AI and Energy Efficiency

AI’s appetite for energy is soaring, but the technology could become a key tool in improving operational efficiency and installation timeframes for clean energy projects.

As generative AI booms, so does its energy appetite. Research from BloombergNEF warns that data centre electricity demand could double by 2050, consuming nearly 9% of global power generation.

While renewables are expected to ramp up alongside, there’s growing concern that corporate priorities—driven by the rapid rise of AI, manufacturing reshoring, and industrial electrification—could skew more toward performance than sustainability.

Fossil fuel firms are already deploying AI to scale operations, and without careful governance, AI could reinforce private profits at the expense of public climate goals.

Speaking at the Innovation Zero conference in London in May, Emily Shuckburgh—director of Cambridge Zero, the University of Cambridge’s flagship climate change initiative—emphasises: “I am deeply concerned about the hubris within much of the AI community, which ignores energy consumption.

“AI data centres already consume a single-digit percentage of global electricity, and that figure is rising rapidly. If this trend continues, we have a major problem.”

Part of the same panel, Google DeepMind’s climate action lead Sims Witherspoon adds: “AI is electricity intensive. There’s no shying away from that.

“So, we need to be smart about when and where we run models. It’s important with any technology, especially one as powerful as AI, for there to be a constant evaluation of the risks and the rewards.

“The benefits need to always vastly outweigh the risks of using the technology.”

Witherspoon outlines a strategic three-part role for AI in climate action: understanding, optimising and accelerating.

Google DeepMind, for instance, has used AI to advance plasma physics, an essential step in achieving commercially viable fusion energy.

Fusion generates energy by fusing light atomic nuclei into heavier ones inside superhot plasma, which is held in place by magnetic coils to prevent it from touching and damaging the reactor walls. The AI developed by DeepMind controls all 19 magnetic coils in a fusion reactor using a single neural network, cutting research time significantly.

Witherspoon adds: “There’s a way to optimise AI to use the most efficient hardware and run models only when the environmental cost is justified.”

She elaborates that Google queues non-essential workloads to run when grids are greener, leveraging tools to shift processing in time and location for a lower carbon footprint.

Both Shuckburgh and Witherspoon emphasise that AI’s effectiveness hinges on three pillars: clearly defined problem statements, high-quality data and well-articulated performance benchmarks. Without these, AI may not be the right tool—or worse, a wasteful one.

Additionally, Shuckburgh stresses that the discussions around AI’s use must include ethical dimensions, particularly around data equity and inclusion: “If we’re looking at climate-related solutions in the developing world, ensuring fairness in how the AI model is generated—and the data it relies on—is critically important.

“You need that human-in-the-loop element to make sure we’re not exacerbating inequalities.”

With adoption rising, many companies are already deploying AI to enhance energy efficiency.

Another big tech giant heavily investing in AI—Amazon Web Services (AWS), in response to questions from edie, highlights that it’s embedding AI across its clean energy strategy to help accelerate the deployment of solar power.

One example is the Baldy Mesa Solar Plus Battery Storage Project, where AWS machine learning tools like Amazon SageMaker process up to 33 billion data points annually to predict energy prices and optimise battery dispatch. During last year’s California heatwave, the system adjusted in real-time to deliver carbon-free energy precisely when demand spiked.

AWS is also backing Maximo, an AI-powered robot created by AES to accelerate solar farm construction in Bellefield, California. Maximo uses AWS’s RoboMaker to simulate and refine robotics, helping slash installation times and costs by up to 50%. Moreover, the robot is built to handle extreme heat and poor lighting—conditions that would be unsafe for humans at the desert site.

“We view AI as an important technology that can help accelerate our progress toward a more sustainable future.

“So, we’re designing AI tools that help avoid waste, carbon emissions and reduce energy use across our operations,” an AWS spokesperson tells edie.

Microsoft, too, is betting on AI to solve both energy efficiency and regulatory challenges.

To help power its own AI systems sustainably, Microsoft is exploring nuclear energy—specifically small modular reactors (SMRs). These next-gen nuclear plants could meet the surging energy demands of Microsoft’s data centres.

But getting one approved is a mammoth regulatory task, with past applications running to 12,000 pages and costing $500m.

To cut down on bureaucracy and time, Microsoft is now training a generative AI model, in partnership with nonprofit Terra Praxis, to auto-generate regulatory paperwork. The tech giant is expecting the AI system to slash human hours by up to 90%.

Back in the UK, the newly formed National Energy System Operator (NESO) is also looking to AI for grid management.

Also speaking at the Innovation Zero conference, NESO’s head of delivery Sangeeta Agrawal explains that the technology is already being deployed in live pilots.

She says: “For us (NESO), making decisions faster in the moment is where AI comes in. The data exists; the tech exists—but it’s still mostly in the hands of a few skilled people. So, we’re working to democratise access.”

By October 2025, NESO aims to have AI-driven decision support tools in its control room under the Volta project, enabling smarter scheduling and transparency. A second initiative, Vanguard, will help planners model complex energy system scenarios with speed and accuracy. Both depend heavily on NESO’s growing Data & Analytics Platform (DAP).

Agrawal describes this as part of a three-pronged ambition: “First, becoming a leading adopter of AI; second, acting as a champion for AI across the GB energy sector; and third, turning our experience into a global blueprint for how AI can help decarbonise power systems.”

Earlier this year, the UK Government announced its plans to ‘turbocharge’ the development of the national AI sector, in line with its headline commitment to achieve clean power by 2030.

It has also established a dedicated AI Energy Council, jointly chaired by the Science and Energy Secretaries, to work with energy companies and examine how best to meet the energy demands of AI technology.

But all roads lead back to data—its quality, availability and governance.