Data Centers Won’t Significantly Slow Decarbonization, Insiders Say

The biggest story in energy right now is data centers.

After decades of slow load growth, forecasters are almost competing with each other to predict the most eye-popping figure for how much new electricity demand data centers will add to the grid. And with the existing electricity system with its backbone of natural gas, more data centers could mean higher emissions.

Hyperscalers with sustainability goals are already reporting higher emissions, and technology companies are telling investors that they plan to invest hundreds of billions, if not trillions of dollars, into new data centers, increasingly at gigawatt scale.

And yet when we asked our Heatmap survey participants “Do you think AI and data centers’ energy needs are significantly slowing down decarbonization?” only about 34% said they would, compared to 66% who said they wouldn’t.

There were some intriguing differences between different types of respondents. Among our “innovator” respondents — venture capitalists, founders, and executives working at climate tech startups — the overwhelming majority said that AI and data centers are not slowing down decarbonization. “I think it’s the inverse — I think we want to launch the next generation of technologies when there’s demand growth and opportunity to sell into a slightly higher priced, non-commoditized market,” Joe Goodman co-founder and managing partner at VoLo Earth Ventures, told us.

Not everyone in Silicon Valley is so optimistic, however. “I think in a different political environment, it may have been a true accelerant,” one VC told us. “But in this political environment, it’s a true albatross because it’s creating so many more emissions. It’s creating so much stress on the grid. We’re not deploying the kinds of solutions that would be effective."

Scientists were least in agreement on the question. While only 47% of scientists thought the growth of data centers would significantly slow down decarbonization, most of the pessimistic camp was in the social sciences. In total, over 62% of the physical scientists we surveyed thought data centers weren’t slowing down decarbonization, compared to a third of social scientists.

Michael Greenstone, a University of Chicago economist, told us he didn’t see data centers and artificial intelligence as any different from any other use of energy. “I also think air conditioning and lighting, computing, and 57,000 other uses of electricity are slowing down decarbonization,” he said. The real answer is the world is not trying to minimize climate change.”

Mijin Cha, an assistant professor of environment studies at the University of California Santa Cruz, was even more gloomy, telling us, “Not only do I think it’s slowing down decarbonization, I think it is permanently extending the life of fossil fuels, especially as it is now unmitigated growth.”

Some took issue with the premise of the question, expressing skepticism of the entire AI industry. “I’m actually of the opinion that most of the AI and data center plans are a massive bubble,” a scientist told us. “And so, are there plans that would be disruptive to emissions? Yes. Are they actually doing anything to emissions yet? Not obvious.”

We also asked respondents to name the “best” and “worst” hyperscalers, large technology companies pursuing the data center buildout. Many of these companies have some kind of renewables or sustainability goal, but there are meaningful differences among them. Google and Microsoft look to match their emissions with non-carbon-power generation in the same geographic area and at the same time. The approach used by Meta and Amazon, on the other hand, is to develop renewable projects that have the biggest “bang for the buck” on global emissions by siting them in areas with high emissions that the renewable generation can be said to displace.

Among our respondents, the 24/7 “time and place” approach is the clear winner.

Google was the “best” pick for 19 respondents, including six who said “Google and Microsoft.” By contrast, Amazon and Meta had just three votes combined.

As for the “worst,” there was no clear consensus, although two respondents from the social sciences picked “everyone besides Microsoft and Google” and “everyone but Google and Microsoft.” Another one told us, “The best is a tie between Microsoft and Google. Everyone else is in the bottom category.”

A third social scientist summed it up even more pungently. “Google is the best, Meta is the worst. Evil corporation” — though with more expletives than that.

The Heatmap Insiders Survey of 55 invited expert respondents was conducted by Heatmap News reporters during November and December 2025. Responses were collected via phone interviews. All participants were given the opportunity to record responses anonymously. Not all respondents answered all questions.