signaling a new era of technological and economic change.