Can AI’s climate benefits outweigh its costs?

The surging resource demands of artificial intelligence make headlines almost daily. The United Nations recently noted that by 2030, data centers could be using three times as much electricity as the combined annual demand of Pakistan, Bangladesh, and Nigeria — countries that are collectively home to more than 650 million people.

Many industry watchers worry that the greenhouse emissions from AI’s growing energy use will make climate change significantly worse. Others suggest that AI will help solve the climate crisis by finding answers to longstanding challenges — rapidly decarbonizing the economy, improving the reliability of electric grids, cutting the cost of renewables, capturing carbon, and more. As a New York Times headline asked, “Will A.I. Ruin the Planet or Save the Planet?”

MIT Sloan School of management professor John Sterman and Koru Labs president Jennifer Turliuk, SFMBA ’25, explore this question in “The Net Climate Impact of Artificial Intelligence: Balancing Current Costs With Future Climate Benefits.” The paper, which provides a framework for evaluating the climate impact of AI to sort out what various futures might look like, is forthcoming in a new book, “Sustainable Climate-Resilience: Incremental Pathways for Cities and Communities,” and available now as a preprint.

Sterman and colleagues have embedded that framework in En-ROADS, an interactive climate simulator developed by Climate Interactive and MIT Sloan. En-ROADS allows users to visualize how various policy and technology choices might shape the planet’s future — from the global temperature, sea level rise, and extreme weather to changes in energy consumption, global health, and GDP.

With En-ROADS’s new features, policymakers, corporate and financial leaders, advocates, educators, and the public at large can now explore the direct and indirect impacts of AI on energy use, greenhouse gas emissions, and climate change.

Sterman and Turliuk emphasized that their analyses focused only on AI’s climate impacts. “AI and data centers also generate significant environmental and social impacts beyond climate change,” Sterman said, citing surging electricity prices, water use, air pollution, and demand for critical metals, “along with a host of security, privacy, social justice, and ethical issues.”

“AI is a remarkable technology, both promising and problematic. But, by itself, it does not solve the climate crisis and these other problems,” said Sterman, co-faculty director of the MIT Sloan Sustainability Initiative. “We need sound policies to do that.”

The direct impacts of AI on climate change

Sterman and Turliuk reviewed the most current data around the climate impacts of AI and found that carbon emissions from AI and data centers today are responsible for roughly 0.5% of global CO₂ emissions from energy.

“AI and data center emissions are making climate change worse today, and they are growing rapidly,” Sterman said.

The direct impact of the greenhouse emissions from AI and data centers on the climate is expected to be on the order of a tenth of a degree Celsius by 2100, according to En-ROADS projections, and the indirect impacts could be much larger. Every tenth of a degree of warming worsens the harms from climate change and makes it harder to limit warming to the upper limit allowed under the Paris climate agreement: no more than 2° C (3.6° F).

Three factors limit the direct impact of AI on climate:

• Emissions from AI and data centers are still small compared with those of the total economy, including emissions from transportation, buildings, industry, agriculture, and other sources, such as deforestation and methane leakage from the fossil fuel supply chain.
• Renewable energy is growing rapidly, helping to decarbonize electricity production across the economy, including power for AI and data centers.
• AI and data centers are becoming more efficient, thanks to better algorithms, chips, and data center design.

Yet, greater efficiency also cuts the other way: The more efficient AI and data centers become, the lower the cost of training and using AI, which will stimulate greater AI use and offset some of the benefits of those efficiency gains. That feedback is known as the direct rebound effect.

Using the En-ROADS simulator, Sterman demonstrated that even faster reductions in the energy intensity of AI would lead to a negligible reduction in global warming by 2100, assuming that no new climate policies are implemented.

The indirect impacts of AI on climate change

AI is widely expected to boost productivity and speed economic growth. But faster economic growth means more energy demand and more greenhouse emissions, worsening climate change. This is the indirect rebound effect of AI.

Forecasts vary widely, with some arguing that AI will increase economic growth modestly while others say it will accelerate dramatically. En-ROADS allows users to explore how AI’s effect on productivity and economic growth could affect the climate.

For example, assuming that AI can eventually boost gross world product by 25% — a figure that’s well below other, more optimistic projections — the additional production of goods and services would raise energy demand and emissions significantly, increasing expected warming from 3.3° C (about 5.9° F) by 2100 to 3.6° C (6.5° F).

