The Electric Dependency of AI: A Return to Gas
For years, big technology companies have maintained a façade of sustainability, promoting data centers powered by renewable energy and pledging commitments toward climate neutrality. However, as artificial intelligence (AI) surges, this image is increasingly being called into question. Electricity demand is climbing at an unprecedented rate that the current grid struggles to meet, and the fuel filling the void isn’t wind or solar—it’s natural gas.
The Rising Electricity Gobblers
The statistics paint a stark picture. Major players like Google and Microsoft consume around 24 terawatt hours (TWh) of electricity annually—more than the energy needs of over a hundred countries. Despite their announcements of record clean energy contracts, emissions are still on the rise. Google, for instance, has seen a startling 48% increase in emissions over the past five years, while Microsoft’s emissions have surged by 31% since 2020. An independent analysis has criticized these technologies as having “poor” or “very deficient” climate integrity, underscoring that the cloud requires more than just a clean image; it’s a question of energy physics.
Electron Fever: Demand Surge
This phenomenon isn’t negligible. According to a report from the Open Energy Outlook initiative, the electricity demands from data centers and cryptocurrency mining could spike by a staggering 350% between 2020 and 2030. This surge would see their share of total U.S. consumption rise from 4% to 9%. Even Goldman Sachs has cautioned that the energy consumption of data centers could grow by 160% by the decade’s end.
Market Imbalance and Rising Costs
The ensuing pressure has already disrupted market dynamics. In December 2024, capacity prices in the PJM region—home to the world’s highest density of data centers—soared from $30 to $270 per MW-day in just one auction. This dramatic rise will inevitably impact the electricity bills of approximately 67 million consumers. John Ketchum, CEO of NextEra Energy, described these times as a “golden era of energy demand,” but also warned that “new electrons cannot reach the grid quickly enough.” As a result, natural gas is reappearing to fill the gap.
The 24/7 Challenge
Why can’t renewable energy suffice? While renewables are becoming more competitive, they cannot provide a constant energy supply that AI demands 24/7. Manuel Losa, from Pictet Asset Management, aptly noted that if firm energy is required continuously, “the only way to achieve this is with gas.” The technical issue lies in the “firmness” of renewables—solar and wind generation cannot reliably ensure constant supply without significant energy storage or robust transmission networks.
Solutions on the Horizon
As traditional electrical planning assumed a 1-2% growth annually, the surge linked to data centers is causing increases of 20-30% in some areas. The immediate answer appears to be ramping up gas-fired generation. However, even that faces hurdles. Gas turbine manufacturing has become a bottleneck, with delivery times extending up to seven years for new units.
Despite the pressing issues, renewables aren’t disappearing. Google has signed agreements for 1.2 gigawatts of new wind and solar energy in the U.S. Yet, even with such contracts, there’s no guarantee that hourly consumption will align with the availability of clean energy.
The Need for Change
To navigate these challenges, battery storage and upgrades to existing grid systems are critical, but they require years to deploy while AI’s growth is rapid. Companies historically focused on renewables, like NextEra, are now looking toward gas solutions for immediate relief. Meanwhile, Google has inked a deal to develop small modular reactors (SMRs), aiming for 500 MW emissions-free by 2030, though any new nuclear capacity won’t be ready for several years.
The Burden of Transition
Today’s situation reveals a clash between two transitions: the rapid advancement of digital technology versus the sluggish pace of energy transition. Five years ago, natural gas was viewed as a temporary solution; today, it supports AI expansion.
As the Open Energy Outlook initiative points out, the goal should not be to choose between digital progress and network stability. However, if energy planning doesn’t adapt swiftly—incorporating more storage, transmission, and improved market designs—the inevitable outcome may mean increased reliance on gas, escalating emissions, and soaring costs. AI promises significant efficiency and intelligent decarbonization, but for now, its relentless growth prolongs the need for fossil fuels.
In conclusion, while a digital future looms just ahead, the energy transition appears to be lagging considerably behind, allowing gas to burn brightly once again in the backdrop of technological progress.