AI data centers—with their massive energy and water needs—have become a lightning rod for criticism from climate change advocates. But could the technology also serve as a powerful mitigation and adaptation tool?
We asked affiliates of Harvard Business School’s Business and Environment Initiative how artificial intelligence could help manage the effects of climate change and what business opportunities might emerge. Here’s what they said.
Christian Kaps: Managing data center demand volatility
For most of the history of electricity grids, supply followed demand. Power plant operators coordinated their output to match users’ fluctuating consumption every hour of every day.
As such, decarbonizing electricity over the past two decades was almost entirely focused on the supply side through building renewable generation like solar and wind. But because these sources are intermittent and uncontrollable, their growing share of generation capacity has created integration challenges. Despite significant investment in transmission infrastructure and battery storage, solar, in particular, causes volatile prices.
Meanwhile, demand-side tools remain surprisingly underused and relatively unsophisticated—during scarcity events, grid operators may, for example, resort to mass text messages asking citizens to stop charging EVs.
AI data centers may change this paradigm. Their enormous power requirements are driving the first significant rise in electricity demand across industrialized nations in decades. And because data centers can be built faster than power plants, this growth risks energy scarcity, higher bills, and reliability issues.
Crucially, however, most data centers are built and operated by large firms—some, like Amazon Web Services, Google, and Microsoft, running thousands globally. Both LLM training and inference can plausibly occur anywhere across these networks, providing responses that can reach users within split seconds.
This means operators could, if willing and able, use daily energy data to train LLMs to predict where renewable supply is abundant, enabling them to shift compute power. Demand would follow supply rather than the reverse.
The combined pressure on data center operators to expand compute power and on utilities to deliver more power may have created the right conditions for the new contracting frameworks and coordinating infrastructure to be designed and built that large-scale demand response requires.
If realized, this would allow cleaner electricity because more renewable volatility could be used to power a larger share of total consumption and establish a template that electric vehicles, heat pumps, and other electrification technologies could follow—charging and running when clean energy is plentiful.
Christian Kaps is an assistant professor of business administration in the Technology and Operations Management Unit.
Vikram Gandhi: Speeding decarbonization, improving agriculture
The technology behind AI can certainly help address the negative externalities of climate change, from mitigation to adaptation. And we're already seeing some of that. Here are five key opportunities that I see:
Unifying strategic data. Part of the problem right now is that policy planning, resource allocation, and capital distribution are currently done in isolation in countries around the world. AI could be a fantastic platform for integrating that information and bringing those data streams together to develop better solutions.
Calibrating energy. There’s a massive opportunity for AI to optimize energy grids and increase efficiency by matching demand and supply more effectively. As more electricity is needed, the AI opportunities grow by the day.
Improving agriculture. The issue of water and its scarcity isn’t appreciated as it should be. If you look at food, there are huge opportunities for AI to improve agricultural practices, weather forecasting, yield management, and water use.
Speeding decarbonization. If you look at the key areas of innovation necessary to move from a carbon-based economy to a non-carbon-based one, we'll need to:
Develop storage at scale and at reasonable cost
Expand carbon capture and sequestration
Develop alternative fuels
AI can help bring that technology to market faster. Digitization accelerated during COVID. Without the pandemic, it would have taken years for those technologies to develop and behavioral changes to take hold. Similarly, AI adoption can help us reach decarbonization solutions faster.
Helping cities and countries adapt. AI could help governments improve weather forecasting and implement adaptation strategies to deal with extreme climate events such as floods, heat waves, droughts, and wildfires.
In addition, AI, done correctly, could help the developing world—which needs to double its infrastructure, food supply, mobility, and grid capacity in the not-too-distant future—grow in a lower-emissions trajectory.
Vikram Gandhi is the Gerald P. Kaminsky Senior Lecturer of Business Administration.
Jennifer Turliuk: Think net-impact with AI
Without a doubt, AI’s immense energy and resource needs are contributing to climate change. However, the technology could potentially help reduce climate change through applications such as optimizing electricity use, creating novel innovations, and building more efficient systems.
Finding the middle ground will challenge leaders to think differently. Currently, they will need to navigate these tensions without robust legislation and other guardrails to guide them.
In an ideal world, policymakers would implement carbon pricing, require disclosure of AI use and impacts, regulate AI-related emissions and resource use, and support research and development, training, and greening the grid. Until then, here’s what leaders can do now:
Normalize net-impact thinking. Leaders will need to think more holistically about initiatives and products and look for projects and pathways where climate benefits exceed harms. For example, perhaps new capital investments should consider factors such as:
Installed base of chips. The number of existing inefficient semiconductors that will require energy to power.
Time value of carbon. Measuring the long-term damage of the emissions that a project creates today.
Rebound effects. How to conserve energy and resources even as new solutions emerge.
Load flexibility. How can a project use only the electricity it needs? How could AI help with this?
Could your organization offset the resource costs of AI use by embracing solar, wind, and other renewable sources more broadly? Are there opportunities to for your infrastructure, software, or hardware to use less energy?
Apply AI only where it’s most beneficial. The climate harms of AI are real and measurable, while the potential climate benefits of AI are speculative. Until regulations can set parameters around AI use, businesses should be ethical stewards of “climate-responsible AI.”
Does every customer interaction need a chatbot? Does every operational element need to become “agentic”? Probably not.
Look for opportunities that leverage AI capabilities to provide the most business value—ideally with positive climate and societal impact.
Jennifer Turliuk is an executive fellow at HBS and president of Koru Labs.
Image by Ariana Cohen-Halberstam with assets from Adobe Stock.
