AI Has an Energy Problem: Can Renewables, And AI Itself, Be the Solution?

Electricity demand from data centers jumped 17% in 2025, while electricity consumption grew even faster, surging 50%, according to the International Energy Agency (IEA). The agency now expects global data-center electricity consumption to roughly double to approximately 950 TWh by 2030.

That would put data centers at around 3% of global electricity consumption by the end of the decade. Even if the number may sound small at a global level, the local impact is considerably larger.

AI-focused data centers can require hundreds of megawatts of power, while the largest facilities being planned or built can reach gigawatt scale. The U.S. Department of Energy has noted that hyperscale projects seeking power capacities between 300 MW and 1,000 MW or more are putting additional pressure on local grids, particularly as utilities may have only one to three years to meet those connection demands.

But the technologies being developed alongside the AI boom could also help address some of those pressures, from integrating more renewable power and optimizing energy storage to making increasingly complex grids more efficient. The question is whether those two sides can develop at the same speed.

AI's electricity demand is rising fast

Data centers are not synonymous with AI. They also run cloud computing, storage, networking and conventional digital services. But AI is increasingly becoming one of the most important drivers of new electricity demand.

Electricity consumption from AI-focused data centers is expected to triple by 2030, according to the IEA. That growth is occurring despite rapid improvements in the efficiency of AI hardware and software.

While power consumption per AI task is declining rapidly, more people are using it, and for increasingly energy-intensive applications such as video generation, reasoning and agentic AI. Some of those applications can consume hundreds or thousands of times more energy per query than simple text generation. As this demand spikes, so does the technology's carbon footprint.

To mitigate this growing environmental impact on an organizational level, proactive governance is becoming essential.

"One thing leaders can do today on the environmental front is put AI governance in place at their organizations," Ben Szuhaj, Senior AI Strategist at KUNGFU.AI, an AI-focused management consulting and engineering firm, told International Business Times.

"Good AI governance is, itself, an environmental best practice because it cuts down on wasteful use and steers adoption toward applications of AI that are more likely to justify the environmental costs."

The physical infrastructure required to support that growth is under increasing pressure too. The IEA's 2025 Energy and AI report found that a conventional data center might have a power demand of around 10 MW to 25 MW, while a hyperscale AI-focused facility can reach 100 MW or more.

For electricity grids, the issue is where that electricity is required, and how quickly. Data centers tend to be highly concentrated geographically, meaning their impact on local electricity systems can be much greater than their global share of consumption suggests.

The U.S. is particularly exposed. The IEA expects data centers to account for nearly half of the country's electricity-demand growth through 2030. The rapid growth of AI applications is a major factor behind rising data-center power requirements, according to the U.S. Department of Energy's Berkeley Lab, which warns these will drive more than 40% of U.S. electricity demand growth through the end of the decade.

Experts warn infrastructure is struggling to keep up with surging energy demand.

"The timing mismatch is real," explains Michael Grasso, CEO of GridRails, a management platform used to coordinate and settle distributed energy resources. "AI-related electricity demand can increase within a few years, while new generation, transmission, storage and other major infrastructure projects often take much longer to plan, permit and build".

This disparity raises serious environmental concerns for the immediate future.

Renewable energy can meet part of the demand – but not automatically

The rapid expansion of renewable power infrastructure could soon supply a substantial portion of the electricity required to fuel global data centers. According to recent data from the IEA's comprehensive energy and technology reviews, renewable sources – chiefly solar, wind, and hydropower – are on track to fulfill nearly half of the global growth in data center power demand through 2030.

In their 2025 Energy and AI report, the IEA noted that clean energy accounted for roughly 27% of the total electricity consumed by these facilities worldwide, while coal generated approximately 30%, followed closely by natural gas at 26% and nuclear power at 15%.

"Hyperscalers are investing heavily in clean energy generation and storage, thus accelerating the deployment of critical clean energy infrastructure", noted Szuhaj.

He expects this pattern to persist, noting it is "the fastest way to secure the electricity they need at predictable prices."

The massive scale-up of green energy is moving quickly. However, expanding the production and use of renewables does not automatically guarantee the continuous, 24/7 power supply required by modern data infrastructure.

Clean energy can't scale in a vacuum, meaning low-emission firm power like nuclear and geothermal, alongside transitionary natural gas, will remain integrated into the near-term supply mix, according to the IEA.

AI as part of the solution

Artificial intelligence holds significant promise as a mechanism for optimizing the broader electrical ecosystem, moving beyond its role as a resource consumer.

Existing AI applications, if widely adopted, could reduce data center emissions by three or four times, according to estimates from the IEA. Systems like methane leak detection, faster grid fault detection, and industrial efficiency gains could all contribute to this.

AI-driven models can also improve the reliability of renewable energy sources.

Because wind and solar generation are inherently intermittent, grid operators face growing pressure to predict supply availability with precision. By analyzing complex meteorological datasets, historic asset behavior, and real-time grid inputs, AI can refine these generation forecasts.

"AI has the potential to make the energy system itself much more efficient. AI can be used to better predict how much electricity renewable sources will produce and when," said Markus Levin, co-founder of XYO, a decentralized data network for AI and robotics.

For instance, these systems can analyze weather forecasts to predict a solar farm's daily output while automatically managing power shortages or grid surpluses.

The IEA calculates that widespread integration of machine learning systems into power plant operation and maintenance routines could yield up to $110 billion in annual operational savings by 2035, primarily by reducing fuel waste and lowering baseline overhead.

However, technology alone can't solve the problem unless it is paired with proper infrastructural integration.

"AI can help reduce emissions and improve energy-system efficiency, but only if we treat it as both a new source of electricity demand and integrate it into tools for managing that demand," Grasso highlighted. Specifically, the technology can optimize the grid by dynamically shifting energy consumption away from high-stress, peak demand periods.

Given that building physical high-voltage lines typically requires years of regulatory approvals and construction, deploying software-driven optimizations may be able to alleviate immediate grid constraints while data center demand continues to scale.

AI's energy problem is becoming an energy-system problem

AI can make energy systems more efficient, but it also faces an efficiency paradox: as the amount of energy needed for each computation falls, total energy use can still rise because AI is being adopted faster and used for increasingly demanding applications.

Efficiency gains alone, therefore, may not be enough to offset growing demand. AI's wider use in the energy system also faces practical barriers, including limited access to high-quality data and a shortage of the digital skills needed to deploy these tools at scale.

New renewable generation is of limited use if transmission capacity cannot deliver the power to where it is needed, while batteries cannot provide continuous supply through prolonged periods of low renewable output.

"What we can do is think beyond simply building more giant data centers and waiting for new power infrastructure to catch up," noted Levin.

He cited the example of Bitcoin, where computing was distributed across independently operated sites rather than concentrated in one place.

With AI's electricity demand expected to keep growing, the need for new generation, storage and grid capacity is increasingly urgent. Renewable energy can provide a significant part of that additional power, while AI itself could help utilities manage demand and make renewable generation more predictable.

But neither can address the problem on its own. As data centers continue to expand, the challenge will be keeping electricity supply, grid infrastructure and demand growth moving at a pace that can support them all.