Capital expenditure by the five largest technology companies in AI data centers is anticipated to rise a further 75% this year, pushing total investment past $700 billion. This fuels an escalating demand for electricity. Hyperscale AI data centers alone require 120 kW of rack power, according to IndexBox. Global data center electricity consumption reached approximately 415 terawatt hours in 2023, about 1.5% of the world's total, according to CapacityGlobal. While global renewable energy capacity expands rapidly, AI's unprecedented power demands grow even faster, threatening clean energy targets. Without significant breakthroughs in energy efficiency or rapid deployment of new clean power, the AI boom risks increasing reliance on fossil fuels or straining grids, potentially undermining global climate goals.
The Green Energy Push Meets AI's Appetite
In 2025, over 90% of new electricity capacity worldwide came from renewable sources, according to Egypt Oil & Gas. Major tech companies also maintain ambitious sustainability pledges; Google, for example, commits to running its business 24/7 on carbon-free energy by 2030, according to ESG Today. However, the sheer scale and speed of AI data center expansion suggest this new renewable capacity, much of which is intermittent, is insufficient to meet AI's specific 24/7 dispatchable clean energy demands. This situation may force tech giants to rely on grid power that is not fully carbon-free, undermining their sustainability goals.
Innovative Solutions for AI Power Demands
The extreme power density required by AI drives the development of novel energy solutions. Aalo Atomics is developing mass-manufactured compact nuclear power systems for data centers, aiming for its first controlled nuclear reaction by July 4, 2026, according to Columbia University. These systems offer a potential path to localized, continuous power.