This increase would dramatically worsen sea level rise, wildfires, extreme weather, declines in crop yields, and other damage from climate change — damage that feeds back to harm our health, security, and prosperity, Sterman said.

He used electric vehicles to illustrate the problem of indirect rebound. An EV can save its owner several thousand dollars per year in lower gas and maintenance costs.

“When I ask executives what they would do with those extra thousands,” Sterman said, “they often say ‘buy more stuff’ and ‘take my family to Disney World.’ But the emissions from that trip to Orlando can outweigh the reduction from driving your EV, increasing your carbon footprint and worsening climate change.”

“Indirect rebound is the big issue,” Sterman said. “Individuals, corporations, and governments all want faster growth in their income, sales, and economy,” and AI may help achieve those goals. “But unless we implement policies and actions that rapidly lower emissions across the economy as a whole, then any economic boost from AI will generate a lot more emissions, and the climate will get notably worse,” he said.

Won’t AI help cut emissions everywhere?

Proponents argue that AI will speed innovations that can help cut emissions, including emissions-free energy, efficient buildings, low-carbon production of cement, steel, and fertilizers, practical carbon-capture technologies, and more.

If AI can do this, the thinking goes, then perhaps the economy can grow faster while emissions fall.

The problem with that line of thinking is one of focus and timing, Sterman and Turliuk note.

• Focus: Today, the vast majority of AI isn’t used to solve climate change and other critical environmental and social problems.
• Timing: AI might lead to new innovations that cut emissions, but they will take time. The delay isn’t merely from how long it takes for AI to help us invent new and better technologies, but from the time required to go from the lab to pilot projects, raise capital, win community acceptance, and build the organizations, production capacity, and supply chains needed to scale globally.

To illustrate, Sterman tested a scenario in which novel ideas generated by AI were able to cut the cost of new nuclear power plants in half. The climate benefits, however, were negligible: Nuclear power did grow far more rapidly, squeezing out many coal and gas plants, but it also displaced the deployment of renewables, substituting one clean energy source for another.

At some point in the future, AI-enabled innovations might lower AI’s direct and indirect emissions enough to offset earlier emissions. “But even if emissions eventually fall back to what they would have been, sea levels won’t fall back,” Sterman said. “The forests burned, the crops lost, and all the businesses and homes destroyed along the way by the extra warming from AI won’t reappear.”

Action item: Enact and support strong climate policies

Both problems can be addressed with strong climate policies — carbon pricing, together with incentives to speed energy efficiency and decarbonization across the economy, end deforestation, and more. These policies would cut emissions today while steering more AI research toward climate solutions.

Turliuk pointed to corporate disclosure as an immediate lever. “Many large corporations are no longer on track for their net-zero targets, and some have stated that this is due to AI,” she said.

Governments at all levels should require disclosure of AI’s environmental impacts, ensure companies pay the full, system-wide costs of data centers, and prevent AI greenwashing. Together with carbon pricing and incentives for efficiency and decarbonization, these steps will reduce emissions and other environmental damage from AI — helping to limit the fires, flooding, crop failures, and supply chain disruptions that threaten businesses.

Business leaders, policymakers, and educators can use En-ROADS to explore AI’s direct and indirect climate impacts, test their own assumptions, and build scenarios for limiting global warming even as AI transforms their organizations.

To do so, interested leaders can join a free webinar addressing the climate impact of AI with John Sterman and Climate Interactive on Aug. 27, 2026, and explore En-ROADS for themselves.

John Sterman is a professor of management at the MIT Sloan School of Management and a professor in the MIT Institute for Data, Systems, and Society. He is also the director of the MIT Sloan System Dynamics group and the MIT Sloan Sustainability Initiative. Sterman’s research centers on improving decision-making in complex systems, including corporate strategy and operations, energy policy, public health, environmental sustainability, and climate change.

Jennifer Turliuk, SFMBA ’25, focuses on reducing climate change and AI harms through advocacy, research, innovation, and education. She is president of Koru Labs, which helps mission-driven organizations and leaders turn climate and AI challenges into measurable, ethical impact. Turliuk has lectured about climate and AI through appointments as a short-term lecturer at MIT, an executive fellow at Harvard Business School, and preceptor of the Google Data Center Community AI Fellowship.