AI companies eye fossil fuels to meet booming energy demand

data center dark hallway green shade fluorescent light
Energy-intensive data centers were responsible for an estimated 4% of the US’ overall energy use in 2022, according to the International Energy Agency.

It takes massive amounts of energy to power the data center brains of popular artificial intelligence models. That demand is only growing. In 2024, many of Silicon Valley’s largest tech giants and hoards of budding, well-funded startups have (very publically) aligned themselves with climate action–awash with PR about their sustainability goals, their carbon neutral pledges, and their promises to prioritize recycled materials. But as AI’s intensive energy demands become more apparent, it seems like many of those supposed green priorities could be jeopardized.

A March International Energy Agency forecast estimates input-hungry AI models and cryptocurrency mining combined could cause data centers worldwide to double their energy use in just two years. Recent reports suggest tech leaders interested in staying relevant in the booming AI race may consider turning to old-fashioned, carbon-emitting energy sources to help meet that demand.

AI models need more energy to power data centers 

Though precise figures measuring AI’s energy consumption remain a matter of debate, it's increasingly clear complex data centers required to train and power those systems are energy-intensive. A recently released peer reviewed data analysis, energy demands from AI servers in 2027 could be on par with those of Argentina, the Netherlands, or Sweden combined. Production of new data centers isn’t slowing down either. Just last week, Washington Square Journal reports, Amazon Web Service Vice President of Engineering Bill Vass told an audience at an energy industry event in Texas he believes a new data center is being built every three days. Other energy industry leaders speaking at the event, like Former U.S. Energy Secretary Ernest Moniz, argued renewable energy production may fall short of what is  needed to power this projected data center growth.

“We’re not going to build 100 gigawatts of new renewables in a few years,” Moniz said. The Obama-era energy secretary went on to say unmet energy demands brought on by AI, primarily via electricity, would require tapping into more natural gas and coal power plants. When it comes to meeting energy demands with renewables, he said, “you’re kind of stuck.”

Others, like Dominion Energy CEO Robert Blue say the increased energy demand has led them to build out a new gas power plant while also trying to meet a 2050 net-zero goal. Other natural gas company executives speaking with the Journal, meanwhile claim tech firms building out data setters have expressed interest in using a natural gas energy source.

Tech companies already have a checkered record on sustainability promises

A sudden reinterest in non-renewable energy sources to fuel an AI boom could contradict net zero carbon timelines and sustainability pledges made by major tech companies in recent years. Microsoft and Google, who are locked in a battle over quickly evolving generative AI tools like ChatGPT and Gemini, have both outlined plans to have net negative emissions in coming years. Apple, which reportedly shuttered its long-running car unit in order to devote resources towards AI, aims to become carbon neutral across its global supply chains by 2030. The Biden administration meanwhile has ambitiously pledged the US to have a carbon pollution free electricity sector by 2035.

[ Related: Dozens of companies with ‘net-zero’ goals just got called out for greenwashing ]

Critics argue some of these climate pledges, particularly those heralded by large tech firms, may seem impressive on paper but have already fallen short in key areas. Multiple independent monitors in recent years have criticized large tech companies for allegedly failing to properly disclose their greenhouse gas emissions. Others have dinged tech firms for heavily basing their sustainability strategies around carbon offsets as opposed to potentially more effective solutions like reducing energy consumption. The alluring race for AI dominance risks stretching those already strained goals even further.

AI boom has led to new data centers popping up around the US

Appetites for electricity are rising around the country. In Georgia, according to a recent Washington Post report, expected energy production within the state in the next ten years is 17 times larger than what it was recently. Northern Virginia, according to the same report, could require the energy equivalent of several nuclear power plants to meet the increased demand from planned data centers currently under construction. New data centers have popped up in both of those states in recent years. Lobbyists representing traditional coal and gas energy providers, the Post claims, are simultaneously urging government offices to delay retiring some fossil fuel plants in order to meet increasing energy demands. Data centers in the US alone were responsible for 4% of the county’s overall energy use in 2022 according to the IEA. That figure will only grow as more and more AI-focused facilities come online.

At the same time, some of the AI industry’s-starkest proponents have argued these very same energy intensive models may prove instrumental in helping scale-up renewable energy sources and develop technologies to counteract the most destructive aspects of climate change. Previous reports argue powerful AI models could improve the efficiency of oils and gas facilities by improving underground mapping. AI simulation modes, similarly could help engineers develop optimal designs for wind or solar plants that could bring down their cost and increase their desirability as an energy source. Microsoft, who partners with OpenAI, is reportedly already using generative AI tools to try and streamline the regulatory approval process for nuclear reactors. Those future reactors, in theory, would then be used to generate the electricity needed to quench its AI models' energy thirst.

Fossil-fuel powered AI prioritizes long-term optimism over current day climate realities 

The problem with those more optimistic outlooks is that they remain, for the time being at least, mostly hypothetical and severely lacking in real-word data. AI models may increase the efficiency and affordability of renewable resources long term, but they risk doing so by pushing down on the accelerator of non-renewable resources right now. And with energy demands surging in other industries outside of tech at the same time, these optimistic longer-term outlooks could serve to justify splurging on natural gas and goal in the short term. Underpinning all of this is a worsening climate outlook that the overwhelming majority of climate scientists and international organizations agree demands radical action to reduce emissions as soon as possible. Renewable energy sources are on the rise in the US but tech firms looking for easier available sources of electricity to power their next AI projects risk setting back that progress.