Compute and Consequence:
AI Energy Demand in a Rapidly Evolving Grid Landscape
Technology companies have been leaders in corporate clean energy procurement, helping accelerate the deployment of wind and solar across the world. However, technology companies’ rush to lead in the development of Artificial Intelligence (AI) is now driving investment in new fossil-based gas plants and even extending coal power generation to meet projected energy demands of new and proposed data centers.
Between January 2023 and January 2025, new gas-fired power plants actively proposed by utilities and independent developers jumped by 70% to nearly 200 gigawatts (GW). Many of the proposed gas projects are directly associated with data center demand. If these plans are fully realized, today’s gas fleet would be increased by nearly one-third.
While many technology companies are claiming 100% renewable or clean energy goals, those goals are often achieved through matching consumption of energy with renewable energy certificates (RECs). However, many RECs do not incentivize new renewable capacity or, due to time of use and location discrepancies, fail to sufficiently cancel the emissions from data centers reliant on gas or coal-fired electricity. This disconnect between greenhouse gas (GHG) emissions from operating and planned data centers and the emissions reductions actually achieved through RECs can undermine the climate goals of both technology companies and power utilities.
Amplifying this issue is significant uncertainty and speculation around the number of new data centers actually needed to serve for decades to come. growth in AI. Similar to prior technology booms, as the market for new technology solidifies beyond its frothy stage, actual demand will likely be much less than the public hype suggests. Not only are demand projections currently overestimated due to companies seeking bids from multiple providers, but competition will likely winnow the number of AI competitors with successful market offerings. Further, the projected hourly energy needs of data centers are likely to be reduced substantially through demand flexibility and operational efficiencies still to come.
Investors have a great deal at stake in understanding these dynamics. If surging electricity demand is overestimated and is met primarily by expanding gas-fired generation and maintaining coal power, the U.S. risks locking in tremendous GHG emissions for decades to come. This potentially dramatic expansion of fossil power comes at a time when rapid decarbonization is necessary to prevent climate-related catastrophes from becoming worse and permanently locking in market value destruction. This paper provides perspective on the AI boom and its energy and risk implications and offers a practical framework through which investors, technology companies, and utilities can work together to evolve best practices and drive clean energy solutions aligned with grid decarbonization.