Rethinking power: how AI is reshaping energy demands in data centers
Today, artificial intelligence is revolutionizing virtually every industry, but its rapid adoption also comes with a significant challenge: energy consumption.
Data centers are racing to accommodate the surge in AI-driven demand and are consuming significant amounts of electricity to support High-Performance Computing, cloud computing services, and the many digital products and services we rely on every day.
Why are we seeing such a spike in energy use? One reason is heavy reliance on graphics processing unit (GPU) chips, which are much faster and more effective than processing tasks. More than just an advantage, this efficiency has now made GPUs the new standard for training and running AI models and workloads.
Yet it also comes at a high cost: soaring energy consumption. Each GPU now requires up to four times more electricity than a standard CPU, an exponential increase that is quickly – and dramatically – changing demands for energy in the data center.
Head of Sustainability, DataBank.
For example, consider these recent findings:
The New York Times recently described how OpenAI hopes to build five new data centers that would consume more electricity than the three million households in Massachusetts.
According to the Center on Global Energy Policy, GPUs and their servers could make up as much as 27 percent of the planned new generation capacity for 2027 and 14 percent of total commercial energy needs that year.
A Forbes article predicted that Nvidia’s Blackwell chipset will boost power consumption even further – a 300% increase in power consumption across one generation of GPUs with AI systems increasing power consumption at a higher rate.
These findings raise important power-related questions: Is AI growth outpacing the ability of utilities to supply the required energy? Are there other energy options data centers should consider? And maybe most importantly, what will data center’s energy use look like in both the short- and long-term future?
Navigating Power Supply and Demand in the AI Era
Despite growing concerns, AI has not yet surpassed the grid’s capabilities. In fact, some advancements suggest that AI energy consumption could even decreases. Many AI companies expended vast amounts of processing power to train their initial models, but newer players like DeepSeek now claim that their systems operate far more efficiently, requiring less computing power and energy.
However, AI’s sudden rise is only one factor in a perfect storm of energy demands. For example, the larger electrification movement, which has introduced millions of electric vehicles to the grid, and the reshoring of manufacturing to the U.S., is also straining resources. AI adds another layer to this complex equation, raising urgent questions about whether existing utilities can keep pace with demand.
Data centers, as commercial real estate, are also subject to the age-old adage, “location, location, location.” Many power generation sites – especially those harnessing solar and wind – are located in rural places in the United States, but transmission bottlenecks make it difficult to move. That power to urban centers where demand is highest. Thus far, geodiversity and urban demand have not yet driven data centers to these remote areas.
This could soon change. Hyperscalers have already demonstrated their willingness and agility in building data centers in the Arctic Circle to take advantage of natural cooling to reduce energy use and costs. A similar shift may take hold in the U.S., with data center operators eyeing locations in New Mexico, rural Texas, Wyoming, and other rural markets to capitalize on similar benefits.
Exploring Alternative Energy Solutions
As strain on the grid intensifies, alternative energy solutions are gaining traction as a means of ensuring a stable and sustainable power supply.
One promising development is the evolution of battery technology. Aluminum-ion batteries, for example, offer several advantages over lithium-based alternatives. Aluminum is more abundant, sourced from conflict-free regions, and free from the geopolitical challenges associated with lithium and cobalt mining. These batteries also boast a solid-state design, reducing flammability risks, and their higher energy density enables more efficient storage, which helps smooth out fluctuations in energy supply and demand – often visualized as the daily “duck curve.”
Nuclear energy is also re-emerging as a viable solution for long-term, reliable power generation. Advanced small modular reactors (SMRs) offer a scalable, low-carbon alternative that can provide consistent energy without the intermittency of renewables.
However, while test sites are under development, SMRs have yet to begin generating power and may still be five or more years away from large-scale deployment. Public perception remains a key challenge, as strict regulations often require plants to be situated far from populated areas, and the long-term management of nuclear waste continues to be a concern.
Additionally, virtual power plants (VPPs) are revolutionizing the energy landscape by connecting and coordinating thousands of decentralized batteries to function as a unified power source. By optimizing the generation, storage, and distribution of renewable energy, VPPs enhance grid stability and efficiency. Unlike traditional power plants, VPPs do not rely on a single energy source or location, making them inherently more flexible and resilient.
Securing a Sustainable Power Future for AI and Data Centers
While it’s hard to predict what lies ahead for AI and how much more demand we’ll see, the pressure is on to secure reliable, sustainable power, now and into the future.
As the adoption of AI tools accelerates, data centers must proactively seek sustainable and resilient energy solutions. Embracing alternative power sources, modernizing grid infrastructure, and leveraging cutting-edge innovations will be critical in ensuring that the power needs of AI-driven industries can be met – now and in the years to come.
We list the best web hosting services.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
Today, artificial intelligence is revolutionizing virtually every industry, but its rapid adoption also comes with a significant challenge: energy consumption. Data centers are racing to accommodate the surge in AI-driven demand and are consuming significant amounts of electricity to support High-Performance Computing, cloud computing services, and the many digital products…
